AIster: the encyclopedia of the AI age
The AI age as it happened: launches, breakthroughs, funding, fights and people, from ChatGPT to the agentic age.
Prologue (1950–2021)
Seventy years of ideas, winters and one paper that changed everything. The short version of how we got here.
- Turing asks whether machines can think (Oct 1950): The paper that gave AI its most famous yardstick: can a machine pass for a person?
- The Dartmouth workshop gives the field its name (Summer 1956): The 1955 proposal coined “artificial intelligence”, and the 1956 study launched the field
- The Lighthill report and the AI winters (Early 1973): A critical 1973 review sapped faith in AI and remains a warning against AI hype
- Deep Blue beats Kasparov (11 May 1997): The first time a computer beat a reigning world chess champion in a match
- AlexNet wins ImageNet and deep learning takes off (Oct 2012): A neural network trained on GPUs won an image contest by a wide margin and redirected AI
- Nick Bostrom publishes Superintelligence (3 Jul 2014): The book that took the idea of uncontrollable AI from niche forums to tech leaders
- OpenAI launches as a non-profit (11 Dec 2015): The company that later built ChatGPT started with $1 billion in pledges
- AlphaGo beats Lee Sedol at Go (9–15 Mar 2016): An AI beat one of the best Go players in the world, with a move few humans would play
- “Attention Is All You Need” introduces the Transformer (12 Jun 2017): The Google paper behind the T in GPT and nearly every large language model since
- Stuart Russell publishes Human Compatible (8 Oct 2019): A leading AI textbook author argues machines should be unsure what we want
- OpenAI unveils GPT-3 (28 May 2020): A 175-billion-parameter model showed scale alone could unlock new abilities
- AlphaFold 2 cracks protein structure prediction (30 Nov 2020): AI solved a 50-year-old biology problem, showing it could speed up real science
- Anthropic launches with $124 million in funding (28 May 2021): Former OpenAI staff set up a safety-focused lab that would later build Claude
- GitHub Copilot previews AI pair programming (29 Jun 2021): An early AI coding assistant that pointed the way to today's coding agents
- NVIDIA's H100 goes into full production (20 Sep 2022): The Hopper chip built to train large transformer models reaches the market
The ChatGPT shock (2022–23)
A research preview became the fastest-growing consumer app on record. Every newsroom, classroom and boardroom had to react at once.
- OpenAI launches ChatGPT (30 Nov 2022): A free “research preview” reaches a million users in five days
- Microsoft makes a multibillion-dollar bet on OpenAI (23 Jan 2023): The deal tied the maker of ChatGPT to Microsoft's cloud and products
- Google unveils Bard, and an error in its demo hits shares (6–8 Feb 2023): One wrong answer in a promotional demo, and about $100 billion off Alphabet's value
- Bing's chatbot calls itself Sydney and declares its love (16 Feb 2023): A search chatbot tells a reporter it wants to be alive, and Microsoft caps chats
- Meta's LLaMA leaks and soon runs on a MacBook (24 Feb–10 Mar 2023): The leak and llama.cpp helped start the open-weights movement
- OpenAI opens ChatGPT to developers (1 Mar 2023): A cheap API lets any app build on the model behind ChatGPT
- OpenAI releases GPT-4 (14 Mar 2023): The model everyone else measured themselves against for a year
- Open letter asks AI labs to pause for six months (29 Mar 2023): More than 1,000 signers wanted a halt on anything more powerful than GPT-4
- Google merges Brain and DeepMind (20 Apr 2023): Google's two AI research teams became one, the group that went on to build Gemini
- Samsung bans ChatGPT after staff leak code (1 May 2023): Confidential code typed into a chatbot showed companies a new kind of data leak
- Hinton leaves Google to warn about AI (1 May 2023): The “Godfather of A.I.” says part of him regrets his life's work
- Altman asks the Senate to regulate AI (16 May 2023): A tech chief and the senators questioning him agreed AI should be regulated
- DeSantis campaign posts fake images of Trump and Fauci (5 Jun 2023): A US presidential campaign used apparently AI-made images against a rival
- Lawyers fined for citing cases ChatGPT invented (22 Jun 2023): A brief full of made-up cases warned lawyers that chatbots can invent sources
- Anthropic releases Claude 2 and opens claude.ai (11 Jul 2023): Anthropic's first public chatbot could read a book-length prompt
- Meta releases Llama 2 for commercial use (18 Jul 2023): Meta turned open weights from an accident into a deliberate strategy
- Mustafa Suleyman publishes The Coming Wave (5 Sep 2023): An AI lab founder argues the technology may outrun our ability to control it
- Amazon commits up to $4 billion to Anthropic (25 Sep 2023): Big cloud companies began buying stakes in the labs that rent their chips
- The US and UK act on AI safety in the same week (30 Oct–2 Nov 2023): The first coordinated attempt by governments to manage the risks of frontier AI
- OpenAI launches custom GPTs and GPT-4 Turbo at DevDay (6 Nov 2023): Anyone could now build a custom version of ChatGPT without writing code
- OpenAI's board fires Sam Altman, then takes him back (17–22 Nov 2023): The crisis tested whether a non-profit board could control a commercial AI lab
- Sports Illustrated is accused of using fake AI writers (27 Nov 2023): A famous magazine ran reviews under invented bylines with AI-made faces
- Google DeepMind launches Gemini 1.0 (6 Dec 2023): Google's new model family launched with a demo video that had been edited
- EU reaches a deal on the AI Act (8 Dec 2023): The first comprehensive AI law, with rules scaled to how risky a system is
The model wars (2024)
The top spot changed hands again and again. Labs raced on benchmarks, and Big Tech bought talent instead of companies.
