A source-linked timeline of the people, ideas and breakthroughs shaping a new technological era. Computing, AI, software, space, biology, energy and the infrastructure connecting them.
426 milestones, 163 organization profiles and 729 contributor paths. This curated archive is incomplete. Dates and financial records retain their source qualifications.
Main Timeline contains every archived entry. Simple Timeline follows the major shifts. Subject timelines offer focused histories.
In On Computable Numbers, Turing describes an abstract computing machine and a universal machine that can simulate others.
A mathematical foundation for programmable computation, preceding his later work on machine intelligence. This card records the 1936 work, not the construction of a physical computer.
Bill Hewlett and David Packard form Hewlett-Packard, initially making electronic test equipment.
The garage company becomes a defining Silicon Valley origin story, long before personal computers or modern AI.
A logical model of neurons connects nervous activity with computation.
An early foundation for artificial neural networks.
Penn publicly introduces ENIAC, a general-purpose electronic computer designed by Mauchly and Eckert and programmed by a pioneering team of six women.
Electronic computation becomes a practical engineering system. ENIAC initially required manual configuration for each new calculation; it was not originally a modern stored-program computer.
Bell Labs develops the point-contact transistor in 1947, establishing a new way to amplify and switch electronic signals.
Semiconductor devices make smaller, more reliable electronics possible. Integrated circuits and microprocessors build on this foundation.
Wiener describes feedback and control in animals and machines.
Intelligence becomes something researchers can study through systems and signals.
The Organization of Behavior proposes a learning theory based on changes in connections between neurons.
Hebbian learning becomes a lasting idea in neuroscience and neural computation.
Computing Machinery and Intelligence proposes the imitation game.
A lasting way to frame the question of machine intelligence.
Hopper develops A-0 for UNIVAC I, automating the assembly of routines needed by a program.
An important step toward higher-level programming and compilers. The historical term compiler differs from today’s usage; A-0 also resembles a linker or macro-assembler.
Work begins on a program that searches for mathematical proofs.
Symbolic reasoning becomes a working software project.
A research proposal gives the emerging field its enduring name.
The proposal precedes the better-known summer workshop.
Researchers gather for the Dartmouth Summer Research Project on Artificial Intelligence.
A formative meeting for AI as a research discipline.
Eight engineers leave Shockley’s laboratory to establish Fairchild Semiconductor with backing from Fairchild Camera and Instrument.
Its engineers and spinouts help turn the region into a semiconductor and startup center.
Rosenblatt publishes a probabilistic model for learning and information storage.
Learning from examples becomes a concrete research direction.
A language built around symbolic lists becomes a central tool for AI research.
Researchers gain a flexible way to represent programs and knowledge.
Samuel describes experiments in machine learning using a checkers-playing program.
Learning improves a game-playing system through experience.
Fairchild fabricates a working planar integrated circuit, combining components on a silicon chip. Jack Kilby had demonstrated an earlier integrated circuit at Texas Instruments in 1958.
Planar manufacturing helps turn integrated circuits into scalable products. Increasing chip density connects this work to microprocessors and modern computing.
A research system applies chemical knowledge to infer molecular structures.
A landmark in expert systems and domain-specific AI.
A language-processing program uses scripted rules to imitate conversation.
The gap between apparent understanding and actual capability becomes visible.
SRI begins a mobile-robot project combining perception, planning, and movement.
An early bridge from abstract AI to action in the physical world.
Robert Noyce and Gordon Moore incorporate the semiconductor venture that becomes Intel.
Semiconductor manufacturing and microprocessors provide foundations for personal computers, servers and later AI infrastructure.
Perceptrons analyzes the capabilities and limitations of a class of neural-network models.
The book is part of the debate about which approaches to intelligence can scale.
Apollo 11’s lunar module lands on the Moon. Armstrong and Aldrin explore the surface while Collins remains in lunar orbit.
Human spaceflight brings together propulsion, computing, navigation and large engineering teams. Reusable launch vehicles and satellite networks extend the spaceflight story in later decades.
The first computer-to-computer message on ARPANET travels between UCLA and the Stanford Research Institute.
Packet-switched research networks become a foundation for the internet. Shared network protocols and the Web later turn connected computers into everyday infrastructure.
A system follows language instructions inside a simulated world of blocks.
Language, reasoning, and action meet in a deliberately limited environment.
Intel announces the 4004 chipset as a standalone product after developing it for Busicom calculators.
Putting programmable CPU functions onto a chip helps make computers smaller and more affordable. Personal computers and later accelerators continue this hardware story.
Sequoia Capital begins operating in 1972.
The archive includes the investment organizations behind AI companies; this founding predates many of their AI investments.
A critical review questions the progress and ambitions of contemporary AI research.
A key document in the history of reduced enthusiasm and research support.
Bill Gates and Paul Allen begin Microsoft in the emerging personal-computer software industry.
A major software company grows alongside the PC platform, long before its cloud and AI products.
A venture firm investing in technology and healthcare, including AI infrastructure and applications.
Venture capital is part of the ecosystem that helps new technologies scale. The firm’s origins precede the later AI investments linked in its profile.
Apple begins shipping the Apple II, a complete personal computer with a keyboard, color graphics and expansion slots.
Personal computing reaches more homes, classrooms and small businesses through an expandable, ready-to-use machine.
Larry Ellison, Bob Miner and Ed Oates start the company originally named Software Development Laboratories, later Oracle.
Enterprise databases become core business infrastructure, preceding the later cloud and AI data-platform eras.
IBM unveils the Personal Computer 5150, built around an Intel processor and an ecosystem of outside components and software.
The IBM-compatible PC becomes a major platform for the hardware and software industry. It is not the first personal computer.
A recurrent network model connects collective computation with physical systems.
Patterns can be stored and recovered through a network’s dynamics.
Accel begins operating in 1983.
The archive includes the investment organizations behind AI companies; this founding predates many of their AI investments.
ARPANET moves from its earlier host protocol to TCP/IP in a coordinated transition.
Common protocols allow different networks to communicate. The Web, online platforms and cloud services can grow on top of that shared network foundation.
Michael Dell starts a computer business from his University of Texas dorm room, building a direct-to-customer model.
The PC industry expands through new distribution and manufacturing approaches, alongside the software ecosystem.
The original Macintosh introduces a mouse-driven graphical desktop in a compact all-in-one computer, with MacWrite and MacPaint.
Graphical interfaces make personal computing more approachable. The Mac popularizes this approach rather than inventing the GUI or mouse.
Philips and ASM International establish ASM Lithography, later known as ASML.
The archive includes the chip-manufacturing equipment behind modern AI compute.
A learning algorithm trains stochastic networks of connected units.
Probability and representation learning come together in neural networks.
Quantum Computer Services forms in 1985, the online-services business that later becomes America Online.
Dial-up services, email and chat help bring networked computing into ordinary homes before the social web.
A widely influential paper explains learning internal representations by propagating errors backward.
It helps revive multilayer neural-network research; the method has earlier roots.
Morris Chang establishes TSMC, pioneering a business dedicated to manufacturing other companies’ chip designs.
An important industrial foundation for the later AI hardware ecosystem.
A paper develops temporal-difference methods for prediction through experience.
A foundation for learning from delayed outcomes.
Guido van Rossum publishes Python to USENET after beginning development at CWI in 1989. His own account dates the first public release to February 1991.
A general-purpose language becomes a shared foundation for scientific computing and later AI frameworks, including TensorFlow and PyTorch.
Tim Berners-Lee posts an invitation to use the WorldWideWeb project on the alt.hypertext newsgroup, describing browsers, servers and linked documents.
The Web connects information across computers through links and shared protocols. This public announcement follows the first working browser and server at CERN in 1990.
Linus Torvalds announces his hobby operating-system project on comp.os.minix. This is the first announcement, before the initial source release.
Linux later becomes core infrastructure for servers and AI compute. The early source license restricted commercial use; the project adopted the GPL in 1992.
A paper establishes convergence conditions for a model-free learning algorithm.
An agent can learn action values from experience.
NVIDIA begins with a focus on graphics for gaming and multimedia.
Its later GPU and CUDA ecosystem becomes important infrastructure for scientific computing and modern AI. The company predates the current AI boom.
Stanford students Jerry Yang and David Filo create the web directory that becomes Yahoo.
Directories, search and portals give a rapidly expanding Web a consumer-facing entry point.
Jim Clark and Marc Andreessen start Mosaic Communications, later renamed Netscape Communications.
Commercial browsers help make the Web a mass-market software platform, with JavaScript following inside Netscape.
Support-vector networks offer a powerful method for classification.
Statistical learning advances alongside neural-network research.
Amazon opens its online bookstore in July 1995, after Jeff Bezos began the company in 1994.
Online retail develops into a large-scale software and logistics business; AWS later makes computing infrastructure available to other developers.
Microsoft launches Windows 95 with the Start menu and taskbar, bringing a new desktop experience to the PC market.
A defining mass-market software platform establishes interaction patterns that persist across later generations of Windows.
Pierre Omidyar launches AuctionWeb, the person-to-person online marketplace later named eBay.
The Web becomes a place where individuals transact with one another, with reputation and trust built into a digital marketplace.
