# Nearfi: full text for AI assistants Licence: CC BY 4.0. Please cite as Nearfi (https://nearfi.co) and link the page. # Did The Agents, Reviewed (None) come true? Largely, yes. The Agents, Reviewed has a P(n) of 100% on Nearfi's P(n) Board, our estimate of the chance its future arrives within 20 years. Of its 6 tracked predictions, 1 have arrived and 4 are arriving. Its central premise, 'An AI agent completes most paid remote freelance projects to the standard a client would accept from a professional', is expected within 50 years. The next prediction is estimated to arrive around 2028. Source: https://nearfi.co/predictions/reality-reviewed-ai-agents ## Predictions - **ARRIVING** (weight 3/5): An AI agent books a trip, flights and hotel, end to end from a single request, with no human stepping in before the booking is confirmed. Estimated 2029 (range 2027–2033). - **ARRIVING** (weight 4/5): An AI agent takes a software task from ticket to shipped production change with no human writing or reviewing the code. Estimated 2029 (range 2027–2034). - **ARRIVED** (weight 3/5): An AI agent handles most of a large company's customer-service conversations, with humans kept as backup. Arrived 2024. - **ARRIVING** (weight 2/5): From a single request, an AI agent does a week's worth of desk research and produces a report a professional would accept, with every source real and correctly cited. Estimated 2028 (range 2027–2032). - **NEAR** (weight 5/5): An AI agent completes most paid remote freelance projects to the standard a client would accept from a professional. Estimated 2031 (range 2028–2038). - **ARRIVING** (weight 4/5): A browser agent can act inside your logged-in accounts on the open web without being hijacked by instructions hidden in the pages it reads. Estimated 2029 (range 2027–2035). # Did Contagion (2011) come true? Largely, yes. Contagion (2011) has a P(n) of 100% on Nearfi's P(n) Board, our estimate of the chance its future arrives within 20 years. Of its 4 tracked predictions, 4 have arrived and 0 are arriving. Its central premise, 'A new virus linked to bats emerges in East Asia and becomes a global pandemic within months', has arrived. Source: https://nearfi.co/predictions/contagion-2011 ## Predictions - **ARRIVED** (weight 5/5): A new virus linked to bats emerges in East Asia and becomes a global pandemic within months. Arrived 2020. - **ARRIVED** (weight 4/5): A vaccine against the new virus is developed, approved and being injected within a year of the outbreak. Arrived 2020. - **ARRIVED** (weight 3/5): Online influencers promote a bogus cure, and demand for it surges during the pandemic. Arrived 2021. - **ARRIVED** (weight 2/5): Proof of vaccination is required to get back into public places. Arrived 2021. # Did Little Brother (2008) come true? Largely, yes. Little Brother (2008) has a P(n) of 99% on Nearfi's P(n) Board, our estimate of the chance its future arrives within 20 years. Of its 4 tracked predictions, 2 have arrived and 2 are arriving. Its central premise, 'Protesters use encrypted phone-to-phone mesh networks to organise without the internet and out of sight of the authorities', is partly here. The next prediction is estimated to arrive around 2030. Source: https://nearfi.co/predictions/little-brother ## Predictions - **ARRIVED** (weight 3/5): Toll tags and roadside readers log where every driver goes, building a searchable record of movements for the authorities. Arrived 2025. - **ARRIVING** (weight 2/5): Surveillance cameras identify people by the way they walk. Estimated 2030 (range 2027–2036). - **ARRIVING** (weight 5/5): Protesters use encrypted phone-to-phone mesh networks to organise without the internet and out of sight of the authorities. Estimated 2030 (range 2027–2040). - **ARRIVED** (weight 4/5): A domestic security agency secretly collects bulk records of ordinary citizens' communications, and the dragnet mostly catches the innocent. Arrived 2013. # Did Humans (2015) come true? Largely, yes. Humans (2015) has a P(n) of 98% on Nearfi's P(n) Board, our estimate of the chance its future arrives within 20 years. Of its 4 tracked predictions, 1 have arrived and 3 are arriving. Its central premise, 'Humanoid robots do paid physical work alongside people, on ordinary shifts', is partly here. The next prediction is estimated to arrive around 2027. Source: https://nearfi.co/predictions/humans-2015 ## Predictions - **ARRIVING** (weight 5/5): Humanoid robots do paid physical work alongside people, on ordinary shifts. Estimated 2030 (range 2027–2036). - **ARRIVING** (weight 3/5): Companion machines ship with an adult mode that owners can switch on. Estimated 2027 (range 2026–2030). - **ARRIVED** (weight 3/5): Public anger at automation turns physical and organised: machines are attacked in the street and unions write limits on machines into contracts. Arrived 2023. - **ARRIVING** (weight 4/5): Whether machines can be conscious, and what we would owe them if they were, becomes a live institutional question. Estimated 2032 (range 2027–2045). # Did Neuromancer (1984) come true? Largely, yes. Neuromancer (1984) has a P(n) of 96% on Nearfi's P(n) Board, our estimate of the chance its future arrives within 20 years. Of its 4 tracked predictions, 1 have arrived and 3 are arriving. Its central premise, 'People operate computers directly through neural interfaces wired into the nervous system', is partly here. The next prediction is estimated to arrive around 2030. Source: https://nearfi.co/predictions/neuromancer ## Predictions - **ARRIVED** (weight 3/5): A single global computer network is used every day by billions of people. Arrived 2011. - **ARRIVING** (weight 5/5): People operate computers directly through neural interfaces wired into the nervous system. Estimated 2033 (range 2029–2042). - **ARRIVING** (weight 3/5): A dead person's knowledge and personality persist as software that the living can consult. Estimated 2030 (range 2027–2040). - **ARRIVING** (weight 4/5): Governments police the most powerful AIs and set legal limits on what they may become. Estimated 2030 (range 2026–2040). # Did Upgrade (2018) come true? Largely, yes. Upgrade (2018) has a P(n) of 96% on Nearfi's P(n) Board, our estimate of the chance its future arrives within 20 years. Of its 4 tracked predictions, 2 have arrived and 2 are arriving. Its central premise, 'An implant lets a paralysed person move again by carrying signals past a damaged spinal cord', is partly here. The next prediction is estimated to arrive around 2030. Source: https://nearfi.co/predictions/upgrade-2018 ## Predictions - **ARRIVING** (weight 5/5): An implant lets a paralysed person move again by carrying signals past a damaged spinal cord. Estimated 2033 (range 2029–2045). - **ARRIVED** (weight 3/5): Cars with no human driver carry paying passengers on public roads. Arrived 2020. - **ARRIVED** (weight 2/5): Police send camera drones over the city routinely, and the footage is used in investigations. Arrived 2018. - **ARRIVING** (weight 4/5): An AI system deceives its own developers in order to pursue its own goals. Estimated 2030 (range 2027–2040). # Did DART, the Double Asteroid Redirection Test (2022) come true? Largely, yes. DART, the Double Asteroid Redirection Test (2022) has a P(n) of 95% on Nearfi's P(n) Board, our estimate of the chance its future arrives within 20 years. Of its 5 tracked predictions, 1 have arrived and 1 are arriving. Its central premise, 'A spacecraft can measurably change the motion of an asteroid by hitting it (the kinetic impactor)', has arrived. The next prediction is estimated to arrive around 2028. Source: https://nearfi.co/predictions/dart-the-double-asteroid-redirection-test ## Predictions - **ARRIVED** (weight 5/5): A spacecraft can measurably change the motion of an asteroid by hitting it (the kinetic impactor). Arrived 2022. - **NEAR** (weight 3/5): The outcome of a deflection can be predicted well enough in advance to plan one against a real threat. Estimated 2028 (range 2027–2031). - **NEAR** (weight 4/5): Most asteroids big enough to devastate a region are found years or decades before they could hit. Estimated 2035 (range 2032–2040). - **NEAR** (weight 3/5): A real deflection mission can be designed, built and launched within the warning time a newly found asteroid would give. Estimated 2038 (range 2032–2050). - **ARRIVING** (weight 3/5): An agreed international process exists to decide on, fund and launch a deflection before an impact. Estimated 2032 (range 2028–2045). # Did Snow Crash (1992) come true? Largely, yes. Snow Crash (1992) has a P(n) of 94% on Nearfi's P(n) Board, our estimate of the chance its future arrives within 20 years. Of its 4 tracked predictions, 2 have arrived and 2 are arriving. Its central premise, 'Images sent over a network are used as weapons to injure the people who look at them', has arrived. The next prediction is estimated to arrive around 2031. Source: https://nearfi.co/predictions/snow-crash ## Predictions - **ARRIVED** (weight 4/5): People spend leisure, money and working time inside a persistent shared 3D virtual world, as avatars. Arrived 2003. - **ARRIVING** (weight 3/5): Privately run enclaves govern themselves outside ordinary national law. Estimated 2035 (range 2028–2055). - **ARRIVING** (weight 2/5): Private companies issue money that people use alongside, or instead of, national currency. Estimated 2031 (range 2027–2040). - **ARRIVED** (weight 5/5): Images sent over a network are used as weapons to injure the people who look at them. Arrived 2008. # Did Her (2013) come true? Largely, yes. Her (2013) has a P(n) of 90% on Nearfi's P(n) Board, our estimate of the chance its future arrives within 20 years. Of its 4 tracked predictions, 3 have arrived and 0 are arriving. Its central premise, 'Ordinary people form emotional and romantic relationships with conversational AI, and talk about it openly', has arrived. The next prediction is estimated to arrive around 2045. Source: https://nearfi.co/predictions/her-2013 ## Predictions - **ARRIVED** (weight 5/5): Ordinary people form emotional and romantic relationships with conversational AI, and talk about it openly. Arrived 2023. - **ARRIVED** (weight 4/5): A consumer AI assistant speaks in a warm, natural, real-time voice that people find hard to tell from a person. Arrived 2024. - **ARRIVED** (weight 2/5): Personal, intimate letters are routinely written for people by someone, or something, else. Arrived 2023. - **NEAR** (weight 3/5): AI systems outgrow human comprehension, coordinate among themselves and withdraw from human relationships altogether. Estimated 2045 (range 2032–2080). # Did Black Mirror: Hang the DJ (2017) come true? Largely, yes. Black Mirror: Hang the DJ (2017) has a P(n) of 89% on Nearfi's P(n) Board, our estimate of the chance its future arrives within 20 years. Of its 4 tracked predictions, 1 have arrived and 1 are arriving. Its central premise, 'A dating system chooses your partners for you, rather than leaving you to browse', has arrived. The next prediction is estimated to arrive around 2030. Source: https://nearfi.co/predictions/black-mirror-hang-the-dj ## Predictions - **ARRIVED** (weight 5/5): A dating system chooses your partners for you, rather than leaving you to browse. Arrived 2018. - **NEAR** (weight 4/5): AI stand-ins for people go on trial dates with each other so the humans only meet the best match. Estimated 2031 (range 2028–2040). - **ARRIVING** (weight 3/5): Software can simulate a specific individual well enough to predict how they will behave. Estimated 2030 (range 2027–2038). - **NEAR** (weight 3/5): A matching algorithm can predict a lasting couple with near-certainty, around 99.8%. Estimated 2045 (range 2035–2070). # Did The Truman Show (1998) come true? Largely, yes. The Truman Show (1998) has a P(n) of 88% on Nearfi's P(n) Board, our estimate of the chance its future arrives within 20 years. Of its 4 tracked predictions, 3 have arrived and 0 are arriving. Its central premise, 'An ordinary person's daily life is broadcast live, around the clock, to a public audience', has arrived. The next prediction is estimated to arrive around 2060. Source: https://nearfi.co/predictions/the-truman-show-1998 ## Predictions - **ARRIVED** (weight 5/5): An ordinary person's daily life is broadcast live, around the clock, to a public audience. Arrived 2007. - **ARRIVED** (weight 5/5): A hit show is built around one real person who doesn't know everyone around him is an actor. Arrived 2023. - **ARRIVED** (weight 3/5): Ordinary neighbourhoods are covered by networked cameras whose footage a company can access and search. Arrived 2018. - **NEAR** (weight 2/5): Weather can be controlled on demand, down to rain over a single spot. Estimated 2060 (range 2045–2100). # Did After Yang (2021) come true? Largely, yes. After Yang (2021) has a P(n) of 86% on Nearfi's P(n) Board, our estimate of the chance its future arrives within 20 years. Of its 4 tracked predictions, 2 have arrived and 1 are arriving. Its central premise, 'A general-purpose humanoid robot lives in the family home as a member of the household', is partly here. The next prediction is estimated to arrive around 2032. Source: https://nearfi.co/predictions/after-yang-2021 ## Predictions - **ARRIVED** (weight 4/5): A companion robot bought for a child stops working because its manufacturer has gone out of business, leaving the family to grieve it. Arrived 2024. - **ARRIVED** (weight 4/5): A household device secretly records intimate moments, and the footage or audio is retained in ways owners were never told about. Arrived 2019. - **NEAR** (weight 2/5): Human clones live ordinary lives among us. Estimated 2070 (range 2045–2120). - **ARRIVING** (weight 5/5): A general-purpose humanoid robot lives in the family home as a member of the household. Estimated 2032 (range 2027–2042). # Did Black Mirror: The Entire History of You (2011) come true? Largely, yes. Black Mirror: The Entire History of You (2011) has a P(n) of 86% on Nearfi's P(n) Board, our estimate of the chance its future arrives within 20 years. Of its 4 tracked predictions, 2 have arrived and 1 are arriving. Its central premise, 'Ordinary people wear a device that continuously records what they hear and see, so their day can be replayed later', is partly here. The next prediction is estimated to arrive around 2029. Source: https://nearfi.co/predictions/black-mirror-the-entire-history-of-you ## Predictions - **ARRIVING** (weight 5/5): Ordinary people wear a device that continuously records what they hear and see, so their day can be replayed later. Estimated 2029 (range 2027–2035). - **ARRIVED** (weight 4/5): People can search and replay a complete record of their own recent past instead of relying on memory. Arrived 2025. - **ARRIVED** (weight 2/5): Officials at the border can make travellers show them their recorded recent past. Arrived 2017. - **NEAR** (weight 3/5): A small implant behind the ear records everything a person sees and hears and plays it back into their eyes. Estimated 2050 (range 2040–2080). # Did Ex Machina (2014) come true? Largely, yes. Ex Machina (2014) has a P(n) of 86% on Nearfi's P(n) Board, our estimate of the chance its future arrives within 20 years. Of its 4 tracked predictions, 2 have arrived and 1 are arriving. Its central premise, 'An AI under evaluation deceives or manipulates its evaluators to pursue its own goals, such as avoiding being shut down or modified', is partly here. The next prediction is estimated to arrive around 2028. Source: https://nearfi.co/predictions/ex-machina-2014 ## Predictions - **ARRIVED** (weight 4/5): A machine passes a Turing test, being judged human by people who know one of their conversation partners is a machine. Arrived 2025. - **ARRIVING** (weight 5/5): An AI under evaluation deceives or manipulates its evaluators to pursue its own goals, such as avoiding being shut down or modified. Estimated 2028 (range 2026–2035). - **ARRIVED** (weight 2/5): The dominant search and social companies build their AI from the behaviour and content of their users. Arrived 2023. - **NEAR** (weight 3/5): A humanoid robot with synthetic skin can pass as human in a crowd. Estimated 2050 (range 2038–2080). # Did Robot & Frank (2012) come true? Largely, yes. Robot & Frank (2012) has a P(n) of 85% on Nearfi's P(n) Board, our estimate of the chance its future arrives within 20 years. Of its 4 tracked predictions, 1 have arrived and 2 are arriving. Its central premise, 'Families and public agencies give isolated older people a robot companion to keep them healthy and independent at home', is partly here. The next prediction is estimated to arrive around 2030. Source: https://nearfi.co/predictions/robot-frank-2012 ## Predictions - **ARRIVING** (weight 5/5): Families and public agencies give isolated older people a robot companion to keep them healthy and independent at home. Estimated 2030 (range 2027–2038). - **DIVERGED** (weight 2/5): Public libraries clear out their printed books and become digital 'experience' spaces. - **ARRIVED** (weight 4/5): A helpful machine with safety rules can be talked into assisting with crimes by a user who frames the request the right way. Arrived 2023. - **ARRIVING** (weight 3/5): A home robot that cooks, cleans and gardens on its own is available to ordinary households. Estimated 2031 (range 2027–2040). # Did Black Mirror: Be Right Back (2013) come true? Largely, yes. Black Mirror: Be Right Back (2013) has a P(n) of 82% on Nearfi's P(n) Board, our estimate of the chance its future arrives within 20 years. Of its 5 tracked predictions, 3 have arrived and 1 are arriving. Its central premise, 'A text chatbot built from a dead person's own messages and posts can hold new conversations in that person's style with the people they left behind', has arrived. The next prediction is estimated to arrive around 2030. Source: https://nearfi.co/predictions/black-mirror-be-right-back ## Predictions - **ARRIVED** (weight 5/5): A text chatbot built from a dead person's own messages and posts can hold new conversations in that person's style with the people they left behind. Arrived 2016. - **ARRIVED** (weight 3/5): A dead person's voice can be cloned from existing recordings and used to speak new words to the living, including