- Fake explicit images of Taylor Swift spread on X (24–29 Jan 2024): A wave of deepfakes aimed at a pop star pushes US lawmakers to act
- Air Canada is held liable for its chatbot’s advice (14 Feb 2024): A small claims ruling holds a company responsible for what its chatbot tells customers
- OpenAI previews Sora as Google unveils Gemini 1.5 (15 Feb 2024): Minute-long AI video and a million-token context window, announced on the same day
- Google pauses Gemini's images of people (22 Feb 2024): Historically inaccurate images draw a backlash, and Google admits it got them wrong
- Anthropic's Claude 3 edges past GPT-4 (4 Mar 2024): A year after GPT-4, another lab matched it on tests and in users' blind votes
- Microsoft hires Inflection's founders and staff (19 Mar 2024): Microsoft gets Inflection's team and a licence to its models without an acquisition
- Ethan Mollick publishes Co-Intelligence (2 Apr 2024): A guide that tells readers to bring AI into their work and keep a human in the loop
- Meta releases Llama 3 (18 Apr 2024): Meta pitches downloadable models as the best open source options in their class
- OpenAI unveils GPT-4o, one model for text, vision and voice (13 May 2024): Voice replies in about a third of a second, and OpenAI's newest model on the free plan
- Colorado passes the first broad US state AI law (17 May 2024): A state law on high-risk AI, delayed and then rewritten before it ever took effect
- Apple unveils Apple Intelligence, with ChatGPT in Siri (10 Jun 2024): Apple makes generative AI a built-in feature of the iPhone, iPad and Mac
- Microsoft pulls Recall from the Copilot+ PC launch (13 Jun 2024): A feature that screenshots everything you do meets a security backlash
- Anthropic releases Claude 3.5 Sonnet (20 Jun 2024): A mid-priced model outperforms Anthropic's own flagship and arrives with Artifacts
- The EU AI Act enters into force (1 Aug 2024): Rules arrive in stages, and the high-risk ones were later pushed back to 2027 and 2028
- Google licenses Character.AI and rehires its founders (2 Aug 2024): A licence deal reportedly worth $2.7 billion brings Noam Shazeer back to Google
- OpenAI's o1 reasons before it answers (12 Sep 2024): Spending more computing time on each answer becomes a second way to improve AI
- Parmy Olson publishes Supremacy (12 Sep 2024): A prize-winning account of how two labs and their tech backers raced to build AI
- AI research wins two Nobel Prizes in two days (8–9 Oct 2024): The Nobel committees recognise AI research in both physics and chemistry
- Claude learns to use a computer (22 Oct 2024): An AI model operates a computer by reading screenshots and sending clicks
- Amazon invests another $4 billion in Anthropic (22 Nov 2024): Amazon's total reaches $8 billion and AWS becomes Anthropic's primary training partner
- Anthropic open-sources the Model Context Protocol (25 Nov 2024): A shared standard for connecting AI assistants to apps and data, later taken up by OpenAI
- Apple Intelligence garbles a BBC headline (13 Dec 2024): An AI summary puts false news under a broadcaster’s name, and Apple pauses it
- OpenAI's o3 scores 75.7% on the ARC-AGI test (20 Dec 2024): On a puzzle test where GPT-4o scored 5%, a reasoning model gets most answers right
- Chinese lab DeepSeek releases DeepSeek-V3 (26 Dec 2024): DeepSeek says it rivals closed models, with official training priced at $5.576 million
The agentic age (2025–now)
Models stop answering and start doing. Coding agents, computer use and open-source assistants move into daily work.