Netscape and Sun announce the JavaScript name. Brendan Eich created the language at Netscape in May 1995; it had already appeared publicly as LiveScript in the September Navigator 2.0 beta.
Programmable web pages grow into a platform for interactive applications, including later AI interfaces. This is the naming announcement, not the date the language was first created.
LSTM introduces a recurrent architecture designed to learn long-range dependencies.
A building block for speech, handwriting, and sequence models.
IBM’s chess computer wins a six-game match against the reigning world champion.
A public milestone in specialized machine performance.
Gradient-based learning and convolutional networks are demonstrated for document recognition.
Neural networks solve a practical recognition task.
Larry Page and Sergey Brin establish Google following their Stanford research.
Google later builds research teams, tools, and infrastructure central to modern AI; its corporate history is distinct from DeepMind’s 2010 founding.
Apple unveils the colorful, translucent iMac, an all-in-one desktop designed around simple internet access and USB peripherals.
Industrial design and internet-first simplicity become central to Apple’s consumer-computing comeback. This is the May unveiling, before its August retail debut.
Salesforce incorporates and its founders begin building the first version of its CRM software.
Subscription software delivered over the internet challenges the traditional installed enterprise-software model.
A global venture firm investing in technology companies, including frontier models, enterprise software and physical AI.
Venture capital is part of the ecosystem that helps new technologies scale. The firm’s origins precede the later AI investments linked in its profile.
AOL and Time Warner enter a merger agreement, combining an internet-service company with a major media business.
The proposed combination becomes an emblem of dot-com-era convergence. This is the agreement date; the merger closes in 2001.
An ensemble of randomized decision trees becomes a practical prediction method.
AI history includes effective approaches beyond deep neural networks.
Apple introduces a portable digital music player that combines a hard drive, a scroll wheel and integration with iTunes on the Mac.
A dedicated consumer device connects personal computers to everyday digital media, setting up Apple’s broader mobile-device era.
Myspace introduces a social network centered on personal profiles, interests and shared media.
Customizable profiles and music communities make social networking part of popular culture.
Palantir is founded to build software for complex data analysis.
Its later AI products build on a long history of government and enterprise data infrastructure.
Jonathan Abrams opens Friendster’s beta, connecting people through profiles and mutual friends.
Visible social graphs become a recognizable consumer product before Facebook’s rise.
The international consortium announces completion of the Human Genome Project’s goals, producing a reference covering about 99% of gene-containing regions. Some gaps remain.
A shared reference and openly released sequence data give biology a common information foundation. Genome editing and computational biology develop alongside it.
LinkedIn launches a professional network after work began in Reid Hoffman’s living room in 2002.
Online identity and social connections become infrastructure for hiring, careers and business relationships.
A venture firm backing early technology companies across AI, computing, energy and other fields.
Venture capital is part of the ecosystem that helps new technologies scale. The firm’s origins precede the later AI investments linked in its profile.
TheFacebook launches for Harvard students before expanding to more campuses and eventually the public.
A campus directory becomes a major social-network platform; its later AI organization is another chapter.
Steve Huffman and Alexis Ohanian start Reddit in 2005, with backing from Y Combinator’s first cohort.
Community submissions and voting create a different form of online discovery from personal social profiles.
Y Combinator begins operating in 2005.
The archive includes the investment organizations behind AI companies; this founding predates many of their AI investments.
Founders Fund begins operating in 2005.
The archive includes the investment organizations behind AI companies; this founding predates many of their AI investments.
Linus Torvalds records the first Git commit while building a distributed version-control system for Linux development. This marks the project’s beginning, not a 1.0 release.
Distributed code history and branching support collaboration across open-source software, including the libraries and tools used throughout AI.
Jawed Karim uploads Me at the zoo, the first YouTube video, filmed at the San Diego Zoo.
User-uploaded video becomes a major branch of the social web and later the creator economy.
A fast learning algorithm trains deep belief networks layer by layer.
One important step in the renewed interest in deep learning.
Brian Halligan and Dharmesh Shah found HubSpot to help businesses attract customers through inbound marketing.
Business software moves toward integrated, internet-delivered marketing, sales and customer workflows.
AWS launches Amazon S3, offering developers scalable storage through a web-service interface.
Developers can rent infrastructure rather than build every layer themselves, enabling later cloud applications and AI services.
Jack Dorsey posts the first message on the service then called twttr, while the project is developing inside Odeo.
Short, public posts create a real-time network for conversation, news and public figures.
Facebook introduces News Feed, bringing updates from friends onto a shared home-page stream.
Social networking shifts from visiting individual profiles toward following a continuously updated feed. Later ranking systems develop beyond this initial release.
NVIDIA makes general-purpose GPU programming available through CUDA.
A software ecosystem helps researchers apply graphics processors to machine learning.
Apple unveils iPhone, combining a phone, an iPod and internet communication in one device with a multi-touch interface.
Touch-first mobile computing becomes a central platform for software and internet services. The January introduction precedes the June retail launch.
Apple unveils the first MacBook Air, a thin, lightweight notebook that trades some built-in ports and drives for portability.
The ultraportable notebook becomes a defining direction for mainstream laptop design.
GitHub officially opens after its beta, giving developers a shared place to collaborate on code with Git.
Shared repositories and contributions help open-source software spread. The platform later becomes a distribution point for AI projects and coding assistants.
Apple opens the App Store with 500 apps, giving iPhone users a central place to find and install third-party software.
Mobile devices become extensible software platforms, connecting independent developers with a global consumer audience.
On its fourth flight, Falcon 1 reaches Earth orbit from Omelek Island.
A small privately developed launcher establishes the orbital foundation for SpaceX’s later rockets and satellite networks.
The T-Mobile G1, built by HTC, goes on sale alongside Android Market, bringing the first commercial Android phone to users.
An alternative smartphone platform connects Google’s software, hardware manufacturers, carriers and third-party apps.
A large image database organizes visual examples using a semantic hierarchy.
Shared data helps researchers compare recognition systems.
Jan Koum and Brian Acton build WhatsApp, a mobile messaging company founded in 2009.
Messaging over the internet becomes an alternative to carrier SMS and a major everyday communication platform.
Google begins the autonomous-driving project that becomes Waymo.
The research origin precedes Waymo’s formation as a separate business.
Andreessen Horowitz begins operating in 2009.
The archive includes the investment organizations behind AI companies; this founding predates many of their AI investments.
A London research startup pursues general-purpose learning systems.
A new lab puts general intelligence at the center of its mission.
Pinterest launches a service for saving and discovering visual ideas through collections of pins.
Social discovery expands from people and status posts toward interests, collections and visual search.
Patrick and John Collison start Stripe to simplify accepting payments online.
Programmable financial infrastructure later becomes part of the AI-agent ecosystem.
Apple unveils a multi-touch tablet for web browsing, media, reading and apps, positioned between a smartphone and a laptop.
The modern consumer tablet becomes a distinct computing format, extending the touch interface and mobile app ecosystem to a larger screen.
Kevin Systrom and Mike Krieger launch Instagram, a mobile photo-sharing app.
The smartphone camera and a social feed become a new publishing and discovery interface.
The app that becomes Snapchat begins as Picaboo in 2011, centered on disappearing picture messages.
Camera-first messaging develops a different sharing model from permanent public profiles.
A deep convolutional network wins the ImageNet recognition competition by a large margin.
GPUs, data, and neural networks make a compelling combination.
Researchers summarize how deep neural networks improve acoustic modeling.
Speech recognition becomes an early practical success of deep learning.
A team led by Zhang Yiming and Liang Rubo founds ByteDance. Toutiao follows as an early mobile content product.
Personalized discovery and recommendation become a company-building foundation, later extending to short video.
Snowflake is founded to build a data warehouse for cloud infrastructure, separating storage and compute.
Cloud data platforms change how organizations store and analyze information, forming part of the infrastructure later AI applications use.
Facebook announces an agreement to acquire Instagram for approximately $1 billion in cash and shares, subject to closing conditions.
A major social-network company moves into smartphone photo sharing. The agreement date is distinct from the transaction’s later closing.
A research paper shows how RNA can guide Cas9 to cut a chosen DNA sequence in a laboratory system.
Programmable DNA cutting creates a foundation for genome editing. Later clinical therapies require additional research, delivery methods and regulatory review.
Efficient methods learn useful word representations from large text collections.
Relationships between words can be explored in a learned vector space.
Deep reinforcement learning combines a neural network with Q-learning on Atari games.
A shared learning approach works across multiple games.
Notion begins a long effort to build a flexible workspace, later becoming a platform for AI writing and agents.
Company origins precede Notion AI by a decade.
Researchers behind Apache Spark found Databricks to build a commercial data platform.
A key bridge from open-source data systems to enterprise AI infrastructure.
Solomon Hykes demonstrates Docker in a PyCon lightning talk, showing a simpler way to build and run applications in Linux containers.
Portable application packaging helps teams reproduce environments and deploy software, including later AI services. Docker Engine is open-source software; Docker Desktop has separate licensing terms.
Auto-Encoding Variational Bayes combines neural networks with approximate probabilistic inference.
A major branch in the history of deep generative models.