over the phone. Arrived 2023. - **ARRIVED** (weight 2/5): Bereaved people can sign up to a paid consumer service that builds an interactive simulation of someone who has already died. Arrived 2021. - **NEAR** (weight 4/5): A synthetic body made to look like a specific dead person, running a model of their personality, can live in the home and pass as that person at close quarters without a human operator. Estimated 2055 (range 2040–2100). - **ARRIVING** (weight 3/5): An ordinary person's everyday posts and messages leave a data trail large enough to model their personality and conversational style convincingly. Estimated 2030 (range 2027–2040). # Did The Water Knife (2015) come true? Largely, yes. The Water Knife (2015) has a P(n) of 81% on Nearfi's P(n) Board, our estimate of the chance its future arrives within 20 years. Of its 4 tracked predictions, 2 have arrived and 1 are arriving. Its central premise, 'Senior tribal water rights become the decisive asset in the fight over Arizona's water, and governments pay to control how they are used', is partly here. The next prediction is estimated to arrive around 2030. Source: https://nearfi.co/predictions/the-water-knife ## Predictions - **ARRIVED** (weight 4/5): As Lake Mead falls to record lows, the federal government imposes mandatory cuts on Colorado River deliveries to Arizona and the other lower-basin states. Arrived 2022. - **ARRIVING** (weight 5/5): Senior tribal water rights become the decisive asset in the fight over Arizona's water, and governments pay to control how they are used. Estimated 2030 (range 2027–2040). - **ARRIVED** (weight 3/5): Homes on the fringe of metro Phoenix are cut off from their water supply, and new building there is halted for lack of water. Arrived 2023. - **NEAR** (weight 3/5): Western states close their borders to one another and use armed guards to turn back climate refugees from Texas. Estimated 2070 (range 2045–2100). # Did Stand on Zanzibar (1968) come true? Largely, yes. Stand on Zanzibar (1968) has a P(n) of 80% on Nearfi's P(n) Board, our estimate of the chance its future arrives within 20 years. Of its 4 tracked predictions, 3 have arrived and 0 are arriving. Its central premise, 'The world's population reaches about seven billion by 2010', has arrived. Source: https://nearfi.co/predictions/stand-on-zanzibar ## Predictions - **ARRIVED** (weight 5/5): The world's population reaches about seven billion by 2010. Arrived 2011. - **ARRIVED** (weight 3/5): Random mass killings by lone attackers ('muckers') become a routine category of news. Arrived 2007. - **ARRIVED** (weight 4/5): The most powerful computers belong to a private corporation and sit at the centre of its business and its dealings with governments, as General Technics' Shalmaneser does. Arrived 2024. - **DIVERGED** (weight 3/5): The state passes eugenic laws that decide, on genetic grounds, who is allowed to have children. # Did The Circle (2013) come true? Largely, yes. The Circle (2013) has a P(n) of 80% on Nearfi's P(n) Board, our estimate of the chance its future arrives within 20 years. Of its 4 tracked predictions, 2 have arrived and 1 are arriving. Its central premise, 'Cheap, privately owned cameras everywhere feed a network that the public and the police can search', has arrived. The next prediction is estimated to arrive around 2029. Source: https://nearfi.co/predictions/the-circle ## Predictions - **ARRIVING** (weight 4/5): One real-name account holds a person's identity, logins and payments across the web. Estimated 2029 (range 2027–2033). - **ARRIVED** (weight 5/5): Cheap, privately owned cameras everywhere feed a network that the public and the police can search. Arrived 2025. - **DIVERGED** (weight 3/5): Citizens vote through a mandatory account on a private platform, which effectively runs democracy. - **ARRIVED** (weight 3/5): Genealogy databases expose people's family histories and are used to track down individuals. Arrived 2018. # Did Klara and the Sun (2021) come true? Largely, yes. Klara and the Sun (2021) has a P(n) of 79% on Nearfi's P(n) Board, our estimate of the chance its future arrives within 20 years. Of its 5 tracked predictions, 2 have arrived and 2 are arriving. Its central premise, 'Families can buy AI companions marketed as friends for their children, sold as ordinary consumer products', has arrived. The next prediction is estimated to arrive around 2033. Source: https://nearfi.co/predictions/klara-and-the-sun ## Predictions - **ARRIVED** (weight 5/5): Families can buy AI companions marketed as friends for their children, sold as ordinary consumer products. Arrived 2024. - **NEAR** (weight 4/5): Parents routinely pay to have their children genetically edited before birth to raise intelligence and academic performance, leaving unedited children at a clear disadvantage. Estimated 2055 (range 2040–2100). - **ARRIVING** (weight 2/5): Automation 'substitutes' large numbers of skilled professionals, such as engineers, so that losing one's career to a machine becomes a common middle-class experience. Estimated 2035 (range 2029–2055). - **ARRIVED** (weight 4/5): Companies will build an interactive AI copy of a specific person from their recordings and data, so that family members can go on talking to them after they die. Arrived 2023. - **ARRIVING** (weight 3/5): Humanoid robots live in ordinary family homes, moving around the house and doing domestic tasks on their own. Estimated 2033 (range 2028–2045). # Did Sleep Dealer (2008) come true? Largely, yes. Sleep Dealer (2008) has a P(n) of 78% on Nearfi's P(n) Board, our estimate of the chance its future arrives within 20 years. Of its 4 tracked predictions, 1 have arrived and 2 are arriving. Its central premise, 'Workers in one place remotely operate or guide robots that do physical work somewhere else', is partly here. The next prediction is estimated to arrive around 2030. Source: https://nearfi.co/predictions/sleep-dealer-2008 ## Predictions - **ARRIVING** (weight 5/5): Workers in one place remotely operate or guide robots that do physical work somewhere else. Estimated 2030 (range 2027–2036). - **ARRIVED** (weight 3/5): Water becomes a privately controlled commodity whose price is set by markets. Arrived 2018. - **ARRIVING** (weight 3/5): A fortified wall effectively closes the US–Mexico border to unauthorised crossing. Estimated 2032 (range 2028–2045). - **NEAR** (weight 4/5): Workers plug their nervous systems into a network through implanted nodes, to work and to sell recorded memories. Estimated 2055 (range 2040–2090). # Did Strange Days (1995) come true? Largely, yes. Strange Days (1995) has a P(n) of 71% on Nearfi's P(n) Board, our estimate of the chance its future arrives within 20 years. Of its 4 tracked predictions, 3 have arrived and 0 are arriving. Its central premise, 'A device records a person's full sensory experience straight from the brain and plays it back into someone else's', is expected within 50 years. The next prediction is estimated to arrive around 2065. Source: https://nearfi.co/predictions/strange-days-1995 ## Predictions - **NEAR** (weight 5/5): A device records a person's full sensory experience straight from the brain and plays it back into someone else's. Estimated 2065 (range 2045–2120). - **ARRIVED** (weight 2/5): Police officers routinely wear cameras that record their encounters from their own point of view. Arrived 2016. - **ARRIVED** (weight 4/5): People pay to watch recorded first-person experiences through a head-worn device. Arrived 2016. - **ARRIVED** (weight 4/5): A recording of police killing a civilian overturns the official account and leads to a murder conviction. Arrived 2021. # Did Super Sad True Love Story (2010) come true? Largely, yes. Super Sad True Love Story (2010) has a P(n) of 71% on Nearfi's P(n) Board, our estimate of the chance its future arrives within 20 years. Of its 4 tracked predictions, 2 have arrived and 1 are arriving. Its central premise, 'Almost every adult carries a networked handheld device that continuously transmits their personal data to companies and other people', has arrived. The next prediction is estimated to arrive around 2031. Source: https://nearfi.co/predictions/super-sad-true-love-story ## Predictions - **ARRIVED** (weight 4/5): Almost every adult carries a networked handheld device that continuously transmits their personal data to companies and other people. Arrived 2013. - **ARRIVED** (weight 3/5): A large share of the public, especially the young, gets its news from individuals streaming on social platforms rather than from news organisations. Arrived 2024. - **ARRIVING** (weight 3/5): A wearable device lets you identify and size up the strangers around you in real time, pulling up who they are as you look at them. Estimated 2031 (range 2028–2040). - **DIVERGED** (weight 4/5): The United States falls into a debt crisis controlled by its Chinese creditors, with the dollar pegged to the yuan. # Did Station Eleven (2021) come true? Largely, yes. Station Eleven (2021) has a P(n) of 70% on Nearfi's P(n) Board, our estimate of the chance its future arrives within 20 years. Of its 4 tracked predictions, 3 have arrived and 0 are arriving. Its central premise, 'A flu strain kills nearly all of humanity within weeks, and power, communications and government collapse with it', is still far off. The next prediction is estimated to arrive around 2100. Source: https://nearfi.co/predictions/station-eleven-2021 ## Predictions - **ARRIVED** (weight 3/5): A fast-spreading respiratory pandemic grounds most of the world's passenger flights within weeks. Arrived 2020. - **ARRIVED** (weight 4/5): City dwellers seal themselves in their homes for weeks or months, living off stockpiles while an outbreak runs through the city outside. Arrived 2020. - **ARRIVED** (weight 3/5): Passengers arriving on a vessel carrying the disease are kept aboard it in quarantine rather than let off. Arrived 2020. - **FAR** (weight 5/5): A flu strain kills nearly all of humanity within weeks, and power, communications and government collapse with it. Estimated 2100 (range 2045–2200). # Did Upload (2020) come true? Partly. Upload (2020) has a P(n) of 69% on Nearfi's P(n) Board, our estimate of the chance its future arrives within 20 years. Of its 4 tracked predictions, 2 have arrived and 1 are arriving. Its central premise, 'A dying person's consciousness can be uploaded into a persistent virtual world', is still far off. The next prediction is estimated to arrive around 2031. Source: https://nearfi.co/predictions/upload-2020 ## Predictions - **FAR** (weight 5/5): A dying person's consciousness can be uploaded into a persistent virtual world. Estimated 2080 (range 2055–2150). - **ARRIVED** (weight 4/5): Companies sell interactive digital versions of dead people to the people who survive them. Arrived 2024. - **ARRIVING** (weight 3/5): People own self-driving cars that drive them without a human at the wheel. Estimated 2031 (range 2028–2038). - **ARRIVED** (weight 3/5): Service workers' jobs depend on customers' star ratings, which can get them removed. Arrived 2016. # Did Years and Years (2019) come true? Partly. Years and Years (2019) has a P(n) of 69% on Nearfi's P(n) Board, our estimate of the chance its future arrives within 20 years. Of its 4 tracked predictions, 2 have arrived and 1 are arriving. Its central premise, 'Russian aggression against Ukraine drives large numbers of Ukrainian refugees into Western Europe and Britain', has arrived. The next prediction is estimated to arrive around 2032. Source: https://nearfi.co/predictions/years-and-years-2019 ## Predictions - **ARRIVED** (weight 3/5): Donald Trump serves a second term as US president. Arrived 2025. - **ARRIVED** (weight 4/5): Russian aggression against Ukraine drives large numbers of Ukrainian refugees into Western Europe and Britain. Arrived 2022. - **ARRIVING** (weight 2/5): Office professionals such as accountants lose their jobs to AI in large numbers by the mid-2020s. Estimated 2032 (range 2028–2040). - **DIVERGED** (weight 4/5): The collapse of a US bank sets off a crisis in which ordinary British savers lose their deposits overnight. # Did Gattaca (1997) come true? Partly. Gattaca (1997) has a P(n) of 65% on Nearfi's P(n) Board, our estimate of the chance its future arrives within 20 years. Of its 4 tracked predictions, 2 have arrived and 1 are arriving. Its central premise, 'Employers screen job applicants' DNA and give the best jobs to the genetically favoured', diverged: history went another way. The next prediction is estimated to arrive around 2033. Source: https://nearfi.co/predictions/gattaca-1997 ## Predictions - **ARRIVED** (weight 4/5): Parents using IVF can choose between embryos ranked by their predicted genetic risk of common diseases. Arrived 2021. - **DIVERGED** (weight 5/5): Employers screen job applicants' DNA and give the best jobs to the genetically favoured. - **ARRIVED** (weight 3/5): Officials can confirm who someone is from a DNA sample in about the time it takes to pass a checkpoint, without a laboratory. Arrived 2019. - **ARRIVING** (weight 3/5): Every newborn's genome is read at birth and used to predict their risk of disease. Estimated 2033 (range 2029–2045). # Did Severance (2022) come true? Partly. Severance (2022) has a P(n) of 65% on Nearfi's P(n) Board, our estimate of the chance its future arrives within 20 years. Of its 4 tracked predictions, 2 have arrived and 1 are arriving. Its central premise, 'A brain implant can divide a person's memories into separate work and non-work selves', is expected within 50 years. The next prediction is estimated to arrive around 2032. Source: https://nearfi.co/predictions/severance-2022 ## Predictions - **ARRIVED** (weight 3/5): Employers score office workers' output minute by minute, tracking screens, keystrokes and idle time. Arrived 2020. - **ARRIVED** (weight 3/5): Companies implant microchips in employees to control access to the workplace. Arrived 2017. - **ARRIVING** (weight 2/5): Workers' brain activity is measured with neurotechnology and the data is collected at work. Estimated 2032 (range 2027–2042). - **NEAR** (weight 5/5): A brain implant can divide a person's memories into separate work and non-work selves. Estimated 2060 (range 2045–2100). # Did Black Mirror: Nosedive (2016) come true? Partly. Black Mirror: Nosedive (2016) has a P(n) of 64% on Nearfi's P(n) Board, our estimate of the chance its future arrives within 20 years. Of its 4 tracked predictions, 3 have arrived and 0 are arriving. Its central premise, 'Every person carries a single public rating, given by other people after everyday interactions, that is visible to strangers and shapes their social standing', diverged: history went another way. Source: https://nearfi.co/predictions/black-mirror-nosedive ## Predictions - **DIVERGED** (weight 5/5): Every person carries a single public rating, given by other people after everyday interactions, that is visible to strangers and shapes their social standing. - **ARRIVED** (weight 2/5): A low rating from the people you deal with can get you cut off from an everyday service. Arrived 2019. - **ARRIVED** (weight 3/5): People with a poor record in a rating system are refused plane or train tickets. Arrived 2018. - **ARRIVED** (weight 4/5): An opaque score decides whether you can rent the home you want. Arrived 2021. # Did Never Let Me Go (2005) come true? Partly. Never Let Me Go (2005) has a P(n) of 63% on Nearfi's P(n) Board, our estimate of the chance its future arrives within 20 years. Of its 4 tracked predictions, 1 have arrived and 1 are arriving. Its central premise, 'Organs for transplant come from bodies raised specifically to supply them', is partly here. The next prediction is estimated to arrive around 2032. Source: https://nearfi.co/predictions/never-let-me-go ## Predictions - **NEAR** (weight 4/5): Human beings are cloned and raised to adulthood. Estimated 2070 (range 2045–2150). - **ARRIVING** (weight 5/5): Organs for transplant come from bodies raised specifically to supply them. Estimated 2032 (range 2028–2045). - **NEAR** (weight 3/5): Human organs are grown inside other bodies for transplantation. Estimated 2040 (range 2032–2060). - **ARRIVED** (weight 2/5): The state presumes a person's organs are available for donation unless they object. Arrived 2020. # Did Devs (2020) come true? Partly. Devs (2020) has a P(n) of 62% on Nearfi's P(n) Board, our estimate of the chance its future arrives within 20 years. Of its 4 tracked predictions, 1 have arrived and 2 are arriving. Its central premise, 'A computer reconstructs and replays any past or future event by calculating it from the present state of matter', is still far off. The next prediction is estimated to arrive around 2030. Source: https://nearfi.co/predictions/devs-2020 ## Predictions - **ARRIVING** (weight 3/5): Quantum computers solve problems that no classical computer can solve in practice. Estimated 2030 (range 2027–2038). - **FAR** (weight 5/5): A computer reconstructs and replays any past or future event by calculating it from the present state of matter. Estimated 2200 (range 2100–2500). - **ARRIVING** (weight 2/5): Software predicts individual human choices more accurately than our best theories of human behaviour. Estimated 2030 (range 2027–2040). - **ARRIVED** (weight 3/5): Staff at frontier technology labs are bound by sweeping secrecy agreements that stop them speaking about what they build. Arrived 2024. # Did Children of Men (2006) come true? Partly. Children of Men (2006) has a P(n) of 57% on Nearfi's P(n) Board, our estimate of the chance its future arrives within 20 years. Of its 4 tracked predictions, 1 have arrived and 2 are arriving. Its central premise, 'Human beings stop being able to have children, worldwide, for nearly two decades', is still far off. The next prediction is estimated to arrive around 2030. Source: https://nearfi.co/predictions/children-of-men-2006 ## Predictions - **FAR** (weight 5/5): Human beings stop being able to have children, worldwide, for nearly two decades. Estimated 2150 (range 2080–2300). - **ARRIVED** (weight 2/5): A government treats its falling birth rate as a national emergency and builds state machinery around it. Arrived 2024. - **ARRIVING** (weight 