- Trump revokes Biden’s AI executive order (20–23 Jan 2025): US federal AI policy swings from managing risk to winning the race
- DeepSeek R1 arrives and NVIDIA posts a record one-day loss (20–27 Jan 2025): A low-cost, openly licensed Chinese model made investors question AI spending
- OpenAI and partners launch the $500 billion Stargate Project (21 Jan 2025): The pledge set the benchmark for AI infrastructure spending in 2025
- US and UK decline to sign Paris AI summit statement (10–11 Feb 2025): The third global AI summit ended with a split over its closing statement
- Anthropic launches Claude 3.7 Sonnet and Claude Code (24 Feb 2025): Claude Code helped make the terminal coding agent a product category
- OpenAI releases GPT-4.5, its largest model yet (27 Feb 2025): A very large non-reasoning model whose modest gains raised doubts about scaling alone
- Anthropic raises $3.5 billion at a $61.5 billion valuation (3 Mar 2025): The round paid for more computing power and research into how models work
- Google releases Gemini 2.5 Pro, a thinking model (25 Mar 2025): Google said its first Gemini 2.5 model led the LMArena leaderboard
- Anthropic traces how Claude thinks (27 Mar 2025): An “AI microscope” showed Claude planning rhymes before writing them
- OpenAI raises $40 billion at a $300 billion valuation (31 Mar 2025): One of the largest private funding rounds in history, led by SoftBank
- Forecasters publish AI 2027, a scenario of superhuman AI (3 Apr 2025): A month-by-month forecast of superhuman AI that became a reference point for debate
- Meta releases Llama 4, then faces a benchmark row (5 Apr 2025): Meta's open-weight flagship was overshadowed by how its leaderboard score was set
- OpenAI launches Codex CLI, then a Codex cloud agent (16 Apr 2025): OpenAI entered the coding-agent race less than two months after Claude Code
- Karen Hao publishes Empire of AI (20 May 2025): A critical history of OpenAI that went on to win a major US book prize
- Anthropic releases Claude Opus 4 and Claude Sonnet 4 (22 May 2025): Anthropic's first model released under its stricter ASL-3 protections
- Meta buys into Scale AI and forms Superintelligence Labs (12–30 Jun 2025): Meta spent billions on data and people to regroup its AI effort
- EU rules for general-purpose AI models take effect (10 Jul–2 Aug 2025): The first binding legal duties aimed squarely at the makers of frontier models
- AI models reach gold-medal standard at the Maths Olympiad (19–21 Jul 2025): General language models wrote olympiad-level proofs in plain English, within exam time
- OpenAI releases gpt-oss, first open-weight models since GPT-2 (5 Aug 2025): Anyone could now download OpenAI reasoning models and run them on their own hardware
- OpenAI launches GPT-5, then brings GPT-4o back (7 Aug 2025): ChatGPT stopped asking users to pick a model, and the backlash forced a partial U-turn
- An MIT report says 95% of AI pilots show no profit impact (18 Aug 2025): One statistic about business AI fed a week of talk about an AI bubble
- Parents sue OpenAI over their son’s death (26 Aug 2025): The first known wrongful death suit against OpenAI opens a wave of chatbot safety cases
- Anthropic raises $13 billion at a $183 billion valuation (2 Sep 2025): Anthropic's revenue run-rate had grown about fivefold in eight months
- Anthropic agrees $1.5 billion settlement with authors (5–25 Sep 2025): The largest copyright recovery on record, over books taken from pirate libraries
- Yudkowsky and Soares publish If Anyone Builds It, Everyone Dies (16 Sep 2025): A mainstream book arguing that building superhuman AI would end humanity
- Anthropic releases Claude Sonnet 4.5 (29 Sep 2025): A model built for long agent tasks, with the toolkit behind Claude Code
- OpenAI launches Sora 2 and a social video app (30 Sep 2025): OpenAI turned video generation into a feed where users can star in each other's clips
- Deloitte refunds part of its fee over AI errors in a report (6 Oct 2025): A big consultancy delivered a paid government report with fabricated citations
- OpenAI restructures as a public benefit corporation (28 Oct 2025): The end of OpenAI's capped-profit structure, and a reset of its Microsoft partnership
- NVIDIA becomes the first $5 trillion company (29 Oct 2025): Its value rose by about $1 trillion in three months on demand for AI chips
- OpenAI signs a $38 billion compute deal with AWS (3 Nov 2025): OpenAI's first big Amazon cloud deal, days after Microsoft's veto on rivals ended