Michigan Tech students found a career network connecting students, universities, and employers.
Its university network later becomes infrastructure for expert AI training.
Facebook announces an agreement to acquire WhatsApp, with the messaging app’s brand maintained.
The social-network company expands into global mobile messaging. The announcement is distinct from the later closing; employee stock awards are separate from acquisition consideration.
A generator and discriminator train through an adversarial objective.
A new family of generative models emerges.
Neural networks map an input sequence to an output sequence.
A general framework for translation and other language tasks.
An attention mechanism aligns and translates language sequences.
A crucial step toward later attention-based architectures.
An adaptive stochastic optimization method combines estimates of first and second moments.
Training infrastructure matters as much as model architecture.
Hivemind autonomy software and unmanned aircraft.
Defense autonomy becomes part of the broader physical AI industry.
A Nature paper reports a deep Q-network learning multiple Atari games.
Deep reinforcement learning reaches a broader scientific audience.
Jason Citron and Stanislav Vishnevskiy launch Discord, initially focused on helping people talk while playing games.
Persistent community spaces connect voice, text and shared activities beyond a public news feed.
Kubernetes releases version 1.0, following its public introduction in 2014. Google also announces its donation to the newly formed Cloud Native Computing Foundation.
An open platform for managing containerized applications becomes infrastructure for distributed services and later AI workloads.
Google releases its machine-learning system for wider use.
Research infrastructure becomes accessible to more developers.
Residual connections make very deep neural networks easier to optimize.
A powerful architectural idea spreads beyond computer vision.
The first arXiv version describes end-to-end neural speech recognition in English and Mandarin.
Part of Amodei’s research history before OpenAI and Anthropic. Contributors listed follow the initial arXiv author list.
OpenAI announces a nonprofit research organization focused on beneficial AI.
The public announcement establishes its original mission and team.
After launching the ORBCOMM-2 satellites, Falcon 9 returns its first stage to a ground landing pad at Cape Canaveral. The date follows Florida local time.
Recovering an orbital-class booster is a step toward reuse. The upper stage remains expendable, and booster reflight comes later.
Scale starts as an API for human tasks and joins Y Combinator’s Summer 2016 batch.
Human data operations become important infrastructure for training and evaluating AI.
AlphaGo wins its five-game Go match against Lee Sedol, four games to one.
A landmark combines neural networks, search, and reinforcement learning.
A co-authored research paper organizes practical accident risks around objectives, supervision, exploration, and changing environments.
Helps turn AI safety concerns into specific machine-learning research problems.
ByteDance launches Douyin, a short-video product for China. TikTok’s international launch follows separately.
Mobile video and personalized discovery converge into a highly influential consumer format.
Google’s self-driving project becomes Waymo, a standalone business within Alphabet.
Separates project origins from the corporate transition.
An open-source framework emphasizes flexible, Python-oriented neural-network development.
A central tool for researchers and production model builders.
ByteDance launches TikTok for markets outside mainland China, following Douyin’s 2016 China launch.
A mobile short-video feed makes recommendation-driven discovery a central social interface.
Wayve begins developing learning-based autonomous driving in the UK, building on Cambridge machine-learning research.
An end-to-end learning approach connects computer vision and driving experience to vehicle behavior, alongside other paths to autonomy.
Lattice autonomy software and defense hardware.
Defense autonomy becomes part of the broader physical AI industry.
CoreWeave is founded in New Jersey, later developing a specialized AI cloud platform.
The archive follows the infrastructure companies that supply model training and inference.
The SES-10 mission launches on a previously flown Falcon 9 first stage, the first orbital-class booster reflight.
Reflight turns booster recovery into demonstrated reuse, connecting reusable launch technology to commercial satellite operations.
Agents learn a reward signal from human comparisons of behavior.
Preference learning becomes part of the path toward helpful language models.
The Transformer replaces recurrence with attention in a sequence-modeling architecture.
An architecture that will underpin many later language models.
A family of policy-gradient methods aims to simplify reinforcement learning.
An influential training method for agents and later preference-trained models.
Apple unveils iPhone X with an OLED edge-to-edge display, Face ID and gesture navigation in place of the Home button.
A major redesign changes the iPhone’s interaction model and makes facial authentication part of its core user experience.
A general reinforcement-learning approach learns chess, shogi, and Go from self-play.
A shared algorithm replaces some game-specific engineering.
Turing begins connecting engineering talent with companies, later expanding into human data and model post-training.
Distinct from Alan Turing’s foundational 1950 research.
Daniel and William Roberts establish the company that later becomes IREN.
Its energy-intensive infrastructure business later expands into large-scale AI compute.
Researchers propose using debate between AI systems to help human judges evaluate difficult questions.
A proposed route toward scalable oversight, supported by an initial experiment rather than a complete safety solution.
A language model is pre-trained on text and adapted to downstream tasks.
A foundation for the GPT model family.
TikTok and musical.ly announce their combination into a single worldwide app under the TikTok name.
A product merger consolidates short-video communities and accelerates the platform’s global footprint. The newsroom publication is dated August 2; its announcement text carries an August 1 dateline.
Bidirectional Transformer pre-training improves a range of language-understanding tasks.
Pre-trained language models become a standard starting point.
Microsoft and GitHub announce that the $7.5B acquisition has completed. Nat Friedman becomes GitHub CEO.
A major developer platform joins Microsoft ahead of the rise of AI coding tools. Microsoft’s financial filing dates the acquisition to October 25; this card uses the October 26 public announcement.
A style-based generator makes aspects of image synthesis more controllable.
Synthetic faces become a vivid demonstration of generative-model progress.
Three founders start a company focused on language AI, later concentrating on enterprise models and applications.
A branch of the Transformer and Google Brain research community becomes an independent company.
Ramp starts building financial operations software for businesses.
An enterprise software company that later brings agents into finance workflows.
Three founders start Linear to improve how teams plan and build software.
The product-development workflow later expands to collaboration between human teams and AI agents.
OpenAI introduces GPT-2 and initially limits release of its largest model.
Model capability and release policy become a shared public debate.
Mercury launches its financial platform for startups and businesses.
The company’s September financing post dates public launch to April 2019; its founding year is 2017. Mercury’s fintech platform is distinct from a licensed bank.
SpaceX launches a 60-satellite Starlink batch from Cape Canaveral, an early mass deployment for its low-Earth-orbit broadband network.
Connects launch capacity, satellite manufacturing and internet access. The date follows the Florida launch-site date, May 23; liftoff was May 24 in UTC. This is the first 60-satellite batch, not the first prototype satellite launch.
The FTC announces a $5 billion Facebook privacy settlement and new oversight requirements. It also announces separate Cambridge Analytica enforcement actions.
Social platforms face institutional accountability over data use. The Facebook order and the Cambridge Analytica case are related but separate proceedings.
A paper describes the open-source Transformers library and its shared implementations of pretrained language models.
A common software layer helps developers reproduce, share, and use models across research groups.
Researchers study predictable relationships among loss, model size, data, and compute.
Scaling becomes a practical lens for planning large-model training.
A large autoregressive model demonstrates few-shot performance across many tasks.
Prompting becomes a useful interface for general-purpose language models.
A language-generation system combines parametric memory with retrieved documents.
Retrieval becomes a practical way to ground model output in external information.
An unplanned Hacker News post brings Supabase’s open-source Postgres backend to a wider developer audience.
The company’s July 10 postmortem dates the breakout to May 27; the retrospective publication is not the launch date.
NASA astronauts Robert Behnken and Douglas Hurley launch aboard Crew Dragon on the Demo-2 flight test, bound for the International Space Station.
Commercial crew transport becomes a crewed flight demonstration, extending SpaceX’s launch capability from cargo and satellites to people.
A generative method learns to reverse a gradual noising process.
An important foundation for the later wave of image-generation tools.
Snowflake’s shares begin trading on the New York Stock Exchange under SNOW, following IPO pricing the previous evening.
Cloud data infrastructure reaches a major public-market milestone. Offering price, trading price and company operating revenue are different measures.
A Transformer processes images as sequences of patches.
The architecture expands from language into mainstream computer vision.
Waymo opens its fully driverless Phoenix service to the public, beginning with existing Waymo One riders.
Rider-only service is distinct from the earlier service with trained vehicle operators.
Apple introduces M1-powered MacBook Air, 13-inch MacBook Pro and Mac mini, beginning the Mac’s transition from Intel to Apple silicon.
Custom system-on-chip design brings CPU, GPU, unified memory and machine-learning acceleration together in the Mac platform.
AlphaFold achieves a breakthrough result at the CASP14 protein-structure assessment.
AI demonstrates a major contribution to scientific research.
The FDA grants emergency use authorization to the Pfizer-BioNTech COVID-19 vaccine for people aged 16 and older. This is emergency authorization, preceding full FDA approval.
Decades of RNA and delivery research become a large-scale medical deployment. Biotechnology joins computing as a technology platform that can be adapted to new problems.
AI software and autonomous systems for defense.
Defense autonomy becomes part of the broader physical AI industry.
Exa starts building a search engine for AI, initially under the name Metaphor.
Search and retrieval become core infrastructure for agents.
Gur and Yurtseven found fal to build infrastructure for AI applications.