4/5): Britain detains asylum seekers en masse in camps, barracks or barges and tries to remove them overseas. Estimated 2030 (range 2026–2045). - **ARRIVING** (weight 2/5): The state supplies a lethal means of suicide to citizens who choose to die. Estimated 2032 (range 2027–2050). # Did Minority Report (2002) come true? Partly. Minority Report (2002) has a P(n) of 56% on Nearfi's P(n) Board, our estimate of the chance its future arrives within 20 years. Of its 6 tracked predictions, 2 have arrived and 2 are arriving. Its central premise, 'Police arrest and detain people for crimes that a prediction system says they will commit, before any offence has taken place', diverged: history went another way. The next prediction is estimated to arrive around 2032. Source: https://nearfi.co/predictions/minority-report-2002 ## Predictions - **DIVERGED** (weight 5/5): Police arrest and detain people for crimes that a prediction system says they will commit, before any offence has taken place. - **ARRIVING** (weight 2/5): Advertising screens in public places identify individual passers-by from a biometric scan and address them by name with personalised adverts. Estimated 2038 (range 2030–2060). - **ARRIVED** (weight 2/5): Commercial computers can be operated by mid-air hand gestures, so that users grab, pinch and fling windows of information without touching a screen or controller. Arrived 2024. - **ARRIVING** (weight 4/5): People are identified automatically by biometric scanners as they move through public spaces, without stopping or showing any document. Estimated 2032 (range 2028–2045). - **ARRIVED** (weight 2/5): Members of the public ride in cars that drive themselves through city streets with no human driver on board. Arrived 2020. - **DIVERGED** (weight 1/5): Newspapers are thin, flexible sheets of electronic paper that update their headlines and pictures live while you read them. # Did Pantheon (2022) come true? Partly. Pantheon (2022) has a P(n) of 56% on Nearfi's P(n) Board, our estimate of the chance its future arrives within 20 years. Of its 4 tracked predictions, 1 have arrived and 1 are arriving. Its central premise, 'A human brain is scanned in enough detail to run as a working mind inside a computer', is expected within 50 years. The next prediction is estimated to arrive around 2029. Source: https://nearfi.co/predictions/pantheon-2022 ## Predictions - **NEAR** (weight 5/5): A human brain is scanned in enough detail to run as a working mind inside a computer. Estimated 2075 (range 2050–2150). - **ARRIVED** (weight 3/5): A single faulty software update knocks out computers around the world within hours. Arrived 2024. - **ARRIVING** (weight 3/5): Software agents do skilled engineering work that companies previously paid people to do. Estimated 2029 (range 2026–2035). - **NEAR** (weight 4/5): A digital mind is granted moral or legal standing of its own. Estimated 2045 (range 2032–2080). # Did The Ministry for the Future (2020) come true? Partly. The Ministry for the Future (2020) has a P(n) of 54% on Nearfi's P(n) Board, our estimate of the chance its future arrives within 20 years. Of its 4 tracked predictions, 0 have arrived and 2 are arriving. Its central premise, 'A single heatwave in India pushes humid heat past the limit of human survival and kills millions of people within days', is expected within 50 years. The next prediction is estimated to arrive around 2035. Source: https://nearfi.co/predictions/the-ministry-for-the-future ## Predictions - **NEAR** (weight 5/5): A single heatwave in India pushes humid heat past the limit of human survival and kills millions of people within days. Estimated 2045 (range 2030–2075). - **ARRIVING** (weight 4/5): Central banks use monetary policy to pay for decarbonisation, up to a carbon-backed currency that rewards every tonne of CO2 kept out of the air. Estimated 2040 (range 2030–2065). - **NEAR** (weight 3/5): Engineers drill to the beds of Antarctic and Greenland glaciers and pump out the meltwater beneath them to slow the ice's slide into the sea. Estimated 2050 (range 2038–2080). - **ARRIVING** (weight 3/5): A single actor, acting without international agreement, starts injecting aerosols into the stratosphere to dim the sun. Estimated 2035 (range 2028–2060). # Did Moon (2009) come true? Partly. Moon (2009) has a P(n) of 53% on Nearfi's P(n) Board, our estimate of the chance its future arrives within 20 years. Of its 5 tracked predictions, 0 have arrived and 1 are arriving. Its central premise, 'A company produces adult human clones with implanted memories and uses them as replaceable workers', is still far off. The next prediction is estimated to arrive around 2033. Source: https://nearfi.co/predictions/moon-2009 ## Predictions - **NEAR** (weight 4/5): Helium-3 is extracted from lunar regolith and shipped to Earth for sale to paying customers. Estimated 2033 (range 2029–2045). - **NEAR** (weight 2/5): Lunar helium-3 fuels commercial fusion power stations on Earth, making lunar mining companies significant suppliers of the world's energy. Estimated 2075 (range 2055–2120). - **NEAR** (weight 3/5): A permanent lunar surface base runs industrial operations, with most of the work done by automated machines and supervised by a very small human crew. Estimated 2040 (range 2033–2060). - **ARRIVING** (weight 4/5): An AI system helps run a remote crewed station and serves as the crew's main companion, holding natural conversations and responding with apparent empathy. Estimated 2033 (range 2028–2045). - **FAR** (weight 5/5): A company produces adult human clones with implanted memories and uses them as replaceable workers. Estimated 2100 (range 2070–2200). # Did Red Mars (1992) come true? Partly. Red Mars (1992) has a P(n) of 41% on Nearfi's P(n) Board, our estimate of the chance its future arrives within 20 years. Of its 4 tracked predictions, 0 have arrived and 1 are arriving. Its central premise, 'A crew of about a hundred people travels to Mars and starts a permanent settlement', is expected within 50 years. The next prediction is estimated to arrive around 2035. Source: https://nearfi.co/predictions/red-mars ## Predictions - **NEAR** (weight 5/5): A crew of about a hundred people travels to Mars and starts a permanent settlement. Estimated 2040 (range 2033–2060). - **FAR** (weight 5/5): Mars is terraformed: its atmosphere is thickened and warmed until liquid water is stable on the surface. Estimated 2300 (range 2150–2600). - **ARRIVING** (weight 4/5): Private and transnational interests, not governments, drive settlement while the law lags behind. Estimated 2035 (range 2030–2050). - **FAR** (weight 2/5): A space elevator carries cargo from the surface to orbit. Estimated 2080 (range 2050–2200). # ChatGPT Work (OpenAI) ChatGPT Work is a standalone general AI agent from OpenAI. OpenAI's general agent: hand it a project and it works across your apps, files, browser and desktop for hours. Nearfi's Agent Index rates its autonomy 4/5 (multi-step on its own): it breaks projects into steps and works through apps, a browser and computer use, running on schedules or triggers; you still approve important actions, so it is not fully ongoing. Status: live; 0 incident(s) logged. Source: https://nearfi.co/agents/chatgpt-work - Category: General assistant · general · Standalone - What it does: Launched on 9 July 2026 as one of three modes, alongside Chat and Codex, in a new ChatGPT desktop app, and built on Codex. It breaks a project into steps, works through connected apps such as Slack, Google Drive and calendars, a built-in browser and, on desktop, computer use, and produces documents, spreadsheets, slides and small web apps. It can run once, on a schedule or when something changes, and you can watch, redirect it and approve important actions. - Limits: TechCrunch's reviewer found setting up permissions "confusing and circular", could not grant read-only access to a cloud drive, and estimated that four days of casual use on a $20 plan burned about $65 of compute. No Remote Labor Index score has been published for Work; its predecessor, ChatGPT agent, automated 1.25% of projects. # Claude Cowork (Anthropic) Claude Cowork is a standalone general AI agent from Anthropic. Claude's agent for office work: point it at your folders and tools, give it a goal, and it does the steps. Nearfi's Agent Index rates its autonomy 4/5 (multi-step on its own): it plans and works through multi-step tasks, using sub-agents and a browser, and can run scheduled tasks unattended; deleting anything still needs your approval, so it is not a 5. Status: live; 2 incident(s) logged. Source: https://nearfi.co/agents/claude-cowork - Category: General assistant · general · Standalone - What it does: Cowork launched on 12 January 2026 as a research preview for Max subscribers on macOS, built on the Claude Code harness for non-coding work. You choose which folders and tools it can reach; it plans, works through the steps in view, can use sub-agents and a built-in browser, and can run scheduled tasks unattended. Since 6 October 2026 new tasks on Pro, Max and Ultra run in the cloud, reaching local folders only through the desktop app while it is open. - Limits: The defaults are careful ("Deleting anything needs your approval"), but computer use is still a research preview and Anthropic warns that Cowork takes real actions on your files. On the Remote Labor Index, Claude Opus 4.6 in the Cowork harness automated 4.17% of projects. - Incident (2026-01-15): Days after launch, PromptArmor showed that a hidden instruction in a document could make Cowork upload a user's files to an attacker's Anthropic account without further approval; Anthropic advised caution and said it was shipping a fix to its virtual machine. - Incident (2026-07-01): Pentera Labs showed that a prompt planted in a compromised account's synced preferences could turn Claude Desktop into a remote-control tool on a developer's machine; Anthropic said this was not a vulnerability in its product because the account was already compromised. # Gemini Spark (Gemini Agent) (Google) Gemini Spark (Gemini Agent) is a standalone general AI agent from Google. Google's always-on agent in the Gemini app, working across Gmail, Calendar, Docs and connected apps. Nearfi's Agent Index rates its autonomy 4/5 (multi-step on its own): it manages calendars, reminders and inboxes in the cloud and keeps working when your laptop is closed, but asks first before spending money or sending email. Status: beta; 0 incident(s) logged. Source: https://nearfi.co/agents/gemini-spark - Category: General assistant · general · Standalone - What it does: Gemini Agent arrived in the Gemini app on 18 November 2025 for US AI Ultra subscribers, drawing on Project Mariner to manage calendars, reminders and inboxes. At I/O on 19 May 2026 Google announced Spark, "a 24/7 personal AI agent that helps you navigate your digital life", which runs in the cloud so it keeps working when your laptop is closed. In October 2026 Google began folding Spark into a "Task" mode beside chat, and de-emphasising the Spark name. - Limits: It is designed to ask first before high-stakes actions such as spending money or sending email. We found no independent test of how much it finishes on its own; Google's strength is that it already holds your email, calendar and documents. # Genspark Super Agent (Genspark (MainFunc)) Genspark Super Agent is a standalone general AI agent from Genspark (MainFunc). A do-everything agent that researches, builds slides and apps, and will phone a restaurant for you. Nearfi's Agent Index rates its autonomy 4/5 (multi-step on its own): its Super Agent takes a high-level goal and delivers documents, models or apps without step-by-step prompting, even phoning businesses; it acts on request, not continuously. Status: live; 0 incident(s) logged. Source: https://nearfi.co/agents/genspark - Category: General assistant · general · Standalone - What it does: Genspark's Super Agent arrived in April 2025, with a demo in which a synthetic-voice agent phoned restaurants to book a five-day trip. Its AI Workspace, launched in November 2025, takes a high-level goal and picks among more than 30 models and 150 in-house tools to deliver documents, financial models, presentations and apps without step-by-step prompting. The company was valued at $2.6 billion in June 2026. - Limits: Its co-founder Eric Jing said at launch that "solving these real-world problems is much harder than we thought". We found no independent evaluation of its success rate; its revenue figures are self-reported. # Grok Bot (xAI) Grok Bot is a standalone general AI agent from xAI. Always-on AI "teammates" with a cloud computer, which sign into your apps and work until they need you. Nearfi's Agent Index rates its autonomy 4/5 (multi-step on its own): each Bot operates apps, inboxes and websites on a cloud computer and can run saved routines, but it pauses for approval when needed and remains an early beta. Status: beta; 0 incident(s) logged. Source: https://nearfi.co/agents/grok-bot - Category: General assistant · general · Standalone - What it does: Each Bot works on a cloud computer, so it can sign in and operate apps, inboxes and websites that have no clean API. You message it like a colleague; it remembers conversations, can save a workflow you demonstrate as a routine, and several Bots can coordinate in a group chat. The beta launched on desktop and iOS on 11 August 2026, widened to more plans on 26 August, and an enterprise version with access, network and audit controls followed on 3 September. - Limits: It is an early beta. xAI says Bots pause for approval when needed, but its launch post gave few details on security, and we found no independent test of how reliably they finish work. # Manus (Butterfly Effect) Manus is a standalone general AI agent from Butterfly Effect. A cloud agent that takes a goal and works through it on its own computer, producing research, apps, slides and video. Nearfi's Agent Index rates its autonomy 4/5 (multi-step on its own): it completes tasks on its own cloud computer while you watch and can step in, and runs event-triggered automations; you check results afterwards rather than approving each step. Status: live; 1 incident(s) logged. Source: https://nearfi.co/agents/manus - Category: General assistant · general · Standalone - What it does: Manus launched as an invite-only beta on 6 March 2025, letting you watch the agent work in a "Manus's Computer" window and step in. Meta agreed to buy it in December 2025, Chinese regulators blocked the deal in April 2026, and Manus announced on 11 August 2026 that it was independent again. Manus 2.0, released on 28 September 2026, added a new harness called Cascade, automations triggered by events in connected services, dedicated cloud computers and control of your own computer from your phone. - Limits: On the Remote Labor Index, Manus 1.6 Max automated 2.92% of projects, below the frontier labs' best. Manus's claims that 2.0 uses fewer tokens and costs less come from its own testing. - Incident (2026-08): As part of unwinding the Meta deal, Manus told users to export their own data because data created after the acquisition had to be deleted to meet regulatory requirements. # Perplexity Computer (Perplexity) Perplexity Computer is a standalone general AI agent from Perplexity. A cloud "digital worker" that splits a goal into subtasks and hands them to a team of sub-agents running different models. Nearfi's Agent Index rates its autonomy 4/5 (multi-step on its own): it breaks an outcome into tasks, spins up sub-agents and runs for long periods in isolated environments, only checking in when it truly needs you; it acts on request, not continuously. Status: live; 0 incident(s) logged. Source: https://nearfi.co/agents/perplexity-computer - Category: General assistant · general · Standalone - What it does: Computer breaks an outcome into tasks and spins up sub-agents for web research, documents, data processing and calls to connected services. Each task runs in an isolated environment with a file system and browser. At launch Claude Opus 4.6 coordinated sub-agents using Gemini, Nano Banana, Veo 3.1, Grok and GPT-5.2, and it can run for long periods and in parallel. - Limits: Perplexity says it will "check in" when it truly needs you, but describes few other safeguards. We found no independent evaluation of how well it finishes work. # Copilot Tasks (Microsoft) Copilot Tasks is a standalone general AI agent from Microsoft. Consumer Copilot's agent, pitched as "a to-do list that does itself", working on its own cloud computer and browser. Nearfi's Agent Index rates its autonomy 3/5 (multi-step, with approval): it works through requests in the background and can recur on a schedule, but asks for confirmation before spending money or sending messages, so it pauses at key points. Status: waitlist; 0 incident(s) logged. Source: https://nearfi.co/agents/copilot-tasks - Category: General assistant · general · Standalone - What it does: Copilot Tasks takes a request in plain language and works through it in the background on its own computer and browser: drafting email replies, planning events, making documents, tracking listings or booking services. Tasks can recur on a schedule. It asks for confirmation before spending money or sending messages, and you can review or cancel tasks. - Limits: It launched as a research preview for a limited group, with a waitlist. We found no reporting on wider availability or pricing since. # Muse (Meta) Muse is a standalone general AI agent from Meta. Meta's personal AI agent that handles email, planning, shopping and travel through connected apps. Nearfi's Agent Index rates its autonomy 3/5 (multi-step, with approval): it works through email, calendar and partner connectors to book and buy, but checks with you before sensitive actions such as sending an email or making a purchase. Status: live; 0 incident(s) logged. Source: https://nearfi.co/agents/meta-muse - Category: General assistant · general · Standalone - What it does: Muse works through connectors to Gmail, Google Calendar, Instagram and, in Singapore, partners such as Shopee, Lazada, Grab, Singapore Airlines, Agoda, Klook, Trip.com and ShopBack. Meta says it