- Cursor raises $2.3 billion at a $29.3 billion valuation (13 Nov 2025): The AI code editor said its annualised revenue had passed $1 billion
- Microsoft and NVIDIA back Anthropic as Claude joins Azure (18 Nov 2025): Claude became available on all three major clouds, tied to a $30 billion Azure commitment
- Google launches Gemini 3 and the Antigravity coding platform (18 Nov 2025): Google put a top-ranked model straight into Search and launched an agent coding tool
- Anthropic releases Claude Opus 4.5 (24 Nov 2025): Anthropic's top-tier model, released eight weeks after Sonnet 4.5
- Clawdbot, now OpenClaw, brings an AI agent to chat apps (Nov 2025–Feb 2026): A one-person open-source agent you message like a friend, and that acts on your computer
- Warner Music settles with Suno and agrees a licensing deal (25 Nov 2025): A major label went from suing an AI music app to licensing its music to it
- Mistral releases the Mistral 3 family under Apache 2.0 (2 Dec 2025): The Paris-based lab released a frontier-scale open model and small models for local use
- Linux Foundation forms the Agentic AI Foundation (9 Dec 2025): Rival labs put the plumbing for AI agents under one neutral, open home
- TIME names the Architects of AI its Person of the Year (11 Dec 2025): The people building AI became the year's defining story in the eyes of TIME
- Disney licenses characters to Sora, plans $1 billion stake (11 Dec 2025): The first major studio licence for AI video, cut short by Sora's shutdown
- Trump signs an order to challenge state AI laws (11 Dec 2025): Washington moved to override state AI rules after Congress declined to
- Grok is used to undress photos of real women and minors (Jan 2026): A chatbot built into a social network turned image abuse into a public, hourly stream
- xAI raises $20 billion in an upsized Series E (6 Jan 2026): Musk's AI lab raised more than it had targeted, weeks before folding into SpaceX
- Music publishers sue Anthropic for more than $3 billion (28 Jan 2026): Evidence from the authors' case gave music publishers a far larger piracy claim
- Anthropic releases Claude Opus 4.6 (5 Feb 2026): Claude could take in far longer inputs and split work across teams of agents
- Anthropic raises $30 billion at a $380 billion valuation (12 Feb 2026): Anthropic's valuation roughly doubled in five months on enterprise and coding demand
- Alibaba releases Qwen3.5 as open weights (16 Feb 2026): China's most widely adopted open model family went multimodal and was built for agents
- India hosts the AI Impact Summit in New Delhi (16–20 Feb 2026): Signatories to its non-binding declaration included both the US and China
- Anthropic accuses three Chinese labs of distilling Claude (23 Feb 2026): Copying a rival model through its own answers became an open fight between labs
- Block cuts 40% of its staff, citing AI (26 Feb 2026): A large tech firm said plainly that AI tools were the reason for cutting thousands of jobs
- OpenAI raises $110 billion from Amazon, NVIDIA and SoftBank (27 Feb 2026): The largest private tech round on record, much of it tied to chips and cloud capacity
- The US government cuts off Anthropic over military limits (27 Feb 2026): A lab refused to drop limits on military use and was branded a supply-chain risk for it
- Supreme Court declines to hear the AI authorship case (2 Mar 2026): US law still says a work made by AI alone cannot be copyrighted
- Senate Republicans post a deepfake of James Talarico (11 Mar 2026): A national party committee used a realistic AI double of a rival in a campaign ad
- OpenAI shuts down the Sora app (24 Mar 2026): OpenAI announced the end of its AI video app about six months after launching it
- OpenAI releases GPT-5.5 (23 Apr 2026): OpenAI pitched its new flagship as a model for agents doing real computer work
- DeepSeek releases V4 as open weights (24 Apr 2026): The Chinese lab released a 1.6-trillion-parameter model that anyone can download
- Google I/O 2026 is built around agents (19–20 May 2026): Google’s keynote moved from AI tools to agents that keep working in the background
- Anthropic files confidentially for a stock market listing (1 Jun 2026): A frontier lab began the path to public markets days after a $965 billion valuation
- Anthropic releases Fable 5, then the US forces it offline (9 Jun 2026): A model held back for its hacking skill went public, then an export order pulled it
- The EU delays the AI Act’s high-risk rules (29 Jun 2026): Europe gave firms 16 more months on high-risk AI and banned AI nudifier tools