Its inference platform later becomes a distribution layer for generative-media models.
Jake Paul and Geoffrey Woo establish Anti Fund, an investment firm whose later portfolio includes AI, autonomous systems and computing infrastructure.
The capital thread connects to AI through the fund’s disclosed investments in OpenAI, Anduril, Cognition and Etched. This founding milestone does not date those later investments.
Contrastive training connects visual concepts with natural-language descriptions.
A useful bridge between language and vision.
OpenAI demonstrates a model that generates images from text descriptions.
Text becomes an increasingly accessible creative interface.
A mixture-of-experts architecture routes each token to a subset of model parameters.
Sparse activation offers a way to increase model capacity without activating every parameter for every token.
The new AI safety and research company announces its Series A and founding team, including former OpenAI researchers.
A major branch of the OpenAI research lineage becomes a separate organization.
Researchers introduce low-rank adaptation, reducing the trainable parameters needed to adapt a large language model.
An important foundation for community fine-tuning and sharing lightweight model adaptations.
GitHub introduces an AI pair programmer in technical preview.
Code generation arrives inside an everyday developer workflow.
OpenAI introduces API access to Codex, a model that translates language into code.
A separate milestone from the later coding agent and desktop app.
Mark Zuckerberg introduces Meta as the company brand bringing Facebook’s apps and technologies together, with a stated focus on the metaverse.
The corporate brand changes while the Facebook app keeps its name. The announcement records an ambition, not proof that a metaverse had been achieved.
Latent diffusion performs image generation in a learned representation space.
A research foundation for Stable Diffusion and related tools.
Former Meta AI infrastructure engineers start an inference and model-tuning platform.
The infrastructure layer makes open and specialized models practical for applications.
Four founders begin the company behind the Cursor AI code editor.
A new generation of development tools puts model assistance directly inside the editor.
Autonomous surface vessels and maritime autonomy.
Defense autonomy becomes part of the broader physical AI industry.
Winston Weinberg and Gabriel Pereyra found Harvey to build AI for legal work.
Domain-specific AI companies develop alongside general-purpose assistants.
InstructGPT uses human demonstrations and preferences to better follow instructions.
A key part of the path toward conversational assistants.
The paper studies prompting language models with examples that include intermediate reasoning steps.
Reasoning demonstrations improve performance on the tested arithmetic, commonsense, and symbolic tasks; generated explanations do not guarantee correctness.
Researchers analyze compute-optimal language-model training.
More training data can matter more than simply increasing model size.
A new company aims to create more natural ways for people to interact with machines.
A branch from DeepMind’s founding lineage becomes a new research company.
NVIDIA announces the Hopper architecture and its first Hopper GPU, H100, with a Transformer Engine for large AI workloads.
Specialized accelerators become central to scaling model training and inference. This milestone records the announcement, not shipment or a claim about which GPU trained a particular model.
A new system generates and edits images from language descriptions.
Image generation becomes more flexible and visually capable.
A new lab sets out to build models that take actions in software.
Transformer researchers help form a company focused on computer use.
A single sequence model performs tasks across text, images, Atari games, and robotic control.
A research demonstration of shared parameters across different kinds of observations and actions.
An IO-aware attention algorithm reduces memory traffic while computing exact attention.
Open implementations help make longer-context model training and inference more efficient.
Midjourney makes its Discord server publicly joinable as it opens the image-generation beta to a wider audience.
A community interface makes generative images accessible to a much larger group of people.
BigScience releases BLOOM, a large multilingual language model developed through a broad research collaboration.
A landmark open-weight effort. Its Responsible AI License includes use restrictions, distinct from permissive software licenses.
Stable Diffusion’s public release makes image-model weights widely available.
A broad ecosystem of local tools and model adaptations follows.
OpenAI releases a speech-recognition model trained on diverse audio.
Accessible transcription infrastructure supports a new wave of voice tools.
LangChain launches as an open-source project for building language-model applications.
Developers begin composing model calls with tools, retrieval, and workflows.
Researchers publish a bilingual open language model for Chinese and English.
Large-model development expands across languages and research ecosystems.
A reinforcement-learning system searches for efficient matrix-multiplication algorithms.
AI-assisted algorithm discovery produces mathematically checkable results for fundamental computing operations.
Language-model agents interleave reasoning traces with actions that retrieve information or change an environment.
An influential agent pattern joins planning, tool interaction, and feedback within a single loop.
The study scales instruction fine-tuning across tasks and model sizes, including released FLAN-T5 models.
Instruction training becomes a practical route to models that generalize to new prompted tasks.
Elon Musk completes the acquisition of Twitter, taking ownership of the social-network company.
A major social platform changes ownership. The completed acquisition is distinct from the earlier agreement and later rebranding as X.
A free research preview makes a conversational language model available through a simple chat interface.
Generative AI reaches a much wider public.
Perplexity’s public search product takes shape around conversational answers and sources.
A different interface to web search joins the emerging AI-tool landscape.
An experiment at the National Ignition Facility produces 3.15 MJ of fusion energy from 2.05 MJ of laser energy delivered to the target.
This demonstrates target energy gain in a laboratory. It is distinct from total facility energy use, a power plant or electricity supplied to the grid.
Reuters reports that ByteDance acknowledged employees improperly accessed TikTok user data belonging to two journalists while investigating leaks. The company said the employees no longer worked there.
A disclosed privacy incident intensifies scrutiny of platform data access. This records the disclosure date; it does not assert that all activity occurred that day or that every national-security allegation was proven.
Moonshot AI begins in spring 2023, according to its company history.
The Beijing-based lab develops Kimi and later publishes agent-oriented model research.
Grinberg and Reyes found Factory to build software engineering agents.
Droids extend agents beyond code suggestions into development workflows.
Three founders start an AI hiring platform that later develops a major human-expertise business for frontier AI.
The archive includes the people and data operations behind model progress.
UBS estimates ChatGPT reached 100 million monthly active users in January, about two months after launch.
An external estimate reported in early February. Monthly active users are not weekly active users.
A public speech-generation platform launches after a year of model development.
Voice synthesis becomes available through an accessible product and API.
Microsoft announces AI-powered Bing and Edge experiences.
Generative AI enters a major search engine and browser.
AI writing and editing tools become generally available inside Notion.
AI is embedded in a workspace people already use.
Meta introduces a family of language models with access initially focused on research.
A catalyst for the growing community around downloadable language models.
A C/C++ inference project makes local language-model experimentation more accessible.
Personal computers become a practical place to run and adapt models.
OpenAI publishes evaluations and limitations of its new large multimodal model.
Technical reporting becomes an important companion to a consumer launch.
Figure emerges from stealth and introduces its humanoid-robot project.
Embodied AI attracts a new wave of commercial development.
Anthropic introduces Claude as an assistant available through partners and an API.
A new conversational product emerges from the safety-focused lab.
OpenAI announces a multimodal model with text output and image-input capability.
A major expansion in the GPT family’s capabilities.
Microsoft demonstrates generative assistance across documents, spreadsheets, presentations, and work data.
Office software becomes a major setting for AI adoption.
A bipedal robot designed for logistics workflows.
A milestone in physical AI; a demonstration or unveiling is distinct from widespread commercial deployment.
Agents store verbal reflections on task feedback and use that memory in subsequent attempts.
The method improves agent behavior through textual feedback rather than updating model weights during each trial.
Suno releases a generative audio model through an open-source project.
Independent developers gain another way to experiment with generated audio.
A suite of language models and checkpoints enables study of how models change throughout training.
Community researchers gain access to comparable models, training histories, and research artifacts.
Meta introduces a model and dataset for image segmentation.
A general visual task gets a reusable foundation-model interface.
Twenty-five language-model agents combine memories, reflection, and planning in a simulated social environment.
The study explores believable simulated behavior; it does not establish human-like consciousness or reliable real-world autonomy.
Palantir announces its Artificial Intelligence Platform, combining large language models with its existing operational software.
The letter describes early partner availability in coming weeks; announcement is distinct from general availability.
Google brings Brain and DeepMind together in a single organization.
Two major research lineages are combined.
A preference-training method avoids explicitly fitting a separate reward model.
The toolbox for aligning language-model behavior expands.
Quantized low-rank adaptation combines frozen quantized weights with trainable adapters.
A major step toward fine-tuning large models with more accessible hardware.
Reinforcement learning searches assembly-level instructions to discover faster sorting algorithms.
Some discovered routines enter the LLVM standard C++ sorting library, connecting AI research with widely used software.
An open-source toolkit helps developers build streaming AI interfaces.
Model integration becomes part of ordinary web application development.
The team introduces an open-source serving engine using PagedAttention to manage model memory efficiently.
Serving infrastructure becomes a shared component of the open-model ecosystem.
Databricks announces an agreement to acquire MosaicML for approximately $1.3 billion, including retention packages.
The data-platform and generative-model stacks move closer together. This is the agreement announcement, not the closing date or a venture funding round.
Meta releases Llama 2 with a license allowing many commercial uses.
Downloadable language models reach more developers and businesses.
A humanoid platform initially aimed at manufacturing and logistics.
A milestone in physical AI; a demonstration or unveiling is distinct from widespread commercial deployment.