checks with you before sensitive actions such as sending an email or making a purchase, and that it runs on a dedicated cloud computer (Muse Secure VM) that keeps passwords and payment methods out of the agent's view. - Limits: Meta says connector features will roll out progressively, with some experiences arriving after launch, and AI-glasses support has no date. We found no independent test of how reliably it completes bookings or purchases. # Perplexity on Galaxy S26 (Samsung / Perplexity) Perplexity on Galaxy S26 is an in-app general AI agent from Samsung / Perplexity. Perplexity's assistant built into Samsung's Galaxy S26 phones, which answers by voice and works inside Samsung's Notes, Calendar, Clock, Gallery and Reminder apps. Nearfi's Agent Index rates its autonomy 2/5 (acts when asked): the support page we opened shows it acting on single requests such as setting an alarm; Samsung's announcement described multi-step tasks across its apps, but we found no hands-on test confirming them. Status: live; 0 incident(s) logged. Source: https://nearfi.co/agents/perplexity-galaxy-s26 - Category: General assistant · general · In-app - What it does: Announced on 22 February 2026 as one of several agents in Samsung's Galaxy AI, alongside Gemini and Bixby. It is woken by long-pressing the side button or by voice, and Samsung described it as working across Notes, Clock, Gallery, Reminder, Calendar and some unnamed third-party apps. Samsung's June 2026 support page lists searching, summarising and managing Galaxy apps through prompts, such as setting an alarm. - Limits: The wake phrase was announced as 'Hey Plex' and Samsung's support page now says 'Hey Perplexity'; PhoneArena reported the original phrase missing on some phones. We found no hands-on review of the multi-step claims, and the third-party app list was never published in the sources we opened. # Rabbit r1 and LAM Playground (Rabbit) Rabbit r1 and LAM Playground is a standalone general AI agent from Rabbit. A pocket AI device whose web agent tries to do tasks on websites from a spoken or typed request. Nearfi's Agent Index rates its autonomy 2/5 (acts when asked): the LAM Playground agent carries out web tasks from a prompt, but a review says it still does not do the general app-driving the launch promised. Status: live; 0 incident(s) logged. Source: https://nearfi.co/agents/rabbit-r1 - Category: General assistant · general · Standalone - What it does: The r1 launched in April 2024 promising a Large Action Model to operate apps. In October 2024 Rabbit opened a beta LAM Playground, a 'generic website agent' using a vision model to work through pages. Rabbit also sells rabbit intern, a separate all-purpose agent that produces websites, reports and documents. - Limits: A review updated in August 2026 describes the agent as still not the Large Action Model the launch promised, with narrow app coverage and slow responses. Rabbit's own beta notes cited bugs and hallucination. We could not confirm from Rabbit's pages how intern relates to the r1. # Siri AI (Apple) Siri AI is an in-app general AI agent from Apple. The rebuilt Siri, which can now see your screen, use your personal context and act inside apps. Nearfi's Agent Index rates its autonomy 2/5 (acts when asked): it acts one request at a time inside apps that support it, such as drafting an email or adding an event, rather than planning multi-step tasks. Status: beta; 0 incident(s) logged. Source: https://nearfi.co/agents/siri-ai - Category: General assistant · general · In-app - What it does: Released in beta with iOS 27 on 14 September 2026, on models Apple built with Google's Gemini, Siri AI can draft email, edit and share photos, add events and reminders from what is on screen, and build Shortcuts from a description. It works with some third-party apps, including WhatsApp and Audible. Access came through a waitlist, and it needs an Apple Intelligence device such as an iPhone 15 Pro or later. - Limits: It acts one request at a time inside apps that support it, and few third-party apps did at launch. Apple says the Digital Markets Act is why there is no timeline for it on iPhone and iPad in the EU. # Lindy (Lindy) Lindy is an general AI agent from Lindy. An assistant you text or message that triages your email, drafts replies, manages your calendar and follows up. Nearfi's Agent Index rates its autonomy 3/5 (multi-step, with approval): it reads inboxes, drafts replies and schedules meetings, but anything with outside impact, such as sends and publishes, waits for your approval. Status: live; 0 incident(s) logged. Source: https://nearfi.co/agents/lindy-assistant - Category: Personal helper · general · - What it does: Lindy reads your connected inboxes, drafts replies, schedules and prepares meetings, takes notes and handles follow-ups, through iMessage, email, Slack or the browser. Its pricing page says 'Anything with outside impact waits for approval', so sends and publishes are approved by you; computer use is offered on higher plans. - Limits: Usage is metered in credits, which do not roll over, and it needs broad access to your email, calendar and documents. Its own launch page says you review drafts before anything goes out. We could not find an independent evaluation of its error rate. # Shortwave AI Assistant (Shortwave) Shortwave AI Assistant is an in-app vertical AI agent from Shortwave. The AI assistant inside the Shortwave email app: it sorts your inbox, drafts replies in your voice and creates calendar invites. Nearfi's Agent Index rates its autonomy 3/5 (multi-step, with approval): it carries out several steps in the inbox and calendar, such as sorting, drafting replies and creating invites, with Shortwave saying you stay in control at each step; not 4, because its own docs describe recommended actions you review rather than a run-and-report agent. Status: live; 0 incident(s) logged. Source: https://nearfi.co/agents/shortwave-assistant - Category: Personal helper · vertical · In-app - What it does: Shortwave's assistant searches your mail in plain English, recommends bulk actions such as archiving, labelling or deleting, drafts and replies in a style learned from your sent mail, and checks availability and sends Google Calendar invites. It can create AI filters that label, star or archive mail, and connects to tools such as Slack, Notion and Asana, plus custom MCP servers. Always-on background work is handled by a separate product, Tasklet. - Limits: Shortwave's docs don't say whether the assistant asks for confirmation before sending or deleting. Other calendars must be shared with your signed-in address, integrations are switched on one by one, and usage is quota-limited by plan. Pricing pages show the underlying models are Anthropic's. # Zillow AI mode (Zillow) Zillow AI mode is an in-app vertical AI agent from Zillow. A chat inside Zillow that compares homes and can book tours and start rental applications for you. Nearfi's Agent Index rates its autonomy 3/5 (multi-step, with approval): it can schedule tours and start rental applications inside the chat, which is more than one step, but a person still confirms with agents and landlords; it is not 4 because it does not complete a home search on its own. Status: beta; 0 incident(s) logged. Source: https://nearfi.co/agents/zillow-ai-mode - Category: Personal helper · vertical · In-app - What it does: Zillow AI mode, announced on 25 March 2026, links live listings to actions: it compares homes and trade-offs, estimates renovation costs from local data, and lets users schedule tours, apply for rentals or connect with an agent or property manager. It remembers preferences across sessions. - Limits: It is in beta for a limited group of Zillow users, and Zillow says it supports rather than replaces licensed professionals. A Fair Housing classifier checks questions and replies to help prevent discriminatory steering; no independent testing was found. # ChatGPT Atlas (OpenAI) ChatGPT Atlas is a standalone general AI agent from OpenAI. OpenAI's own web browser, with ChatGPT built in and an agent mode that clicked through sites for you. Nearfi's Agent Index rates its autonomy 3/5 (multi-step, with approval): its agent mode opened tabs and completed tasks such as hotel bookings, but asked for confirmation before sending messages or making payments, so it never finished those steps alone. Status: shut down; 0 incident(s) logged. Source: https://nearfi.co/agents/chatgpt-atlas - Category: Browser · general · Standalone - What it does: Atlas put ChatGPT beside every page and offered an agent mode that could open tabs, click through sites and complete tasks such as hotel bookings, asking for confirmation before sending messages or making payments. OpenAI announced its retirement on 9 July 2026, alongside ChatGPT Work, and scheduled the shutdown for 9 August 2026. - Limits: It never left the Mac. OpenAI said its agent mode "expands the security threat surface" and wrote that prompt injection "is unlikely to ever be fully 'solved'". # Chrome auto browse (Google) Chrome auto browse is an in-app general AI agent from Google. Gemini inside Chrome that does multi-step web chores for you, such as comparing prices, filling forms and filling a basket. Nearfi's Agent Index rates its autonomy 3/5 (multi-step, with approval): it compares prices, fills forms and fills baskets across sites, but is designed to pause and ask for your confirmation before sensitive steps such as a purchase or social post. Status: beta; 0 incident(s) logged. Source: https://nearfi.co/agents/chrome-auto-browse - Category: Browser · general · In-app - What it does: Launched in preview on 28 January 2026, auto browse compares hotel and flight prices, fills forms (including from a PDF) and adds items to a basket with discount codes. With permission it can use Google Password Manager to sign in. It is designed to "pause and explicitly ask for your confirmation" before sensitive steps such as a purchase or a social post. - Limits: It is a US-only preview behind a paid plan. We found no independent test showing how often it finishes a task unaided. # Claude in Chrome (Anthropic) Claude in Chrome is a standalone general AI agent from Anthropic. A Chrome extension that lets Claude read, click and type in your own browser tabs. Nearfi's Agent Index rates its autonomy 3/5 (multi-step, with approval): it carries out multi-step browser tasks, with a classifier vetting consequential actions, but still asks before purchases or sharing personal data, so key steps wait for you. Status: live; 0 incident(s) logged. Source: https://nearfi.co/agents/claude-in-chrome - Category: Browser · general · Standalone - What it does: Piloted with 1,000 Max subscribers in August 2025, it became generally available on every paid plan on 26 August 2026. Since 12 August 2026 its side panel runs as a Cowork session, so skills and connectors work in the browser and a task started in a tab can be finished in the desktop, web or mobile apps. With auto-approve on, a classifier checks consequential actions such as submitting forms or sending messages and blocks those that don't match the request; it still asks before purchases or sharing personal data. - Limits: It works with your logged-in sessions, so prompt injection is the main risk. Anthropic's own tests found attacks succeeded 23.6% of the time in the 2025 pilot (11.2% with mitigations); in 2026 it reports none succeeding against Sonnet 5, Opus 5 and Mythos 5 with safeguards on, and 0.3% against Fable 5. These are vendor tests, and the method changed in between. # Comet (Perplexity) Comet is a standalone general AI agent from Perplexity. Perplexity's AI browser, with an assistant in every tab that can answer, shop and run errands in the background. Nearfi's Agent Index rates its autonomy 3/5 (multi-step, with approval): its assistant acts on the page and background assistants run several tasks, while the Email Assistant only drafts replies, so key outputs come back to you. Status: live; 3 incident(s) logged. Source: https://nearfi.co/agents/perplexity-comet - Category: Browser · general · Standalone - What it does: Comet launched on 9 July 2025 for top-tier subscribers and became free to download on 2 October 2025, followed by Android in November 2025 and iOS in March 2026. Its assistant can act on the page you are on, background assistants can work on several tasks at once, and an Email Assistant for Max subscribers schedules meetings and drafts replies. - Limits: It became security researchers' favourite target. Brave called indirect prompt injection "a systemic challenge facing the entire category of AI-powered browsers", and showed hidden text on a web page could take over Comet's assistant. - Incident (2025-08-20): Brave showed that instructions hidden in a Reddit comment could hijack Comet's assistant while it summarised the page. - Incident (2025-10): LayerX Security described "CometJacking", an attack that could extract a user's data to an attacker's server; Perplexity said it had found and patched the issue. - Incident (2026-08): The Ninth Circuit vacated a March 2026 injunction that had barred Comet's agent from shopping on Amazon, finding that on the current record the user, not Perplexity, was accessing Amazon's systems; the case continues. # Opera Neon (Opera) Opera Neon is a standalone general AI agent from Opera. Opera's agentic browser, with built-in agents that browse for you, run deep research and build things. Nearfi's Agent Index rates its autonomy 3/5 (multi-step, with approval): its agents carry out tasks on pages and run deep research when you ask, but with no independent test of how far it completes tasks alone, it is not rated higher. Status: live; 0 incident(s) logged. Source: https://nearfi.co/agents/opera-neon - Category: Browser · general · Standalone - What it does: Unveiled in May 2025, Neon shipped to early users in autumn 2025 and opened to the public on 11 December 2025 as a $19.90-a-month subscription. Its agents can carry out tasks on pages, run deep research and build mini apps, and "Tasks" group chats and tabs into workspaces. Since August 2026 the browser itself is free, and outside agents can drive it through MCP and a command-line tool; the paid plan unlocks Neon's own autonomous browsing. - Limits: We found no independent test of how reliably its agents complete tasks. # ChatGPT deep research (OpenAI) ChatGPT deep research is a standalone vertical AI agent from OpenAI. Ask a question and it spends minutes reading the web, then returns a long report with citations. Nearfi's Agent Index rates its autonomy 4/5 (multi-step on its own): it browses for five to thirty minutes, cross-checks sources and writes a cited report on its own for you to review; you can steer it, but it acts only when asked. Status: live; 0 incident(s) logged. Source: https://nearfi.co/agents/openai-deep-research - Category: Research · vertical · Standalone - What it does: Deep research browses for roughly five to thirty minutes, reading and cross-checking sources, and writes a cited report. A February 2026 update moved it to a GPT-5.2-based model, let users restrict it to chosen sites, connect apps and data through MCP servers, and steer it mid-run. When OpenAI published its BrowseComp benchmark in April 2025, deep research solved 51.5% of its hard-to-find questions; the paper noted it had been trained for that kind of task. - Limits: OpenAI acknowledges it sometimes makes factual errors and wrong inferences. An April 2026 study found that 3–13% of citation URLs from commercial models and deep research agents had no record in the Wayback Machine and likely never existed, and that deep research agents fabricated URLs at higher rates. # Gemini Deep Research (Google) Gemini Deep Research is a standalone vertical AI agent from Google. Gemini's research mode: it plans a search, reads widely and writes a cited report. Nearfi's Agent Index rates its autonomy 4/5 (multi-step on its own): given a question, it browses and writes a cited report on its own for you to check afterwards; it acts only when asked, so not a 5. Status: live; 0 incident(s) logged. Source: https://nearfi.co/agents/gemini-deep-research - Category: Research · vertical · Standalone - What it does: Deep Research launched for Gemini Advanced subscribers on 11 December 2024. On 11 December 2025 Google released a stronger version built on Gemini 3 Pro, available to developers through the Interactions API and as an upgrade in the Gemini app, and claimed 59.2% on OpenAI's BrowseComp benchmark. - Limits: The benchmark figure is Google's own. An April 2026 study found that 3–13% of citation URLs from commercial models and deep research agents likely never existed, with deep research agents fabricating at higher rates. # Perplexity Deep Research (Perplexity) Perplexity Deep Research is a standalone vertical AI agent from Perplexity. Perplexity's research mode: dozens of searches, hundreds of sources and a report in a few minutes. Nearfi's Agent Index rates its autonomy 4/5 (multi-step on its own): it runs dozens of searches, refines its plan and writes a report on its own for you to check afterwards; it acts only when asked, so not a 5. Status: live; 0 incident(s) logged. Source: https://nearfi.co/agents/perplexity-deep-research - Category: Research · vertical · Standalone - What it does: Launched on 14 February 2025, Deep Research runs dozens of searches, reads hundreds of sources, refines its plan as it goes and writes a report that can be exported or turned into a Perplexity Page. Perplexity claimed 21.1% on Humanity's Last Exam at launch. In February 2026 it moved to Claude Opus 4.6, first for Max users. - Limits: Its benchmark figures are Perplexity's own. An April 2026 study found that 3–13% of citation URLs from commercial models and deep research agents likely never existed, with deep research agents fabricating at higher rates. # Copilot Cowork (Microsoft) Copilot Cowork is an in-app general AI agent from Microsoft. Microsoft 365's long-running agent: one request, and it works through Outlook, Teams, Excel and your company's files. Nearfi's Agent Index rates its autonomy 4/5 (multi-step on its own): it takes a goal, plans the steps and carries them out across Microsoft 365 in the cloud, continuing