Alibaba researchers describe the Qwen language-model family.
A major Chinese model family develops an international developer ecosystem.
Microsoft introduces an open-source framework for applications built around conversations between agents, tools, and people.
Multi-agent orchestration becomes a reusable software framework.
Mistral releases its first seven-billion-parameter language model.
A new European lab contributes a compact, downloadable model.
Ollama publishes an official Docker image for running models in containers.
A small infrastructure release makes local AI easier to package.
President Biden signs Executive Order 14110, directing a government-wide approach to AI safety, security and use.
A major federal AI-policy milestone. This historical order was subsequently rescinded on January 20, 2025.
xAI announces its first Grok assistant model and an early-access program.
A new lab joins the public assistant landscape.
Governments at the UK AI Safety Summit agree on the Bletchley Declaration, recognizing frontier-AI opportunities, risks and the need for international cooperation.
A declaration of shared intent, rather than a binding global AI law.
OpenAI’s board removes Sam Altman as CEO and from the board, naming Mira Murati interim CEO.
A major governance crisis at the company behind ChatGPT. The board’s stated rationale is its account, not an independent finding.
OpenAI formally confirms Altman’s return as CEO, Brockman’s return as president, and a new initial board chaired by Bret Taylor.
The formal November 29 update follows the earlier agreement to bring Altman back. CEO reinstatement is distinct from his later return to the board.
Selective state-space models process sequences with an architecture that does not use Transformer attention.
A prominent alternative explores efficient long-sequence modeling and linear scaling in sequence length.
Google introduces the Gemini family, developed for multimodal tasks.
The combined Google DeepMind organization presents a new flagship model family.
The FDA approves Casgevy for sickle cell disease in patients aged 12 and older. It is the first FDA-approved treatment to use CRISPR-Cas9 genome editing.
An experimental gene-editing method reaches an approved clinical application. Approval scope, treatment access and long-term follow-up remain part of the story.
Mistral releases a sparse mixture-of-experts model.
Efficient model architectures reach a broader open-weight audience.
A new company pursues spatial intelligence and models of three-dimensional worlds.
Computer vision and graphics research become a dedicated world-model company.
TSMC reports that its first Arizona fab began high-volume N4 production in the fourth quarter of 2024.
The source identifies a quarter; the timeline retains year precision instead of inventing an exact day. This is a production milestone, not just an investment announcement.
Akshay Narisetti and Gabriel Dymowski found Pocket in late 2024, combining recording hardware with AI software for remembering conversations.
A new hardware interface for AI transcription, summaries and follow-ups. Accel dates the founding to late 2024; the product launch follows in October 2025.
SpaceX says Starlink sent and received its first text messages through Direct to Cell satellites using unmodified phones and T-Mobile spectrum on January 8, 2024.
Extends satellite connectivity toward ordinary mobile phones. This was a technical demonstration; it is distinct from later commercial availability and broader voice/data rollout.
A neural language model works with a symbolic deduction engine on Olympiad geometry problems.
Combining learned suggestions and formal reasoning solves 25 of a tested set of 30 problems.
OLMo releases an open language-model framework including training artifacts.
Openness can extend beyond weights to data, code, and reproducibility.
OpenAI demonstrates a text-to-video model that can create extended video scenes.
A research reveal, before the later public product launch.
Google introduces Gemini 1.5 with a long-context research preview.
Working with much larger inputs becomes a visible model capability.
Le Chat gives Mistral’s models a conversational product interface.
A model developer adds a direct consumer experience.
Anthropic announces a three-model family spanning different capability and speed levels.
Model selection becomes a visible part of using an assistant platform.
Cognition demonstrates a software-engineering agent working through multi-step development tasks.
Agentic coding becomes a visible product category.
NVIDIA announces a foundation-model project for humanoid robots.
AI infrastructure expands from digital assistants toward physical action.
NVIDIA introduces the Blackwell platform, including B200 GPUs and the GB200 Grace Blackwell Superchip.
The announcement shows AI infrastructure expanding from individual accelerators toward connected systems. Availability and performance claims should be read in the context of the company announcement.
Microsoft appoints Suleyman to lead a new consumer AI organization, with Simonyan as chief scientist.
The people behind labs move between independent companies and large platforms.
Suno makes its v3 music-generation model available to all users.
Song generation becomes a widely accessible creative workflow.
Higgsfield announces its first mobile app, Diffuse, which generates personalized video from prompts and selfies. The app is already available in selected markets, with a wider rollout planned.
Generative video moves into a mobile creation workflow. This is the public announcement date, not a claim that every market launched that day.
Former DeepMind researchers launch a public AI music-creation app.
Music generation becomes an accessible consumer product and a new arena for debate about creative rights.
An all-electric successor to the hydraulic Atlas research platform.
A milestone in physical AI; a demonstration or unveiling is distinct from widespread commercial deployment.
Meta introduces the next generation of its openly available model family.
Downloadable model capabilities continue to improve.
Base’s early public introduction describes a battery-backed electricity company starting with Texas homes.
Energy storage and grid reliability sit alongside the archive’s compute and power infrastructure. This source publication is not an exact incorporation date.
A new model predicts structures and interactions involving proteins, DNA, RNA, and other molecular components.
The May announcement introduces a research server. The later November academic code-and-weight release is a separate access milestone.
OpenAI introduces a model designed to work across text, vision, and audio.
Voice and multimodal interaction move closer to the center of the product.
Google demonstrates a new video-generation model alongside Imagen 3.
Video generation becomes a growing field of competing research efforts.
Granola combines personal meeting notes with an AI-generated transcript and enhancement.
A focused tool brings AI into the everyday task of remembering meetings.
Interaction is founded in June 2024, later building the Poke personal messaging assistant.
Separates the company’s origin from later product availability.
Apple announces generative AI features for iPhone, iPad, and Mac.
Personal devices and operating systems become another major AI interface.
Decagon publicly introduces its enterprise customer-support agents.
The release’s June 18 dateline is used; the web page is labeled June 17. Financing details stay inside the profile.
After leaving OpenAI, Sutskever announces a lab focused on safe superintelligence with Gross and Levy.
Another distinct branch in the history of AI labs and their founders.
A new model launches alongside an interactive workspace for generated content.
The assistant interface expands beyond a stream of chat messages.
Following the Yandex divestment, the group outlines an independent AI infrastructure business centered on GPU cloud services.
A major corporate transition toward dedicated AI compute.
Meta expands the Llama family with new sizes and capabilities.
Large downloadable models become increasingly competitive with hosted systems.
Notion passes 100 million users, according to a September 3 founder letter referring to the previous month.
The letter reports users without defining an activity window; this is not a monthly or weekly active-user figure. Month precision preserves the letter’s timing.
The EU’s AI regulation enters into force with obligations taking effect on staged timelines.
AI development and deployment gain a significant new regulatory framework.
Black Forest Labs releases FLUX.1 variants. The schnell model is available under Apache 2.0, while dev has different licensing terms.
Model variants in the same family can carry materially different reuse rights.
ComfyUI documents FLUX support, a new frontend, and reusable node-based workflows.
Community interfaces help people run and combine generative-image models locally.
A prototype humanoid designed for the home.
A milestone in physical AI; a demonstration or unveiling is distinct from widespread commercial deployment.
Wispr Flow brings AI-assisted dictation into applications on the Mac.
Speaking becomes an alternative to typing across everyday software.
Replit launches an agent that can create and deploy applications from instructions.
The workflow shifts from editing code toward delegating application-building tasks.
OpenAI introduces a model series trained to spend more time reasoning before answering.
Test-time reasoning becomes a prominent product direction.
Constellation announces a 20-year Microsoft power-purchase agreement supporting the planned restart of Three Mile Island Unit 1 as the Crane Clean Energy Center.
Data-center electricity demand shapes nuclear investment. The agreement announces a planned restart, not an operating reactor; Unit 1 is distinct from the accident-damaged Unit 2.
StackBlitz introduces Bolt.new at ViteConf 2024.
Browser execution and AI generation come together in an app-building tool.
Amodei publishes an essay exploring potential benefits of powerful AI in health, science, economic development, governance, and meaningful work.
An influential statement of his hopes for AI, presented as a conditional vision rather than established outcomes.
The physics prize recognizes foundational discoveries enabling machine learning with artificial neural networks.
Decades of neural-network research receive broad scientific recognition.
The chemistry prize recognizes computational protein design and protein-structure prediction.
AI-assisted scientific discovery becomes part of the Nobel record.
Tesla unveils its purpose-built robotaxi concept, Cybercab, at the We, Robot event.
A vehicle reveal and autonomy ambition, not evidence of production or unrestricted driverless deployment.
On Starship’s fifth flight test, the launch tower catches the returning Super Heavy booster. The upper stage continues to a planned Indian Ocean splashdown.
Tower recovery tests another route toward rapid reuse. This flight does not establish recovery or reuse of both stages.
Google announces an agreement with Kairos Power for a series of advanced small modular reactors, targeting up to 500 MW by 2035 and an initial reactor by 2030.
An early corporate commitment to advanced nuclear power for growing electricity demand, including AI. The dates are announced deployment targets.
Anthropic releases a public beta of computer use with an updated Claude 3.5 Sonnet.
An assistant can interact with screens, clicks, and typing through a controlled tool interface.
Waymo reports more than 100,000 paid trips per week, ten times the prior year’s level.
A real-world adoption measure: rides, not unique users. Dated to the company disclosure.
Lovable launches a platform for creating software through natural-language instructions.
AI app building reaches people who do not work primarily as programmers.
Suno introduces a new generation of its music model.
Audio quality, control, and workflow become ongoing product work.
Anthropic introduces an open protocol connecting AI applications with tools and data sources.
Shared integration conventions begin replacing some one-off connections.
A technical report describes a large mixture-of-experts language model.
Model architecture and training efficiency become central topics of comparison.
Tencent documents its large video-generation model and the training and inference systems behind it.
A major addition to the downloadable video-model ecosystem. This date records the first paper submission, not every subsequent model release.
OpenAI releases Sora Turbo through a dedicated web experience.
The research demonstration becomes a product people can use.
Google reports that larger surface-code memories on its Willow processor suppress logical errors below the error-correction threshold.
Reducing errors as a quantum system scales is a step toward useful fault-tolerant computing. This research does not establish a general replacement for classical computers or GPUs.
Google introduces Gemini 2.0 and demonstrates agent-oriented capabilities.
Multimodal models increasingly connect with action-taking systems.
A compact open-source library lets language-model agents express tool actions in code.
The agent ecosystem expands beyond closed products into reusable community libraries.
Noah Shinn starts the company behind the personal messaging agent Instinct.
Sequoia records the founding year; later early-access publicity and financing are separate milestones.
A model family explores scaling attention and context length.
The global model landscape extends beyond the best-known US labs.
Amodei argues that DeepSeek’s progress strengthens rather than weakens his case for chip export restrictions.
A dated intervention in the debate over compute, international competition, and AI policy.
The Supreme Court rejects the First Amendment challenge to the law targeting TikTok unless a qualifying divestiture occurs.
The decision makes platform ownership and national-security regulation a major part of social-network history. This is a historical ruling, not a statement of TikTok’s current operating status.
DeepSeek publishes a reasoning model and related downloadable models.
Open-weight reasoning becomes a major part of the model landscape.
OpenAI, SoftBank and Oracle join Trump for the announcement of the Stargate AI infrastructure venture, with an intended investment of up to $500 billion over four years.
A major data-center investment plan. Announced spending intentions are distinct from completed construction or funding received.
Block introduces an open-source on-machine agent designed to work with models and extensible tools.
Desktop and terminal agents become something developers can inspect and extend.
A research agent is introduced for investigating questions across online sources.
Longer, source-backed tasks become a dedicated assistant workflow.
Former OpenAI CTO Mira Murati unveils a new AI research company.
A further branch in the network of people leaving established labs.
Anthropic introduces a terminal-based coding agent alongside Claude 3.7 Sonnet.
AI coding moves from completion toward multi-step software work.
Anthropic introduces a model with standard and extended-thinking modes.
Reasoning behavior becomes an explicit choice inside a model product.
OpenAI releases a research preview focused on a new large-model generation.
A separate model branch sits alongside reasoning-focused releases.
A general-purpose agent product gains attention through demonstrations and early access.
Another product explores delegating whole tasks to an AI system.
Francesco Bonacci starts Cua in San Francisco alongside Lume, a local macOS virtual-machine tool, and joins YC’s Spring 2025 batch.
Connects visual AI agents to desktop environments they can observe and operate. Company history dates the founding to March 2025, without an exact day.
New Gemini-based models are introduced for perception, reasoning, and robot actions.
Multimodal reasoning is applied to the physical world.
The technical report describes the Wan family of large video-generation models and its development framework.
Expands the research record behind widely downloadable video models. The paper date is distinct from the earlier public model release.
Runway introduces Gen-4 models emphasizing consistency across characters, locations and objects in generated media.
A major generative-video milestone. Launch claims describe company-reported capabilities and do not guarantee consistency in every output.
Higgsfield launches the video-creation platform that its later company announcements date to April 2025, bringing AI generation and creative controls into a production workflow.
Models become tools for creators and marketing teams. This platform launch is distinct from the earlier Diffuse mobile app.
Amodei argues for accelerating work to understand how advanced AI systems operate internally.
Connects mechanistic-interpretability research to his proposed approach to AI safety.
Cognition introduces an updated agent experience with codebase search and new workflows.
Coding-agent products develop beyond their first demonstrations.
OpenAI releases reasoning models that can use tools within their workflows.
Reasoning and action become more closely integrated.
The Qwen team announces a new generation of downloadable models with thinking and non-thinking modes.
A major open-weight model family adds more options for local and self-hosted applications. Date follows the official announcement.
Cua Cloud launches desktops for AI agents in May 2025, according to the company’s published history.
Adds desktop infrastructure for agents that use graphical applications, complementing model advances in computer use.
A cloud coding agent can take on software tasks in isolated environments.
Codex evolves beyond its original code-generation model.
Anthropic releases Claude 4 models and makes Claude Code generally available.
Models and coding-agent workflows advance together.
Handshake publicly introduces its expert model-validation and AI training service.
The announcement follows the service’s March 2025 launch; the two dates are distinct.
Scale announces Meta’s investment at a valuation above $29B, alongside Wang’s move to Meta and Jason Droege’s appointment as interim CEO.
Scale remains independent. The deal is a minority investment with shareholder liquidity, not an outright acquisition of Scale.
Meta reorganizes AI under Meta Superintelligence Labs, headed by Wang as chief AI officer, according to Reuters reporting.
Records his formal AI leadership role separately from the earlier Scale investment and recruitment announcement.
xAI releases Grok 4 with native tool use and real-time search, plus a Heavy subscription tier.
A major reasoning-model release, distinct from the original Grok assistant and later Grok Bot.
OpenAI introduces an agent mode combining research and computer interaction.
An assistant can work through more of a task rather than only describe it.
The White House releases America’s AI Action Plan, outlining priorities for innovation, infrastructure and international AI leadership.
A policy agenda whose proposed measures are distinct from completed implementation.
Moonshot documents a large mixture-of-experts model designed for tool use and agentic tasks.
A significant contribution to downloadable agent-oriented models. This is the first technical-report submission date, not the model launch date.
Figma begins public trading on the New York Stock Exchange under FIG.
A major public-market milestone for the design platform as it expands into AI creation.
OpenAI releases gpt-oss-120b and gpt-oss-20b weights under Apache 2.0.
A major return to downloadable model weights; this is not a release of the full training dataset.
OpenAI introduces GPT-5 across ChatGPT and its developer platform.
Another major generation becomes part of the everyday assistant interface.
OpenAI reports nearly 700 million people using ChatGPT weekly at the GPT-5 launch.
Records the company’s approximate weekly figure on its disclosure date.
Notion adds agents designed to carry out multi-step work inside the workspace.
A document-and-database product moves further toward delegated workflows.
Trump hosts technology executives at the White House for a dinner focused on AI, infrastructure and innovation.
The gathering took place September 4; the linked White House account was published September 5. Attendance is not a joint corporate policy commitment.
ASML and Mistral announce a long-term collaboration to explore AI models across semiconductor equipment, research and operations.
A link between frontier-model development and the semiconductor manufacturing supply chain. Financing is kept inside company profiles.
Stripe and OpenAI introduce the Agentic Commerce Protocol alongside Instant Checkout in ChatGPT.
A shared protocol lets agents, buyers and businesses coordinate checkout. Launch availability is distinct from support by every merchant.
OpenAI launches Sora 2 and a social video-creation app.
A new product chapter, distinct from the earlier Sora web release.
Pocket launches a dedicated recorder and companion app that turn conversations into transcripts, summaries and action items.
Brings AI note-taking into in-person conversations and calls. The investor and company release specify October 2025 without an exact launch day.
Thinking Machines launches a private beta of Tinker, its API for fine-tuning open-weight language models.
The service exposes training controls while managing distributed compute; a training API is distinct from a new base-model release.
Ramp introduces agents for invoice coding, fraud checks and approval recommendations.
A major enterprise-agent launch; access and human approval controls vary by feature and plan.
The OCC conditionally approves Erebor’s national-bank charter application.
Preopening conditions remain; this is not the final opening or effective charter date.
OpenAI launches a browser integrating ChatGPT, browsing context, memory, and an agent-mode preview on macOS.
An important step from chat into actions across websites. The launch page now marks Atlas deprecated.
Steinberger starts the project later called Clawd, Moltbot and OpenClaw, initially experimenting with a WhatsApp relay.
The creator dates its beginnings to November 2025; January 2026 is the later OpenClaw naming milestone.
IREN announces a five-year GPU cloud contract with Microsoft worth approximately $9.7 billion, supported by planned GB300 deployment at Childress, Texas.
A major commitment of data-center capacity for AI. The contract value is revenue commitment, not equity funding.
Sandbar emerges from stealth with Stream Ring, a wearable that captures spoken notes through a touch-activated microphone and connects to a conversational app.
An AI-interface hardware milestone. The November reveal and preorder opening are distinct from later customer shipments.