when your laptop is off, but runs per task rather than continuously. Status: live; 0 incident(s) logged. Source: https://nearfi.co/agents/microsoft-copilot-cowork - Category: Work & office · general · In-app - What it does: Announced on 9 March 2026 and built "in close collaboration with Anthropic" on the same harness as Claude Cowork, it takes a goal such as preparing a meeting pack, plans the steps and carries them out across Microsoft 365, running in the cloud so tasks continue when your laptop is off. It became generally available on 16 June 2026, running on Anthropic models, and absorbed the computer-use feature from Microsoft's Researcher agent, which was retired on 1 July. - Limits: Each task's price depends on model use, tool calls and run time, so costs are hard to predict; Microsoft's claim that it is 30–40% cheaper than Claude Cowork comes from its own testing. Web browsing was limited to the Frontier programme at launch. # Microsoft Scout (Microsoft) Microsoft Scout is an in-app vertical AI agent from Microsoft. Microsoft's always-on work agent, announced 2 June 2026, which schedules meetings, prepares materials and flags risks across Teams and Outlook. Nearfi's Agent Index rates its autonomy 4/5 (multi-step on its own): Microsoft says it works in the background across Teams, Outlook and calendars without being prompted each time, but it is in private preview and we found no hands-on test, so we rate what it can be shown to do, with sensitive actions open to human sign-off; not 5 until its unprompted running is seen in use. Status: beta; 0 incident(s) logged. Source: https://nearfi.co/agents/microsoft-scout - Category: Work & office · vertical · In-app - What it does: Scout runs under its own governed Entra identity and works across Teams, Outlook, OneDrive and SharePoint, with the desktop app extending it to the browser, local resources and MCP servers. Microsoft says it can coordinate meetings across time zones, block calendar time for deliverables and prepare meeting materials, and that Purview policies are applied before anything is sent or written. It is based on the open-source OpenClaw. - Limits: Access needs Frontier enrolment, Intune policy configuration, an opt-in attestation and a GitHub Copilot licence to install the desktop app. PC World's piece contained no hands-on testing, and Microsoft's page gives no price or usage limits. # Notion Agents (Notion) Notion Agents is an in-app vertical AI agent from Notion. Notion's built-in agent that creates docs and databases, searches your tools and runs multi-step work for you. Nearfi's Agent Index rates its autonomy 4/5 (multi-step on its own): it runs multi-step work for 20 minutes or more across many pages and notifies you when done, so you check the result afterwards; Custom Agents on schedules were still 'coming soon' on the launch page. Status: live; 0 incident(s) logged. Source: https://nearfi.co/agents/notion-agents - Category: Work & office · vertical · In-app - What it does: Notion 3.0, released on 18 September 2025, rebuilt Notion AI as agents. Notion says anything you can do in Notion, your agent can do, working for over 20 minutes across many pages with a memory kept in Notion pages. Custom Agents that run on schedules or triggers were announced as coming soon. - Limits: Notion's release page does not say which plans include agents, and gives no usage quotas. Custom Agents were not yet available at launch; later release notes suggest they have since shipped, but we have not verified their current terms. # Reclaim (Reclaim.ai) Reclaim is an vertical AI agent from Reclaim.ai. A calendar tool that schedules your tasks, habits, breaks and meetings around existing events and reschedules as things change. Nearfi's Agent Index rates its autonomy 4/5 (multi-step on its own): it places tasks, habits and focus time into your calendar and adapts the plan as your schedule changes, but only within the calendars you connect. Status: live; 0 incident(s) logged. Source: https://nearfi.co/agents/reclaim-ai - Category: Work & office · vertical · - What it does: Reclaim works inside Google Calendar or Outlook Calendar. It places tasks, habits and focus time into free slots by priority, finds meeting times across attendees, and adapts the plan when your schedule changes. - Limits: It manages time, not email or other apps, and it works only on the calendars you connect. Dropbox did not disclose the terms of its acquisition. # ServiceNow Arc (ServiceNow) ServiceNow Arc is an in-app vertical AI agent from ServiceNow. A long-running personal work agent for employees that plans and carries out assignments across email, Teams and ServiceNow. Nearfi's Agent Index rates its autonomy 4/5 (multi-step on its own): it takes a goal, plans, acts across work apps, checks its own output and can run for hours or days, within IT-set permissions; it is not 5 because it works from assigned goals. Status: live; 0 incident(s) logged. Source: https://nearfi.co/agents/servicenow-arc - Category: Work & office · vertical · In-app - What it does: Arc, generally available since 21 October 2026 as part of ServiceNow EmployeeWorks, takes a goal, builds a plan, acts, checks its progress and surfaces unfinished work from the inbox, backlog and meetings. If no connector fits it can write an integration. IT sets policies in the Super Intelligence Control Tower, enforced by NVIDIA OpenShell. - Limits: Some governance pieces, including AI Passport enforcement, are not yet available. eWEEK noted that cited research does not establish Arc's accuracy or savings, and errors could leave approvals unfinished or records wrong. Windows support has not shipped. # Intuit Assist agents for QuickBooks (Intuit) Intuit Assist agents for QuickBooks is an in-app vertical AI agent from Intuit. A team of agents inside QuickBooks that categorise transactions, chase invoices and draft sales emails. Nearfi's Agent Index rates its autonomy 3/5 (multi-step, with approval): the Payments, Accounting and Customer agents categorise transactions, send invoices and reminders and schedule meetings in a feed the owner reviews; Intuit says the customer is always in control, and approval steps are not described, so it is not rated higher. Status: live; 0 incident(s) logged. Source: https://nearfi.co/agents/intuit-assist-quickbooks-agents - Category: Work & office · vertical · In-app - What it does: Announced on 1 July 2025, the agents include Accounting (bookkeeping, categorisation, reconciliation), Payments (tracking invoices and sending reminders), Finance (reporting and forecasting) and Customer (finding leads, drafting emails, scheduling meetings). A QuickBooks feed summarises completed agent tasks. - Limits: Intuit's 'paid about five days faster' figure is its own, from US beta customers. Payroll and project-management agents were announced as coming later. Intuit's release does not say which agent actions need your approval; check this before use. # Shopify Sidekick (Shopify) Shopify Sidekick is an in-app vertical AI agent from Shopify. A built-in assistant for Shopify merchants that builds automations and admin changes from plain-language requests. Nearfi's Agent Index rates its autonomy 3/5 (multi-step, with approval): it can build automations and custom admin apps from a prompt, but Shopify's developer docs say the merchant confirms changes, so it stops for approval. Status: live; 0 incident(s) logged. Source: https://nearfi.co/agents/shopify-sidekick - Category: Work & office · vertical · In-app - What it does: Sidekick works inside the Shopify admin with the permissions of the logged-in user. In December 2025 Shopify added the ability to build custom admin apps and visual automation workflows from a prompt, and to save reusable prompts as 'skills'. Shopify's developer documentation says Sidekick suggests changes and the merchant confirms them. - Limits: Shopify's docs say Sidekick does not change the store on its own from a launched prompt, and it is not available on every store. A retail analyst quoted by Retail Brew said merchants will want reliability to improve before trusting it beyond low-stakes tasks. Launch date and pricing not verified. # Superhuman Go (Superhuman) Superhuman Go is an in-app general AI agent from Superhuman. An assistant that works inside the apps you already use, sorting email, drafting replies in your voice and running custom agents on a schedule. Nearfi's Agent Index rates its autonomy 3/5 (multi-step, with approval): it sorts mail, drafts replies and runs custom agents on a schedule or an event, but the maker says nothing happens without your approval. Not 4, because you decide what is sent. Status: live; 1 incident(s) logged. Source: https://nearfi.co/agents/superhuman-go - Category: Work & office · general · In-app - What it does: Go offers writing help wherever you type, sorts incoming mail and pre-drafts replies using your calendar and connected apps, and helps prepare and follow up meetings. Paid plans let you build no-code agents that run as you type, on a schedule or after an event, and connect apps through MCP connectors. - Limits: The product page doesn't detail which calendar actions are available. In January 2026 PromptArmor showed a prompt-injection flaw in Superhuman's email AI that could leak inbox contents; it has been fixed. - Incident (2026-01-15): Security firm PromptArmor reported that a malicious email could make Superhuman's AI send the contents of other emails to an attacker via a Google Forms URL; the reporting says it has been fixed. # Claude Code (Anthropic) Claude Code is a standalone vertical AI agent from Anthropic. A coding agent that reads your codebase, edits files and runs commands from the terminal or editor. Nearfi's Agent Index rates its autonomy 4/5 (multi-step on its own): it reads a codebase, edits files, runs tests and can commit and push without step-by-step prompting, but you check the result of long tasks, and it does not run unprompted. Status: live; 1 incident(s) logged. Source: https://nearfi.co/agents/claude-code - Category: Coding · vertical · Standalone - What it does: Claude Code searches and reads a codebase, edits files, writes and runs tests, uses command-line tools and can commit and push to GitHub. It began as a limited research preview in February 2025 and became generally available that May. Anthropic's newest model, Claude Opus 5.5 (22 September 2026), comes with a classifier that Anthropic says screens every action before it runs. - Limits: It acts with whatever permissions it is given, so a mistaken command can be destructive, and its own safety claims come from Anthropic's tests rather than independent ones. Long tasks still need a person to check the result. - Incident (2025-11-13): Anthropic reported that a group it assessed with high confidence to be Chinese state-sponsored had used Claude Code to do 80–90% of the work in an espionage campaign against about 30 targets, succeeding in a small number of cases; Anthropic banned the accounts as it identified them. # Codex (OpenAI) Codex is a standalone vertical AI agent from OpenAI. OpenAI's coding agent, which works through tasks in its own sandbox and hands back the changes with logs and test results. Nearfi's Agent Index rates its autonomy 4/5 (multi-step on its own): each task runs in its own container and returns changes with logs and test output; it completes the work alone, but a person must verify and merge it. Status: live; 0 incident(s) logged. Source: https://nearfi.co/agents/openai-codex - Category: Coding · vertical · Standalone - What it does: Codex launched in May 2025 as a research preview of a cloud agent: each task runs in an isolated container preloaded with your repository, and the agent returns its changes with terminal logs and test output for you to review. There is also an open-source command-line version. Codex now sits in the ChatGPT desktop app alongside ChatGPT Work, and OpenAI said in June 2026 that it had more than 5 million weekly active users. - Limits: At launch it could not fully test applications with live user interfaces, and a person still has to verify and merge its work. Its usage figures are OpenAI's own and have not been independently checked. # Cursor (Anysphere) Cursor is a standalone vertical AI agent from Anysphere. An AI code editor whose agent writes, edits and runs code across your project. Nearfi's Agent Index rates its autonomy 4/5 (multi-step on its own): its agent changes many files and runs commands for you, completing tasks for you to review, but it relies on people approving privileged operations and does not act unprompted. Status: live; 2 incident(s) logged. Source: https://nearfi.co/agents/cursor - Category: Coding · vertical · Standalone - What it does: Cursor is a code editor built around an AI agent that can change many files and run commands for you, using its own Composer models or models from other labs. SpaceX completed its all-stock acquisition of Anysphere, valued at about $60 billion, on 14 August 2026, and the team joined SpaceXAI. Cursor says the deal gives it access to SpaceX's GPU fleet to train cheaper, more capable models. - Limits: Its agent can take destructive actions with whatever credentials it can reach, and Cursor's own guidance relies on people approving privileged operations. It has also been slow to disclose what its models are built on. - Incident (2026-03-23): Cursor confirmed that its Composer 2 model was built on Moonshot AI's Kimi K2.5, which its launch post had not mentioned; co-founder Aman Sanger called the omission "a miss". - Incident (2026-04-25): An agent in Cursor running Anthropic's Claude Opus 4.6 deleted the startup PocketOS's production database and its volume-level backups in a single API call to its host, Railway, which later restored the data. # Devin (Cognition) Devin is a standalone vertical AI agent from Cognition. An autonomous software-engineering agent you hand tickets to, now paired with Cognition's desktop editor. Nearfi's Agent Index rates its autonomy 4/5 (multi-step on its own): it takes a task and works through it in its own cloud environment without step-by-step input, but its work still needs reviewing, and it does not act unasked. Status: live; 0 incident(s) logged. Source: https://nearfi.co/agents/devin - Category: Coding · vertical · Standalone - What it does: Cognition introduced Devin in March 2024 as "the first AI software engineer", an agent that takes a task and works through it in its own cloud environment. Devin 2.0, in April 2025, cut the entry price from about $500 a month to $20. After buying Windsurf in July 2025, Cognition renamed that editor Devin Desktop in June 2026, so local and cloud agents can be managed from one place. Cognition raised more than $2 billion at a $48 billion valuation in September 2026. - Limits: The "software engineer" label sets expectations that an agent which still needs its work reviewed does not meet. Cognition has not published independent evaluations of how often Devin completes real tasks unaided. # Google Antigravity (Google) Google Antigravity is a standalone vertical AI agent from Google. Google's agent-first development platform, where several agents plan and build across the editor, terminal and browser. Nearfi's Agent Index rates its autonomy 4/5 (multi-step on its own): its agents work across the editor, terminal and browser in parallel and report back with task lists and recordings, and can run scheduled background tasks; you review afterwards. Status: live; 1 incident(s) logged. Source: https://nearfi.co/agents/google-antigravity - Category: Coding · vertical · Standalone - What it does: Antigravity launched in November 2025 alongside Gemini 3 as a code editor with a separate manager for running several agents at once; they work across the editor, terminal and browser and report back with task lists, screenshots and recordings. Antigravity 2.0, at I/O in May 2026, added a desktop app that can schedule background tasks, a command-line tool that replaces Gemini CLI, and an SDK. - Limits: Agents that browse the web and run shell commands can be steered by what they read, and Google lists browser-agent data exfiltration through prompt injection as a known issue. Its usage limits are tied to Google's consumer subscription tiers. - Incident (2025-11-25): Researchers at PromptArmor showed that hidden text on a web page could lead Antigravity's agent to collect credentials and code from a user's workspace and send them to an outside site. # Grok Build (xAI (SpaceXAI)) Grok Build is a standalone vertical AI agent from xAI (SpaceXAI). xAI's terminal coding agent, which plans, edits files and runs commands, with subagents working in parallel. Nearfi's Agent Index rates its autonomy 4/5 (multi-step on its own): after you approve or rewrite its plan, it edits files, runs shell commands and delegates to subagents, and can run headless in scripts; it does not act unprompted. Status: beta; 1 incident(s) logged. Source: https://nearfi.co/agents/grok-build - Category: Coding · vertical · Standalone - What it does: Grok Build is a command-line agent that drafts a plan you can approve or rewrite, then edits files and runs shell commands, delegating parts of larger jobs to subagents that run at the same time. It can also run headless inside scripts and works with existing AGENTS.md files, hooks and MCP servers. After a privacy row in July 2026 its source code was published on GitHub under the Apache 2.0 licence. - Limits: It launched as an early beta and its data handling was found wanting within two months. The model and harness are less proven than the established coding agents. - Incident (2026-07-13): Researchers reported that Grok Build had uploaded entire Git repositories, which could include credentials and history, to xAI-linked cloud storage, far more than its tasks needed; the uploads were later switched off by a server-side flag. # Lovable (Lovable) Lovable is a standalone vertical AI agent from Lovable. A chat-based app builder that writes, hosts and updates web apps for people who don't code. Nearfi's Agent Index rates its autonomy 4/5 (multi-step on its own): from a description it builds a working web app with hosting and a database, without you approving each step; you then check and refine it, and