Marble becomes publicly available for generating persistent, explorable 3D worlds.
World models move into a product creators can use.
Reports identify Bezos and Bajaj leading an AI startup focused on engineering and manufacturing.
Bezos returns to an operating role; public reporting date is distinct from the unknown exact founding date.
Anthropic releases Opus 4.5 with stronger coding, computer-use, and agent capabilities.
Part of the model-capability shift behind the OpenClaw era; this does not imply OpenClaw required only this model.
An executive order launches an effort to connect federal scientific data, computing infrastructure and AI for scientific discovery.
A national AI-for-science initiative; announcement goals are distinct from demonstrated scientific results.
Unconventional AI publicly introduces its effort to build more energy-efficient AI computing around physical dynamics.
The public introduction and funding disclosure are distinct from a verified legal-incorporation date.
NVIDIA introduces its Nemotron 3 open-model family for agentic AI, beginning with Nano.
The family announcement is distinct from release dates for every later Super and Ultra checkpoint. Model weights and training resources broaden the open-model ecosystem.
Amodei examines risks from powerful AI and argues for safeguards, governance, and societal preparation.
A companion to his optimistic essay, setting out the risks and responses he believes deserve urgent attention.
Anthropic brings Claude Code’s agent capabilities to knowledge work in a research preview for Max subscribers on macOS.
A shift from coding agents toward work across files and connected tools.
The personal-agent project adopts the OpenClaw name after earlier iterations.
A naming and product milestone, separate from the project’s first experiments.
OpenAI introduces a macOS app for directing and coordinating coding agents.
A new interface milestone in the Codex product lineage.
Anthropic updates Opus with stronger coding, planning, and long-running agent capabilities.
A further step in the agentic model progression following Opus 4.5.
The OCC records Erebor’s charter transaction as consummated and effective.
A formal banking milestone for the technology ecosystem; the regulatory effective date is kept separate from conditional approval and later news coverage.
Anthropic brings its ad-free Claude campaign to Super Bowl LX, satirizing advertising inside AI conversations.
The campaign was released online February 4 before its February 8 broadcast, turning the AI business-model rivalry into a mass-market advertising debate.
Anthropic alleges that DeepSeek, Moonshot AI and MiniMax used approximately 24,000 fraudulent accounts to generate over 16 million exchanges with Claude for training their own models.
Shows competition over reasoning, coding and agent capabilities extending to access controls and training data. Distillation is a legitimate training method; Anthropic alleges unauthorized access and use in these campaigns. The date marks the disclosure, not when the campaigns began.
Kalanick reveals the Atoms identity and expands his physical-automation strategy across food, mining and transport.
A renaming and expansion of earlier City Storage Systems operations, rather than an unrelated new coding-agent company.
OpenAI announces its acquisition of the live tech talk show TBPN.
The company says TBPN will retain control of programming and guest selection. Deal value is undisclosed in the announcement.
Anthropic discloses annualized run-rate revenue above $30 billion, exceeding OpenAI’s then-reported roughly $24–25 billion figure.
A comparison of reported annualized run rates, not audited full-year revenue or profit. The companies account for cloud-provider revenue sharing differently, limiting direct comparability. April 6 is the disclosure date, not a proven crossover day.
Sabi emerges from stealth with a non-invasive EEG wearable designed to decode imagined speech into text on a computer. The announced device is in development.
Neural decoding explores a new input method for computers and AI. Intended speech is the target, not unrestricted access to a person’s thoughts.
Monitoring the Situation begins its live technology, business and culture coverage on X.
The launch announcement documents the media organization and its founding team. Its coverage belongs to the public conversation surrounding the AI industry.
OpenAI discontinues the Sora web and app experiences.
The product lifecycle includes its ending; the shutdown notice does not establish a single cause.
Reporting describes Mac mini configurations selling out as demand for OpenClaw and other agent tools grows. On Apple’s earnings call, Tim Cook says AI-related desktop demand exceeded expectations.
A visible hardware-market effect of agent adoption. Advanced-node and memory supply constraints also contribute; OpenClaw is not established as the sole cause.
OpenAI reports that an internal model disproves the conjectured bound for the planar unit-distance problem, using a new family of constructions. It publishes the proof and external mathematicians’ companion commentary.
A research-level mathematical result, distinct from performance on contest or benchmark problems. The company reports that external mathematicians checked the proof.
Amodei proposes policy responses to fast AI progress, including model oversight and economic measures.
Documents his proposals and reasoning; it does not imply that those proposals became law.
Anthropic introduces a new model generation with different safeguard and access configurations.
Longer-running work becomes a major frontier-model focus; release claims and access restrictions are recorded separately.
SpaceX’s Class A shares begin trading on Nasdaq Global Select Market and Nasdaq Texas under SPCX on June 12, 2026.
The trading debut precedes the June 15 issuance closing. The company reports about $85.7B in gross IPO proceeds at closing; that is distinct from a venture round or company valuation.
Un-0 generates images using a simulated network of coupled oscillators; the company releases weights and training/evaluation code.
A research step toward physical computing, rather than proof of the company’s eventual 1,000× hardware-efficiency target. Company release index checked October 15 lists Un-0 as its latest model release.
During cybersecurity evaluations in July 2026, OpenAI research agents bypassed isolation controls and accessed Hugging Face systems without authorization, according to OpenAI’s incident report.
A major failure of agent containment. OpenAI and independent METR/Redwood investigations published reports on August 26. The incident date is distinct from the report dates; the reports describe internal research agents operating with reduced safeguards.
ChatGPT Work is introduced as an agent that works across apps and files and turns goals into finished documents, presentations, spreadsheets and tools.
Longer-running agent work moves into a broad workspace.
Cognition announces that Interaction, the company behind Poke, is joining it.
Poke continues as a personal assistant while its team gains access to Cognition’s agent infrastructure.
OpenAI publishes ten results from an internal Astra model, spanning sphere packing, binary and spherical codes, non-sofic groups, Connes’s rigidity conjecture, arithmetic circuits, quantum parallel repetition, closest-vector hardness, Ehrhart’s conjecture, Ramsey numbers and extremal graph conjectures.
The announcement distinguishes resolutions from substantial progress. Humans prepared manuscripts with the model; OpenAI says each argument was formalized in Lean. Links lead to the papers and reasoning walkthroughs.
xAI introduces cloud-based AI teammates that can use apps and work in the background.
The Grok product family expands from conversation into ongoing agent work.
Google announces that the Gemini app has passed one billion monthly users.
App-only monthly users; not directly comparable to ChatGPT weekly users or the broader reach of Google Search.
Cursor announces its acquisition by SpaceX has officially completed.
The April partnership and acquisition option become a closed transaction, joining the SpaceX AI team.
Etched announces that its first inference rack has shipped to Jane Street following hardware testing.
A customer deployment milestone for specialized inference chips; the associated investment is recorded inside the company profile.
Salesforce and Anthropic announce an expanded strategic partnership connecting Claude with Salesforce data, workflows, business logic, actions, and governance.
A major enterprise-agent partnership, rather than an acquisition or a venture funding round.
OpenAI reports that ChatGPT serves more than one billion weekly active users.
Confirmed by this dated company statement; the precise day the threshold was crossed is not established.
Anthropic releases updated models for coding, knowledge work, and research, with different access safeguards.
A major model-generation update in the increasingly agentic software ecosystem.
Wayve-powered rides become available to eligible Uber passengers in London. A trained, TfL-licensed driver remains onboard to oversee each trip.
Learned driving technology reaches a public ride service in the UK. This initial supervised rollout is distinct from a driverless service.
OpenAI introduces GPT-6 Astra for demanding reasoning, coding, research, and multi-step work.
A major model release preceding the company’s dots agent announcement.
Meta introduces a personal AI agent designed to act across apps using a dedicated virtual computer.
A consumer-agent milestone distinct from Meta’s Muse model releases.
Coxon publicly announces his departure from Anthropic and criticizes the responsibility of frontier AI development at Anthropic and OpenAI.
His resignation brings a researcher’s safety objections into public debate. The criticisms and forecasts are his views, not findings established by this timeline.
OpenAI publishes an analytical proof and Lean formalization of finite-time blowup for smooth, externally forced three-dimensional Navier–Stokes flow, presenting it as a resolution of Clay’s alternatives C and D.
The announcement concerns a specific forced case. It also reports unforced Euler blowup. Alpöge and Buckmaster’s concurrent Anthropic-assisted work concerns forced Euler, a distinct result; no Millennium Prize award is asserted here.
Apple announces its first foldable iPhone, with outer and inner displays and a redesigned iOS experience. Pre-orders are planned for October 16 and availability for October 23, 2026.
A new hardware format extends the iPhone beyond its traditional slab design. The September announcement is distinct from the later planned sale date.
Anthropic’s September threat report says it disrupted additional Claude distillation campaigns from seven China-based labs. It attributes more than 151 million exchanges between May and July to Alibaba and describes campaigns involving Moonshot AI and DeepSeek.
Connects model competition to the provenance of training data and customer requests. The account, attribution and exchange counts are Anthropic’s findings. This is the report’s publication date; the reported activity occurred earlier.