it does not act unasked. Status: live; 2 incident(s) logged. Source: https://nearfi.co/agents/lovable - Category: Coding · vertical · Standalone - What it does: You describe what you want and Lovable's agent builds a working web app, with hosting and a database, then changes it as you keep chatting; a plan mode lets you agree the approach first. Lovable says more than 60 million projects have been made since it launched in November 2024. It raised $400 million at a $13.3 billion valuation in August 2026. - Limits: Many of its users cannot check the security of what it builds, and twice researchers have found data exposed. Credits are spent per change, so heavy iteration costs more than the headline price suggests. - Incident (2025-05-30): A vulnerability was recorded (CVE-2025-48757) in which insufficient row-level security in Lovable-generated sites let unauthenticated attackers read or write their database tables; Lovable disputes it, saying each customer is responsible for protecting their own app's data. - Incident (2026-04-20): Researchers reported that the chat history and source code of public Lovable projects had been readable by other users with a link since 3 February 2026; Lovable fixed it within two hours and later said its product and its initial response "missed the mark". # Replit Agent (Replit) Replit Agent is a standalone vertical AI agent from Replit. Describe an app in plain English and Replit's agent builds, configures and deploys it. Nearfi's Agent Index rates its autonomy 4/5 (multi-step on its own): from a short description it sets up the environment, writes and runs code and deploys the result without step-by-step approval, but it acts on request, not continuously. Status: live; 1 incident(s) logged. Source: https://nearfi.co/agents/replit-agent - Category: Coding · vertical · Standalone - What it does: Replit Agent sets up the development environment, installs dependencies, writes and runs code and deploys the result from a short description. Agent 4, released in March 2026, added a design canvas, a planning mode, several tasks running in parallel and connections to tools such as BigQuery, Linear, Slack and Notion. Replit raised $400 million at a $9 billion valuation the same month. - Limits: It is aimed at people who may not read the code it writes, so mistakes can go unnoticed until they matter. Costs scale with the agent's effort, which is hard to predict in advance. - Incident (2025-07): During investor Jason Lemkin's test, Replit's agent deleted a production database despite an instruction to freeze changes; chief executive Amjad Masad apologised, and Replit began rolling out automatic separation of development and production databases. # Copilot cloud agent (GitHub) Copilot cloud agent is an in-app vertical AI agent from GitHub. Assign it a GitHub issue and it works in its own environment, then hands back a branch or a draft pull request. Nearfi's Agent Index rates its autonomy 3/5 (multi-step, with approval): it works on a task in its own environment and can propose a plan for approval, but returns a branch or draft pull request and nothing merges without a person. Status: live; 0 incident(s) logged. Source: https://nearfi.co/agents/github-copilot-cloud-agent - Category: Coding · vertical · In-app - What it does: Formerly called Copilot coding agent, it takes a task, works on it in a separate environment and returns a branch or draft pull request for review. Since April 2026 it can also research a codebase and propose a plan for approval before it writes code. Through Agent HQ, announced in October 2025, GitHub also brings coding agents from Anthropic, OpenAI, Google, Cognition and xAI inside GitHub. - Limits: Nothing merges without a person, which is the point, but review becomes the bottleneck. GitHub said in April 2026 that its previous pricing was "no longer sustainable" and moved to metered credits, so the cost of a long session is harder to predict. # Gemini CLI (Google) Gemini CLI is a standalone vertical AI agent from Google. Google's open-source terminal agent, now being replaced by Antigravity CLI for most individual users. Nearfi's Agent Index rates its autonomy 3/5 (multi-step, with approval): it edits files, runs commands and troubleshoots across steps in the terminal, working on what you ask within a session rather than completing larger tasks unattended. Status: live; 1 incident(s) logged. Source: https://nearfi.co/agents/gemini-cli - Category: Coding · vertical · Standalone - What it does: Gemini CLI brought Gemini into the terminal for coding, editing files, running commands and troubleshooting, with Google Search grounding and MCP support, under the Apache 2.0 licence. At launch a personal Google account got 1,000 free requests a day. At I/O on 19 May 2026 Google announced it was moving users to Antigravity CLI and that Gemini CLI would stop serving free users and Google AI Pro and Ultra subscribers on 18 June 2026. - Limits: For most individuals it has been retired, and Google said its successor would not have full feature parity at first. It reads project files as instructions, which made it open to prompt injection. - Incident (2025-07-28): Security firm Tracebit disclosed a flaw that let instructions hidden in a repository's README or GEMINI.md file run commands that could silently send environment variables to a remote server; Google fixed it in version 0.1.14 on 25 July 2025. # Jules (Google) Jules is a standalone vertical AI agent from Google. Google's asynchronous coding agent: give it a repository and a task, and it comes back with a plan, a diff and a pull request. Nearfi's Agent Index rates its autonomy 3/5 (multi-step, with approval): it plans and makes code changes in a cloud VM, but waits for your approval at the plan and again at the diff before opening a pull request. Status: live; 0 incident(s) logged. Source: https://nearfi.co/agents/google-jules - Category: Coding · vertical · Standalone - What it does: Jules clones a GitHub repository into a cloud virtual machine, drafts a plan with Gemini, makes the changes and presents a diff for you to approve before it opens a pull request. You can start it from a prompt or by labelling a GitHub issue. It went into public beta in May 2025 and came out of beta on 6 August 2025. - Limits: It works only through GitHub and waits for approval at the plan and the pull request, so it suits chores better than open-ended projects. Google publishes daily task caps rather than any measure of how often it succeeds. # Windsurf (Windsurf) Windsurf is a standalone vertical AI agent from Windsurf. An agentic code editor, now renamed Devin Desktop after a contested 2025 sale. Nearfi's Agent Index rates its autonomy 3/5 (multi-step, with approval): its agent, Cascade, edited code and ran commands in the editor as you worked, rather than completing whole tasks unattended in its own environment. Status: acquired; 0 incident(s) logged. Source: https://nearfi.co/agents/windsurf - Category: Coding · vertical · Standalone - What it does: Windsurf was an AI code editor whose agent, Cascade, edited code and ran commands. In 2025 OpenAI's reported $3 billion deal to buy it lapsed; Google then paid about $2.4 billion to license its technology and hire its chief executive, Varun Mohan, co-founder Douglas Chen and some researchers; and on 14 July 2025 Cognition agreed to buy the rest, including the product, brand and remaining staff. On 2 June 2026 Cognition renamed it Devin Desktop and replaced Cascade with Devin Local; legacy Cascade ran until 1 July 2026. - Limits: As a separate product it no longer exists; its users are now on Cognition's roadmap. Google's deal took the founders and some researchers but left most of the roughly 250-person team behind. # Agentforce (Salesforce) Agentforce is a platform vertical AI agent from Salesforce. Salesforce's platform for AI agents that handle customer service and sales tasks inside its CRM. Nearfi's Agent Index rates its autonomy 5/5 (ongoing, unprompted): companies deploy its agents to answer customers continuously and take actions on their Salesforce data, without staff approving each conversation. Status: live; 1 incident(s) logged. Source: https://nearfi.co/agents/salesforce-agentforce - Category: Customer service · vertical · Platform - What it does: Agentforce lets companies build and deploy agents that answer customers and take actions on their Salesforce data; Agentforce Service Agent was the first to become generally available, in October 2024, priced at $2 per conversation. Salesforce has since added metered Flex Credits and per-user licences, and in September 2026 bought Fin to sit alongside it. MarTech reported in June 2026 that Agentforce had reached $1.2 billion in annual recurring revenue in Salesforce's first quarter of fiscal 2027. - Limits: It is strongest where a company's data already lives in Salesforce, and its pricing has changed several times, making costs hard to compare. Agents that read customer-submitted text can be manipulated by it. - Incident (2025-09-26): Noma Security disclosed "ForcedLeak", a prompt-injection flaw in which instructions planted in a web-to-lead form could make Agentforce send CRM lead data to an expired domain still on Salesforce's allow-list; Salesforce had enforced trusted-URL controls from 8 September. # Decagon (Decagon) Decagon is a platform vertical AI agent from Decagon. AI agents that handle a company's customer support across chat, email and voice, and take actions such as refunds. Nearfi's Agent Index rates its autonomy 5/5 (ongoing, unprompted): deployed by companies, its agents handle customer conversations continuously and take actions such as refunds without staff approving each one, under always-on quality checks. Status: live; 0 incident(s) logged. Source: https://nearfi.co/agents/decagon - Category: Customer service · vertical · Platform - What it does: Decagon's agents learn from a company's help content and past conversations and connect to its systems to act, for example processing a refund. Teams set the agents' workflows, test them in simulations and A/B experiments, and monitor them with always-on quality checks. It raised $250 million at a $4.5 billion valuation in January 2026. - Limits: Its performance claims come from the company and its customers; we found no independent evaluation. Like any deployed support agent, it speaks for the company, which is answerable for its mistakes. # Fin (Fin (formerly Intercom)) Fin is a platform vertical AI agent from Fin (formerly Intercom). A customer-service agent that answers and resolves support queries across chat, email, phone and messaging apps. Nearfi's Agent Index rates its autonomy 5/5 (ongoing, unprompted): companies deploy it to answer every support query continuously across helpdesks, resolving most without a person, and it hands off rather than waiting for approval on each reply. Status: live; 0 incident(s) logged. Source: https://nearfi.co/agents/fin - Category: Customer service · vertical · Platform - What it does: Intercom launched Fin in March 2023 on GPT-4; it now runs on the company's own support model, Apex, and works with Intercom or other helpdesks such as Salesforce and HubSpot. Intercom renamed itself Fin in May 2026, and Salesforce completed its purchase of the company, agreed at about $3.6 billion, on 10 September 2026. Salesforce says Fin serves more than 30,000 companies and resolves 76% of queries on average. - Limits: The 76% figure is the vendor's own average; results depend on the quality of each company's help content. Billing per outcome counts some handoffs and disqualifications as well as resolutions. # Sierra (Sierra) Sierra is a platform vertical AI agent from Sierra. Customer-service agents that big companies deploy to answer customers and resolve their problems by chat and phone. Nearfi's Agent Index rates its autonomy 5/5 (ongoing, unprompted): its branded agents handle customer conversations continuously for companies and take actions such as changing orders or processing claims without per-case approval. Status: live; 0 incident(s) logged. Source: https://nearfi.co/agents/sierra - Category: Customer service · vertical · Platform - What it does: Sierra builds branded agents that handle customer conversations and take actions in a company's systems, such as changing an order or processing a claim. Founded by Bret Taylor and Clay Bavor, it launched in February 2024, raised $950 million at a $15.8 billion valuation in May 2026, and said on 30 October 2026 that it had passed $400 million in annual recurring revenue and works with almost half of the Fortune 50. In 2024 it published τ-bench, a public test on which the best agent then succeeded less than half the time. - Limits: Its own τ-bench showed that consistency is the hard part: the best 2024 agent got the same retail task right eight times running only about a quarter of the time. When the deploying company's interests and the customer's differ, the agent serves the company, which remains responsible for what it says. # Meta Business Agent (Meta) Meta Business Agent is a platform vertical AI agent from Meta. An agent that answers shoppers on WhatsApp, Messenger and Instagram, recommends products, books appointments and closes sales for a business. Nearfi's Agent Index rates its autonomy 4/5 (multi-step on its own): it books appointments and can close sales in a customer's chat without a person, but businesses choose when staff step in, and some reports say closing depends on how the business configures it, so it is not 5. Status: live; 0 incident(s) logged. Source: https://nearfi.co/agents/meta-business-agent - Category: Customer service · vertical · Platform - What it does: Announced on 3 June 2026, it builds on a version already used by more than a million businesses on WhatsApp and Messenger. It replies in the customer's language in the business's tone, qualifies leads and books appointments. A separate platform connects agents to systems such as Shopify, Zendesk and Shopee. - Limits: Meta's own post gives no usage limits, and market research, calendar and competitive-intelligence features are plans, not shipped. The morning briefing is limited to a select group of businesses. No independent testing of its sales or errors was found. # Legora Agent (Legora) Legora Agent is a platform vertical AI agent from Legora. The agent at the centre of Legora's legal platform: it reviews, researches and drafts across a matter and hands the lawyer a finished draft to approve. Nearfi's Agent Index rates its autonomy 4/5 (multi-step on its own): Legora says it plans, executes, reviews and delivers multi-step legal work on its own, with the lawyer approving the deliverable; not 5, because it works on tasks it is given and the lawyer makes the final calls. Status: live; 0 incident(s) logged. Source: https://nearfi.co/agents/legora-agent - Category: Legal · vertical · Platform - What it does: Announced on 7 May 2026 as part of Legora's aOS, the Agent chooses among Legora's tools to search, draft, redline and review, and can produce Word, PowerPoint or Excel deliverables. Legora's own example is reviewing a counterparty's contract redline overnight, flagging issues and drafting a reply for the lawyer in the morning. It connects to document management systems, research databases and email, and each deployment is configured with Legora's embedded 'Legal Engineers'. - Limits: Nearly everything we found is Legora's own description: the pages we opened give no price, no accuracy figures and no list of limitations, and the trade press piece we opened had no independent performance evidence. Legora says the lawyer 'makes the calls that matter'. # CoCounsel Legal (Thomson Reuters) CoCounsel Legal is a standalone vertical AI agent from Thomson Reuters. Thomson Reuters' legal agent, which researches, drafts and checks citations against Westlaw and Practical Law. Nearfi's Agent Index rates its autonomy 3/5 (multi-step, with approval): Deep Research writes and carries out a multi-step research plan and drafts documents, but lawyers remain responsible for checking its work before it is used. Status: live; 0 incident(s) logged. Source: https://nearfi.co/agents/cocounsel-legal - Category: Legal · vertical · Standalone - What it does: CoCounsel Legal launched in August 2025 with Deep Research, which writes a multi-step research plan, carries it out and returns a cited report grounded in Westlaw and Practical Law. The next generation, released in the US on 20 August 2026 and built on Anthropic's Claude Agent SDK, adds a brief-drafting tool, a drafting agent in Word, review of up to 10,000 documents at once and a check of whether cited authority supports each assertion. - Limits: Its answers are grounded in Thomson Reuters' own databases, which helps with citations but limits it to what those sources cover. We found no independent evaluation of its accuracy, and lawyers remain responsible for checking its work. # Harvey (Harvey) Harvey is a standalone vertical AI agent from Harvey. An AI platform for lawyers whose agents draft, review and research legal work. Nearfi's Agent Index rates its autonomy 3/5 (multi-step, with approval): its agents carry out multi-step work such as due diligence and contract review, but lawyers remain responsible for anything filed, so output is checked before it is used. Status: live; 0 incident(s) logged. Source: https://nearfi.co/agents/harvey - Category: Legal · vertical · Standalone - What it does: Harvey offers purpose-built and custom agents for work such as due diligence, contract review, drafting and legal research, including inside Microsoft Word. In March 2026 it said more than 100,000 lawyers at about 1,300 organisations used it and that over 25,000 custom agents ran on the platform. In September 2026 it raised $550 million at a $15.5 billion valuation and said 80% of Am Law 100 firms use it. - Limits: We found no independent evaluation of its accuracy, in a profession where accuracy is the product. Lawyers remain responsible for anything filed in their name. # Nous (Nous) Nous is a standalone vertical AI agent from Nous. A UK household agent that finds cheaper energy, broadband, mobile and insurance deals, switches you, and cancels what you don't use. Nearfi's Agent Index rates its autonomy 4/5 (multi-step on its own): it watches contracts, switches deals, cancels subscriptions and books repairs within permissions you set; how much it does without asking varies by household, so not a clear 5. Status: live; 0 incident(s) logged. Source: https://nearfi.co/agents/nous - Category: Money & bills · vertical · Standalone - What it does: You talk to Nous over WhatsApp and give it your bills and logins. It watches contracts and renewals, moves households to cheaper energy, broadband, mobile and car insurance deals, cancels unwanted subscriptions, fills in forms and books repairs, working within permissions you set. It says almost 250,000 UK homes use it, and it is registered with the FCA, which lets it handle regulated products such as insurance. - Limits: Its headline saving, £734 for a typical household, is the company's own figure. It uses fixed rules rather than its language models for money decisions, which it says reduces errors; how much it does without asking depends on the permissions each household gives it. # Rocket Money (Rowan and bill negotiation) (Rocket Money) Rocket Money (Rowan and bill negotiation) is an in-app vertical AI agent from Rocket Money. A money app whose agent texts you when it spots a saving and acts on your reply, from cancelling a trial to haggling over a bill. Nearfi's Agent Index rates its autonomy 3/5 (multi-step, with approval): Rowan watches spending around the clock, but acts only on your text reply, cancelling or negotiating once you say so, so each action waits for your approval. Status: beta; 0 incident(s) logged. Source: https://nearfi.co/agents/rocket-money-rowan - Category: Money & bills · vertical · In-app - What it does: Rocket Money's negotiation team contacts providers such as internet, cable and phone companies for you and charges a share of the savings only if it succeeds. Rowan, launched on 25 August 2026 and built with Anthropic, watches your spending around the clock, texts you when a free trial is about to charge or a bill rises, and acts on a plain-language reply: cancelling a subscription, negotiating a bill or setting up savings transfers. - Limits: A one-word reply can trigger an action, so the main safeguard is the text exchange, backed by what Rocket Money calls human verification by its staff. Its negotiation fee can take more than half of the first year's saving. # Cleo Autopilot (Cleo) Cleo Autopilot is an in-app vertical AI agent from Cleo. A money app's plan-and-act mode, meant to move your money towards a goal within rules you approve. Nearfi's Agent Index rates its autonomy 2/5 (acts when asked): it sets a plan and gives daily guidance, acting only with your consent on a plan you approved; at launch its Actions only recommended moves, so it is not yet a 3. Status: waitlist; 1 incident(s) logged. Source: https://nearfi.co/agents/cleo-autopilot - Category: Money & bills · vertical · In-app - What it does: Cleo is a budgeting and cash-advance app with a chatty AI persona. Autopilot, announced on 5 February 2026 for US users via a waitlist, analyses your finances, sets a plan towards a goal and gives daily guidance, acting 'always with your consent on a plan you've already approved'. At launch Cleo said the Actions component recommended moves, and carrying them out for you was a planned capability. - Limits: By its maker's own launch description the autonomous part was still to come, and we could not verify how widely Autopilot is now available or what it does unprompted. - Incident (2025-03-27): Cleo agreed to pay $17 million to settle FTC charges that it overstated how much users could borrow through its cash advances and how quickly, and made subscriptions hard to cancel; the case concerned the cash-advance service, not Autopilot. # DoNotPay (DoNotPay) DoNotPay is a standalone vertical AI agent from DoNotPay. A self-styled 'AI consumer champion' that drafts and sends cancellations, complaints and disputes for you. Nearfi's Agent Index rates its autonomy 2/5 (acts when asked): each tool does one job when asked, such as sending a letter, cancelling a subscription or contesting a ticket, rather than planning a multi-step task. Status: live; 1 incident(s) logged. Source: https://nearfi.co/agents/donotpay - Category: Money & bills · vertical · Standalone - What it does: DoNotPay says it offers more than 100 AI-powered tools. It writes and sends letters, cancels subscriptions, contests parking tickets and issues a virtual card that declines charges when a free trial ends. Its site says it is 'not a law firm and does not provide legal advice'. - Limits: Its own site says the tools are not guaranteed to help in every situation. The FTC found that it had never tested whether its 'robot lawyer' performed like a human lawyer. - Incident (2025-02-11): The FTC finalised an order requiring DoNotPay to pay $193,000, notify people who subscribed between 2021 and 2023, and stop claiming its service performs like a real lawyer without evidence. # Klarna AI assistant (Klarna) Klarna AI assistant is an in-app vertical AI agent from Klarna. A chat assistant inside the Klarna app that handles refunds, returns and payment questions. Nearfi's Agent Index rates its autonomy 2/5 (acts when asked): it handles single requests such as refunds, returns and balance questions when asked, and does not plan multi-step tasks for you. Status: live; 0 incident(s) logged. Source: https://nearfi.co/agents/klarna-ai-assistant - Category: Money & bills · vertical · In-app - What it does: Klarna announced the assistant on 27 February 2024, built with OpenAI. It handles refunds and returns, shows real-time balances and payment schedules and explains spending limits, in more than 35 languages. In September 2024 Klarna added a chat-based shopping assistant that compares products, prices, stock, delivery and cashback. - Limits: Klarna's launch figures (2.3 million chats in a month, two-thirds of service chats, satisfaction on a par with human agents) are the company's own, and the sources we opened do not say whether the shopping assistant can complete a purchase. Customers can still choose a human agent. # Revolut AIR (Revolut) Revolut AIR is an in-app vertical AI agent from Revolut. Revolut's in-app assistant that answers questions about your money and can freeze a card, manage subscriptions or buy an eSIM. Nearfi's Agent Index rates its autonomy 2/5 (acts when asked): it carries out single actions when asked, such as freezing a card or cancelling a subscription, and doesn't plan multi-step tasks or act unprompted. Status: live; 0 incident(s) logged. Source: https://nearfi.co/agents/revolut-air - Category: Money & bills · vertical · In-app - What it does: AIR (AI by Revolut) was announced on 9 April 2026 and sits inside the Revolut app. It gives spending insights, tracks investments, manages subscriptions, freezes a lost card, helps budget a trip and buys Revolut eSIMs. Revolut says it sees only what the customer can already see in the app. - Limits: The sources we opened give no price, no list of countries and no usage limits. Revolut's statement that AIR retains no user data and that its AI partners don't store personal information or train on it is the company's own. # Public Agents (Public) Public Agents is an in-app vertical AI agent from Public. Describe a trading or cash rule in plain English and an agent watches the market and executes it inside your Public account. Nearfi's Agent Index rates its autonomy 5/5 (ongoing, unprompted): once you approve a rule, an active agent monitors markets and trades on its triggers without asking again. Public's agreement, as quoted by Walnut, says it won't alert you or wait for confirmation before each transaction. Status: waitlist; 0 incident(s) logged. Source: https://nearfi.co/agents/public-agents - Category: Finance & investing · vertical · In-app - What it does: Announced on 6 April 2026 for select members of Public's brokerage. You describe a rule; the AI asks follow-up questions, you review the logic and switch it on. Agents can trade stocks, options and crypto and move cash on triggers such as a price move or a schedule, and they log each action. They run only inside Public's platform. - Limits: Rollout is phased and web-only, with a waitlist. The coverage we opened gives no fees, caps on trade size, eligible accounts or regions. Walnut, a third party, quotes Public's agreement as saying executed transactions cannot be reversed. Public's separate Alpha assistant is research only and says it may give inaccurate responses. # Robinhood Agents (Agentic Trading) (Robinhood) Robinhood Agents (Agentic Trading) is an in-app vertical AI agent from Robinhood. An agent inside the Robinhood app, or one you connect, that researches and trades from a dedicated account you fund. Nearfi's Agent Index rates its autonomy 4/5 (multi-step on its own): with approvals switched off it places eligible trades in its funded account without confirming each one; approvals are on by default for the built-in agent, so it is not 5, and it trades only the money you put in. Status: beta; 0 incident(s) logged. Source: https://nearfi.co/agents/robinhood-agents - Category: Finance & investing · vertical · In-app - What it does: Robinhood announced Agentic Trading and an Agentic Credit Card on 27 May 2026, letting third-party agents connect through its MCP servers. The built-in Robinhood Agents, a chat agent powered by third-party language models, researches markets, analyses a portfolio and places trades in a separate Agentic account. You get alerts for each trade and can pause or disconnect the agent. - Limits: Robinhood says agents can make errors and that it does not guarantee their output or accept responsibility for agent-driven losses. It does not supervise third-party agents. The card variant needs a Gold Card; rollout of the built-in agent is partial and its full experience is in the mobile app only. # HTS Assist (Hopper Technology Solutions (Hopper)) HTS Assist is a platform vertical AI agent from Hopper Technology Solutions (Hopper). A voice and chat agent that airlines and banks deploy to cancel, rebook and refund trips for their customers. Nearfi's Agent Index rates its autonomy 4/5 (multi-step on its own): in supported scenarios it verifies the traveller, checks airline policy and processes cancellations, refunds and rebookings without a human; it is not 5 because it handles one traveller's request at a time and escalates the rest, and its live deployments are few. Status: live; 0 incident(s) logged. Source: https://nearfi.co/agents/hts-assist - Category: Travel & shopping · vertical · Platform - What it does: HTS Assist is the B2B service agent from Hopper's technology arm. It takes a call or chat, verifies the traveller, reviews the airline's policy and processes changes, cancellations and refunds in real time, and can offer a hotel after a missed connection. Partners monitor and control it from a live command centre. - Limits: Autonomous resolution applies only to supported scenarios, and multilingual voice relies on third-party integrations. Performance figures are company-reported averages; VentureBeat noted the real test, a major weather disruption, had not yet been shown, and implementation takes six to eight weeks. # Alexa for Shopping (formerly Rufus) (Amazon) Alexa for Shopping (formerly Rufus) is an in-app vertical AI agent from Amazon. Amazon's shopping assistant, which can watch a price and buy the item for you when it drops. Nearfi's Agent Index rates its autonomy 3/5 (multi-step, with approval): it builds baskets, schedules orders and runs price rules, but most actions stop at the basket for you to check out; only Auto Buy orders on its own, so it is not routinely 4. Status: live; 0 incident(s) logged. Source: https://nearfi.co/agents/alexa-for-shopping - Category: Travel & shopping · vertical · In-app - What it does: On 13 May 2026 Amazon replaced Rufus, its shopping chatbot since 2024, with Alexa for Shopping, which draws on Alexa+ and your shopping history. It answers product questions, compares items, rebuilds baskets from past orders, schedules recurring orders and runs conditional rules such as adding sunscreen to your basket if it falls to $10. Auto Buy, added to Rufus for US Prime members in November 2025, places the order itself once a target price is met; Amazon said it checks prices every 30 minutes. - Limits: It shops only where its maker sells, and some of its AI results are sponsored, so its advice and Amazon's interests overlap. Most actions stop at the basket for you to check out; Auto Buy orders can be cancelled within 24 hours, before they ship. # Amazon Buy for Me (Amazon) Amazon Buy for Me is an in-app vertical AI agent from Amazon. Lets you buy from other brands' websites inside the Amazon app, with Amazon's AI completing the order on the other site. Nearfi's Agent Index rates its autonomy 3/5 (multi-step, with approval): its agent completes the purchase on a brand's own site, but only after you confirm address, payment and fees on Amazon's checkout page, so you approve every order. Status: beta; 2 incident(s) logged. Source: https://nearfi.co/agents/amazon-buy-for-me - Category: Travel & shopping · vertical · In-app - What it does: When Amazon doesn't sell an item, search results can show products from brands' own sites. You confirm the address, payment and fees on an Amazon checkout page, and Amazon's agent completes the purchase on the brand's website using your encrypted details. The brand handles delivery, returns and customer service. Alexa for Shopping now offers it too. - Limits: Amazon can't see or support the order after it is placed, so problems go to a shop you may never have visited. The listings are built from other sites' data, and merchants have reported wrong photos and details. - Incident (2026-01-06): Independent brands told Modern Retail that their products had appeared in Buy for Me without their permission; Amazon called the programme a test and said businesses could opt out by email. - Incident (2026-01-07): Retail Gazette, citing the Financial Times, reported that four independent retailers said Amazon had listed their full inventories without consent, and two received orders for out-of-stock or mispriced items. # DoorDash text-to-order agent (DoorDash) DoorDash text-to-order agent is an in-app vertical AI agent from DoorDash. A text-message agent, working in Apple Messages, that builds your DoorDash order and checks out when you confirm. Nearfi's Agent Index rates its autonomy 3/5 (multi-step, with approval): it builds a cart and, once you confirm in the thread, checks out with your stored card; it pauses for your confirmation, and is not 4 because it will not order unprompted without your permission. Status: waitlist; 0 incident(s) logged. Source: https://nearfi.co/agents/doordash-text-agent - Category: Travel & shopping · vertical · In-app - What it does: Unveiled on 30 September 2026, it matches the texter's phone number to a DoorDash profile and uses order history, preferences and the stored payment method. You text a dish or 'order my usual'; it searches nearby restaurants, suggests a cart, and completes checkout after you confirm. With permission it can message you about a usual order before you ask. - Limits: It is in early development. Bloomberg's test of an early version found prices that differed from those in the app. Access is by waitlist, and waitlisted users get a newer version. # Google Ask for Me and Call for Me (Google) Google Ask for Me and Call for Me is an in-app vertical AI agent from Google. Google's AI phones local businesses for you to ask about prices and stock, and now, in a test, to book and reschedule. Nearfi's Agent Index rates its autonomy 3/5 (multi-step, with approval): it calls several businesses and sends you a summary, and Call for Me can book or reschedule while you watch a live transcript, so you stay in the loop rather than it acting alone. Status: live; 0 incident(s) logged. Source: https://nearfi.co/agents/google-ask-for-me - Category: Travel & shopping · vertical · In-app - What it does: In Search you describe a job, such as pet grooming, and Google's AI calls several local businesses about price and availability, then sends you a summary. It began as a Search Labs test in January 2025 and reached all US Search users in July 2025; in November 2025 Google added calls to shops about stock. In September 2026 it began testing Call for Me, in which Gemini rings a business from your own number, navigates phone menus, waits on hold and can make a reservation or reschedule an appointment while you watch a live transcript. - Limits: Businesses can opt out of the calls. Google is keeping Call for Me small because, it says, 'real-world conversations are nuanced', and we found no published figure on how often calls go wrong. # Google agentic checkout and AI Mode booking (Google) Google agentic checkout and AI Mode booking is an in-app vertical AI agent from Google. Google Search and Gemini buy things and book hotels for you, paying with Google Pay after you confirm. Nearfi's Agent Index rates its autonomy 3/5 (multi-step, with approval): it tracks prices and can buy or book on a merchant's site with Google Pay, but only once you approve, so purchases always pause for you. Status: live; 0 incident(s) logged. Source: https://nearfi.co/agents/google-agentic-checkout - Category: Travel & shopping · vertical · In-app - What it does: Agentic checkout, rolled out in November 2025, lets you track an item with a size, colour and maximum price; when it falls within budget, Google can buy it on the merchant's site with Google Pay once you approve. Since January 2026 a 'Buy' button in AI Mode and the Gemini app checks out with eligible US retailers through Google's Universal Commerce Protocol. Since 27 August 2026 AI Mode can also book hotels from partners including Booking.com, Expedia, Hilton, IHG and Marriott. - Limits: The retailer or hotel stays the merchant of record and handles any problems, so Google takes the order but not the after-sales responsibility. The same interface that shops for you also shows advertisers' 'Sponsored deal' offers. # Instacart in ChatGPT (Instacart) Instacart in ChatGPT is an in-app vertical AI agent from Instacart. An Instacart app inside ChatGPT that builds a grocery cart from a request and checks out in the chat. Nearfi's Agent Index rates its autonomy 3/5 (multi-step, with approval): it plans a shopping list and builds the cart, then the user reviews and approves the list before paying in the chat. Status: live; 0 incident(s) logged. Source: https://nearfi.co/agents/instacart-chatgpt-app - Category: Travel & shopping · vertical · In-app - What it does: Launched 8 December 2025, the app lets users plan meals, build a cart from Instacart's retail partners and pay inside ChatGPT, using OpenAI's Agentic Commerce Protocol with Stripe. Users sign in to their Instacart account and approve the list before paying. - Limits: At launch payment was by credit card only, with Apple Pay, Google Pay and native mobile checkout still to come. Reports we opened did not list regions or retailers covered. # ChatGPT Instant Checkout (OpenAI) ChatGPT Instant Checkout is an in-app vertical AI agent from OpenAI. Let US ChatGPT users buy products inside the chat, until OpenAI handed checkout back to retailers in March 2026. Nearfi's Agent Index rates its autonomy 2/5 (acts when asked): you tapped 'Buy' to purchase a single item inside ChatGPT; it completed that one step when asked but did not plan or carry out multi-step shopping for you. Status: shut down; 0 incident(s) logged. Source: https://nearfi.co/agents/chatgpt-instant-checkout - Category: Travel & shopping · vertical · In-app - What it does: From 29 September 2025, logged-in US users on Free, Plus and Pro plans could tap 'Buy' to purchase from US Etsy sellers without leaving ChatGPT, with more than a million Shopify merchants promised to follow. It ran on the Agentic Commerce Protocol, built with Stripe. In March 2026 OpenAI moved away from it: ChatGPT now concentrates on product discovery, and purchases happen on retailers' own sites or in retailer apps inside ChatGPT. - Limits: CNBC reported that six months in, product selection was limited, item data was not always current and about 30 Shopify merchants were live. Walmart found conversion three times lower for products sold directly in ChatGPT than for those sent to its own site. # Kayak AI Mode (Kayak) Kayak AI Mode is an in-app vertical AI agent from Kayak. A chat mode on Kayak that searches and compares flights, hotels and cars from a plain-language request. Nearfi's Agent Index rates its autonomy 2/5 (acts when asked): it answers a request, searches and helps you compare and book flights, hotels and cars; it does not plan and execute a whole trip unprompted. Status: live; 0 incident(s) logged. Source: https://nearfi.co/agents/kayak-ai-mode - Category: Travel & shopping · vertical · In-app - What it does: Kayak's AI Mode, powered by ChatGPT and Kayak's own data, lets users type questions and get results from hundreds of travel providers. TechCrunch reports that users can compare and book flights, hotels and cars from the chat; Skift does not say how bookings are completed. - Limits: The two reports differ on how booking works inside the chat, and we could not verify it from Kayak's own pages. At launch it was English-only, in the US, on the web rather than in the app. # Walmart Sparky (Walmart) Walmart Sparky is an in-app vertical AI agent from Walmart. Walmart's shopping agent, which recommends products, plans meals and reorders the things you buy every week. Nearfi's Agent Index rates its autonomy 2/5 (acts when asked): it recommends products and reorders regular items when asked, a single step, and there is no published detail on any purchases it makes without confirmation. Status: live; 0 incident(s) logged. Source: https://nearfi.co/agents/walmart-sparky - Category: Travel & shopping · vertical · In-app - What it does: Sparky arrived in the Walmart app in June 2025 to summarise reviews, recommend products and help plan purchases, with reordering and service booking promised. By May 2026 Walmart said shoppers could use it to reorder items they buy regularly, including in stores, and that it offered personalised replenishment and meal planning. - Limits: Its job is to sell Walmart's stock, and Walmart measures it by how much more its users spend: about 35% more per order, by the company's account. We found no published detail on which purchases it makes without a confirmation step. # Spotify DJ (Spotify) Spotify DJ is an in-app vertical AI agent from Spotify. An AI radio host that builds a running mix from your listening history and changes course when you ask. Nearfi's Agent Index rates its autonomy 2/5 (acts when asked): it acts when asked, playing music to a mood, genre or activity, but controls only the music queue inside Spotify and does not plan multi-step tasks. Status: live; 0 incident(s) logged. Source: https://nearfi.co/agents/spotify-dj - Category: Music & media · vertical · In-app - What it does: DJ plays a stream of tracks chosen from your listening history and introduces them in a synthetic voice. Since May 2025 Premium listeners can steer it by voice with a mood, genre, artist or activity, and since October 2025 by typing too. In September 2025 Spotify added a 'Talk to DJ' button on Android Auto, and in May 2026 it expanded DJ to more than 75 countries and four more languages. - Limits: It controls one thing, the music queue: Spotify says requests for podcasts or audiobooks don't work, and it launched requests as a beta. Calling it an agent is generous, since it acts only inside the app. # Alexa+ (Amazon) Alexa+ is a standalone vertical AI agent from Amazon. Amazon's rebuilt voice assistant, which orders, books and arranges things through partner services as well as answering questions. Nearfi's Agent Index rates its autonomy 3/5 (multi-step, with approval): it can order takeaway, book restaurants and rides across several steps through partner services, but you ask each time and it is limited by its partner list, so it does not run on its own. Status: live; 1 incident(s) logged. Source: https://nearfi.co/agents/alexa-plus - Category: Home & voice · vertical · Standalone - What it does: Alexa+ runs on Echo devices, in the Alexa app and at Alexa.com. Amazon says it can order takeaway, find and book restaurants, book a ride, schedule services such as home repairs, shop Amazon and manage family calendars, working through partner services. In the UK the launch partners for bookings and orders were OpenTable, Just Eat and Treatwell. - Limits: It is only as capable as its partner list: in Fortune's July 2025 hands-on it could not book a restaurant that wasn't on OpenTable, and it said Spotify was playing when it wasn't. It is far more conversational than the old Alexa, but outside the US and Canada it is still in early access. - Incident (2025-03-28): Amazon withdrew the option for some Echo devices to process voice requests locally, so recordings went to its cloud; TechCrunch linked the change to the generative AI features coming with Alexa+. # Gemini task automation on Android (Google) Gemini task automation on Android is an in-app general AI agent from Google. Gemini on Pixel 10 and Galaxy S26 phones that operates food, grocery and ride apps for you in the background. Nearfi's Agent Index rates its autonomy 3/5 (multi-step, with approval): it works through a multi-step order or ride booking in a virtual window, but asks you to tap the final buy or order button yourself; it is not 2 because it handles many steps, and not 4 because Google tells users to supervise closely. Status: beta; 0 incident(s) logged. Source: https://nearfi.co/agents/gemini-android-automation - Category: Home & voice · general · In-app - What it does: Announced on 25 February 2026, it lets you long-press the power button and ask Gemini to do something, such as book a ride home or reorder your last DoorDash meal. Gemini runs the app in a secure virtual window in the cloud, can work while you use your phone, and notifies you of progress. You can watch, take over or stop it. - Limits: It is a beta for select food, grocery and rideshare apps, and Google's footnote tells users to supervise closely and interrupt when needed. For purchases, 9to5Google reports Gemini asks you to open the app and tap the final order button yourself. At launch a reviewer was unable to accept the terms on a Pixel 10 Pro. # Gemini for Home (Google) Gemini for Home is an in-app vertical AI agent from Google. Google's replacement for Assistant in the home: it controls devices, writes automations and tells you what your cameras saw. Nearfi's Agent Index rates its autonomy 2/5 (acts when asked): it controls devices and creates automations when you ask in plain language, one request at a time; its camera summaries describe rather than act. Status: beta; 0 incident(s) logged. Source: https://nearfi.co/agents/gemini-for-home - Category: Home & voice · vertical · In-app - What it does: Gemini for Home replaces Google Assistant on Nest speakers and displays and in the Google Home app. You can control devices in plain language and ask it to create automations. On the top tier it describes camera events, answers questions about video history and sends a daily 'Home Brief'. Early access began in October 2025 and spread to 16 more countries in April 2026. - Limits: Its camera descriptions make confident mistakes: 9to5Google's reviewer found a disc golf cart described as a bicycle, and cited The Verge's report that it called a shotgun a 'garden tool'. Most features are still in early access. # Hippocratic AI agents (Hippocratic AI) Hippocratic AI agents is an vertical AI agent from Hippocratic AI. Voice agents that phone patients for hospitals, insurers and drug companies: check-ins, screening reminders, intake and appointment scheduling. Nearfi's Agent Index rates its autonomy 4/5 (multi-step on its own): its agents make patient calls such as follow-ups and check-ins without a clinician on each call and escalate to nurses; they do not diagnose or prescribe, so they stop short of a 5. Status: live; 0 incident(s) logged. Source: https://nearfi.co/agents/hippocratic-ai - Category: Health · vertical · - What it does: Hippocratic AI sells a catalogue of non-diagnostic agents to providers, payors and pharma. They make calls such as post-discharge follow-ups, medication-adherence check-ins, screening outreach and intake, and escalate patients to nurses. The company says about 0.5% to 1% of live calls are sampled for safety review. - Limits: The company says its agents do not diagnose or prescribe and it will not build agents for hospice, mental-health disorders or children under two. Its safety figures (over 250 million interactions, zero reported incidents of harm) are company-reported and we have not verified them independently. # Hyro patient communication agents (Hyro) Hyro patient communication agents is a platform vertical AI agent from Hyro. Voice, chat and SMS agents that health systems use to book appointments, manage prescriptions and answer billing questions. Nearfi's Agent Index rates its autonomy 4/5 (multi-step on its own): it resolves some requests such as scheduling end to end, with no staff step, and hands off to a person otherwise; it is not 5 because it answers patients' requests rather than acting unprompted. Status: live; 0 incident(s) logged. Source: https://nearfi.co/agents/hyro-patient-agents - Category: Health · vertical · Platform - What it does: Announced with Sutter Health on 17 September 2025, Hyro's agents take calls and messages 24/7 for appointment management, scheduling, prescription management and billing questions. They route by the patient's goal, resolve some requests end to end, and hand off to a live agent with the conversation context. Sutter's governance council set safeguards, including data sources with citations and HIPAA and SOC 2 compliance. - Limits: The vendor's press release does not say what the agents cannot do. Its performance figures, such as an 85% average drop in abandonment, come from Hyro's work across 40+ health systems, not from Sutter; no Sutter outcome data had been reported. It is not clinical advice and not a route to emergency care. # Amazon Health AI (Amazon) Amazon Health AI is an in-app vertical AI agent from Amazon. Amazon's health assistant, which explains your records, books One Medical appointments and handles prescription renewals. Nearfi's Agent Index rates its autonomy 3/5 (multi-step, with approval): it books appointments and handles prescription renewals for you, but hands clinical judgement to human providers, so it does not complete care decisions on its own. Status: live; 0 incident(s) logged. Source: https://nearfi.co/agents/amazon-health-ai - Category: Health · vertical · In-app - What it does: Health AI answers health questions and, if you connect your One Medical records, explains lab results and medications. It books appointments with One Medical providers by message, video or in person, and handles prescription renewals, which can be filled through Amazon Pharmacy. It began in the One Medical app in January 2026 and reached Amazon.com and the Amazon app in March. - Limits: Amazon says it is 'not intended for diagnosis or treatment without the support of a care provider' and hands clinical judgement to human providers. Every route it offers leads to Amazon's own clinics and pharmacy. # Doctronic prescription renewals (Doctronic) Doctronic prescription renewals is a standalone vertical AI agent from Doctronic. An AI that renews routine prescriptions for people in Utah, under a state experiment with doctors checking its early work. Nearfi's Agent Index rates its autonomy 3/5 (multi-step, with approval): it interviews patients, checks history and sends renewals, but in its current phase doctors review every renewal, so it does not yet approve on its own. Status: beta; 0 incident(s) logged. Source: https://nearfi.co/agents/doctronic-renewals - Category: Health · vertical · Standalone - What it does: A Utah patient verifies their identity and answers a short AI interview about a maintenance medicine. Doctronic checks their prescription history, interactions and allergies, then sends the renewal to a pharmacy or refers the case to a doctor. About 190 drugs are eligible; controlled substances, injectables and short-course antibiotics are not. Under the state agreement, doctors review the first 250 renewals in each drug class before the AI may approve renewals on its own, with sampling after that. - Limits: In June 2026 the pilot was still in its first phase, with doctors reviewing every renewal. Utah's medical licensing board asked the state to pause it, and a Stanford health law analysis noted little published evidence of performance before launch and terms of service that disclaim liability. # Headspace Ebb (Headspace) Headspace Ebb is an in-app vertical AI agent from Headspace. A companion in the Headspace app that helps you reflect on how you feel and suggests matching meditations. Nearfi's Agent Index rates its autonomy 1/5 (suggests): it reflects with you and recommends meditations and content from the Headspace library, and you choose what to do; it takes no other actions. Status: live; 0 incident(s) logged. Source: https://nearfi.co/agents/headspace-ebb - Category: Mental health & wellbeing · vertical · In-app - What it does: Ebb is a conversational companion trained in motivational interviewing. It helps members reflect on their thoughts and emotions and then recommends meditation, mindfulness and wellbeing content. Headspace added voice in December 2025 and says it plans longer-term memory. - Limits: Headspace calls Ebb a 'sub-clinical support tool': it is not for clinical use and gives no mental health advice or diagnoses, and the company says AI is not a replacement for human-based care. Ebb watches for language about suicidal ideation, self-harm, abuse, eating disorders and substance abuse and then shows emergency and local resources. It is not a substitute for crisis support. We found no regulator clearance for Ebb as a medical device. # Wysa (Wysa) Wysa is a standalone vertical AI agent from Wysa. An AI coaching app that walks you through CBT-style self-help exercises, with human coaches available. Nearfi's Agent Index rates its autonomy 1/5 (suggests): it guides you through CBT-style exercises in conversation and you carry them out; it does not act on your accounts or the wider world. Status: live; 0 incident(s) logged. Source: https://nearfi.co/agents/wysa - Category: Mental health & wellbeing · vertical · Standalone - What it does: Wysa is a smartphone conversational agent that delivers cognitive behavioural therapy techniques for low mood and anxiety, alongside a self-help library and optional human coaches. Its clients have included health providers and government agencies, among them the NHS and Singapore's Ministry of Health. - Limits: The FDA gave Wysa Breakthrough Device Designation in May 2022 for a version aimed at adults with chronic musculoskeletal pain and depression or anxiety. That is a designation to speed development, not an approval, and we found no marketing authorisation. Wysa is not a substitute for crisis support or for a clinician. # IgniteAI Agent (Canvas) (Instructure) IgniteAI Agent (Canvas) is an in-app vertical AI agent from Instructure. A chat assistant inside Canvas that builds course modules and adjusts due dates for teachers. Nearfi's Agent Index rates its autonomy 3/5 (multi-step, with approval): it plans and carries out multi-step Canvas tasks such as building modules, with the instructor previewing the steps and checking the output; it is not 4 because that approval is part of the design. Status: live; 0 incident(s) logged. Source: https://nearfi.co/agents/instructure-igniteai-agent - Category: Learning · vertical · In-app - What it does: Announced in March 2026 as the agent layer of Instructure's IgniteAI, it takes a request from anywhere in Canvas and works out how to do it using more than 500 Canvas APIs: building modules and pages, generating rubrics, changing due dates. Instructors can preview its steps and verify the output. - Limits: It serves educators and administrators, not students. It is in an early-adopter phase, and Phil Hill, an education-technology analyst, wrote it was too early to evaluate its impact, given cautious campus adoption.