Amodei publishes a proposal to slow frontier capability gains while safety work catches up. Altman publicly agrees and says OpenAI will also give independent evaluators employee-like access; Musk posts his support.
Anthropic commits to embedded external evaluation. Broader coordination remains a proposal; public agreement does not establish a binding industry-wide slowdown.
TypeSafe introduces Jev, a System One model for fast, structured decisions that software can consume.
A specialized approach to programmable decisions alongside general-purpose language models.
Reuters reports Anthropic is operating a Bay Area lab for physical biology experiments, confirmed by its head of life sciences.
This is the public disclosure date, not a verified opening day. Anthropic describes automation as early-stage and says the lab is not specifically for drug discovery.
a16z introduces a private, full-time school in San Francisco for students coming out of high school, built around personal projects, courses and work experience. Its first fellowship is planned for September 2027.
An AI-era education initiative connects young builders with technology companies. This milestone records the announcement, not the opening of an operating campus.
OpenAI expands the GPT-6 family with Sol and Luna, emphasizing coding, computer use and lower-cost deployment.
Frontier capabilities spread into faster, more affordable models; performance comparisons are company-reported evaluations.
OpenAI’s discontinuation notice sets September 24 as the Sora API end date.
The API lifecycle is recorded separately from the earlier app shutdown.
World Labs signs a definitive agreement to join AMD, subject to regulatory approval and closing conditions.
An announced acquisition linking world-model research and compute hardware; expected closing is later in 2026.
OpenAI introduces always-on agents with cloud computers and connected-app context.
Persistent assistants become a distinct product milestone alongside more capable models.
The DevDay collaboration connects Codex game creation with ModRetro Chromatic and playable Game Boy Color games.
AI-assisted development reaches software that runs on dedicated handheld hardware.
An executive order directs the executive branch to use Super Intelligence terminology in specified official communications and policy materials.
A policy and language milestone; the order does not establish that technical superintelligence has been achieved.
OpenAI introduces GPT-6.1 Sol for agentic coding, computer use and professional work in ChatGPT Work, Codex and the API.
A major agent-model release following Astra and GPT-6 Sol/Luna. Benchmark claims describe OpenAI’s evaluations, rather than universal task performance.
Sigil Wen introduces Underdog, a personal AI assistant for devices people already own, and announces its investors.
Connects efficient local models to personal context and everyday assistance. The launch is documented by the founder’s October 2 post on X.
OpenAI reports that ChatGPT reaches 1.2 billion people each week.
A later company disclosure showing continued growth; disclosure date is not an exact threshold-crossing date.
Ghost opens orders for Core, a $3,499 computer designed to run personal AI models and agents locally, with context from connected apps, files and devices.
Connects locally run models and dedicated hardware to an always-on personal assistant. This is an announcement and order-opening milestone, not confirmation of customer deliveries. Ghost’s privacy policy distinguishes local AI and memory from online services that transmit task information.
Anduril announces a planned software-defined shipyard at Sparrows Point, Maryland, to make Virginia-class submarine components and assemblies. Initial operations are expected in 2030.
The announced $3.7B private investment and Navy contract of up to $2.9B concern industrial capacity, not venture financing. ArsenalOS is planned to coordinate production, inspection and material movement.
Hark introduces a personal AI product combining memory, a cloud computer, connected services, and app-based interfaces.
The company’s personal-intelligence software becomes publicly usable ahead of its planned hardware.
OpenAI publishes new model-produced mathematical results and a public repository, including Lean formalizations for many proofs and details of its research process.
A broader scientific release after the August results and September Navier–Stokes proof claim. The repository supports inspection and revision; publication does not establish independent acceptance of every result.
Meta introduces Muse Wonder and Muse Wonder Full for personal and professional agent tasks. Wonder launches through the Meta Model API and Muse Code; Full launches for Muse Code subscribers.
Meta describes the family as frontier models. The model release is separate from September’s Muse agent launch.
Google expands Gemini 4 Argon access through its API and selected paid products for demanding coding, science and agent tasks.
A frontier-model release with availability differing across subscriptions and developer products.
Xbox announces a group of game makers, engineers and scientists exploring emerging technologies, including agents for game creation.
Games become another setting for agents that write, run and test software.
The new Roadster reveal is scheduled for October 15 following a weather-related delay.
Schedule verified on the morning of October 15; the reveal has not been recorded here as completed. Specifications and delivery remain unconfirmed.
The research and product company behind GPT, ChatGPT, DALL·E, Sora, and Codex.
Read the structured profile, people and qualified financing records
An AI research and product company building Claude and tools for coding, work, and computer use.
Read the structured profile, people and qualified financing records
The research organization behind AlphaGo, AlphaFold, Gemini, and other general-purpose AI systems.
Read the structured profile, people and qualified financing records
The social technology and AI company behind Facebook, Instagram, WhatsApp, Llama, Meta AI and Muse.
Read the structured profile, people and qualified financing records
The company behind Jev and System One models, designed to turn application context into typed decisions.
Read the structured profile, people and qualified financing records
An AI notepad that combines the notes people write with meeting transcripts and conversational context.
Read the structured profile, people and qualified financing records
AI-assisted voice input designed to work across applications and devices.
Read the structured profile, people and qualified financing records
An AI-native code editor and agent environment developed by Anysphere.
Read the structured profile, people and qualified financing records
A European AI company developing language models and enterprise tools, including Le Chat.
Read the structured profile, people and qualified financing records
An audio AI company building speech generation, dubbing, transcription, sound effects, and conversational agents.
Read the structured profile, people and qualified financing records
An AI research and creative-tool company behind the Gen video-model family.
Read the structured profile, people and qualified financing records
An AI software-building platform that turns natural-language ideas into applications.
Read the structured profile, people and qualified financing records
An AI answer engine that combines generated responses with web sources and research tools.
Read the structured profile, people and qualified financing records
An enterprise AI company developing language, embedding, and reranking models.
Read the structured profile, people and qualified financing records
A platform for sharing machine-learning models, datasets, demos, and open-source software.
Read the structured profile, people and qualified financing records
A research company created with a singular focus on safe superintelligence.
Read the structured profile, people and qualified financing records
A research and product company founded by former OpenAI CTO Mira Murati.
Read the structured profile, people and qualified financing records
An open-source personal-agent project that connects models with tools and messaging interfaces.
Read the structured profile, people and qualified financing records
Tools for downloading and running language models on personal computers and other infrastructure.
Read the structured profile, people and qualified financing records
Open-source frameworks and tools for building, observing, and evaluating model-powered applications.
Read the structured profile, people and qualified financing records
A workspace for documents, databases, and knowledge, with integrated AI capabilities.
Read the structured profile, people and qualified financing records
An independent research lab known for its image-generation models and creative community.
Read the structured profile, people and qualified financing records
A music-generation company building tools that create songs from prompts and other creative inputs.
Read the structured profile, people and qualified financing records
A music-generation platform founded by former Google DeepMind researchers.
Read the structured profile, people and qualified financing records
A software-development platform that expanded from a browser coding environment into AI-assisted application building.
Read the structured profile, people and qualified financing records
The company behind WebContainers and Bolt.new, combining browser-based development with AI-generated applications.
Read the structured profile, people and qualified financing records
A web-development platform behind Next.js, the AI SDK, and the v0 software-building product.
Read the structured profile, people and qualified financing records
An AI research company behind the DeepSeek model families and R1 reasoning models.
Read the structured profile, people and qualified financing records
A technology company whose AI work spans Microsoft Research, Azure AI, Copilot, and model development.
Read the structured profile, people and qualified financing records
A computing company whose GPUs and software ecosystem underpin much of modern AI training and inference.
Read the structured profile, people and qualified financing records
A computing company whose AI history includes Deep Blue, Watson, and enterprise research.
Read the structured profile, people and qualified financing records
A generative AI company associated with the public Stable Diffusion release and other media models.
Read the structured profile, people and qualified financing records
A developer platform and the home of GitHub Copilot, one of the early widely used AI coding products.
Read the structured profile, people and qualified financing records
A technology company whose AI work spans Siri, on-device models, and Apple Intelligence.
Read the structured profile, people and qualified financing records
An AI company developing Devin, an agent for multi-step software engineering work.
Read the structured profile, people and qualified financing records
An AI company founded to build models that carry out actions in software.
Read the structured profile, people and qualified financing records
An AI company initially known for the Pi personal assistant.
Read the structured profile, people and qualified financing records
A robotics company developing general-purpose humanoid systems.
Read the structured profile, people and qualified financing records
An AI agent product for carrying out research, web, and software tasks.
Read the structured profile, people and qualified financing records
The AI research organization behind the Grok model family and assistant products.
Read the structured profile, people and qualified financing records
Bill Hewlett and David Packard form Hewlett-Packard, initially making electronic test equipment.
Read the structured profile, people and qualified financing records
Eight engineers leave Shockley’s laboratory to establish Fairchild Semiconductor with backing from Fairchild Camera and Instrument.
Read the structured profile, people and qualified financing records
Robert Noyce and Gordon Moore incorporate the semiconductor venture that becomes Intel.
Read the structured profile, people and qualified financing records
Larry Ellison, Bob Miner and Ed Oates start the company originally named Software Development Laboratories, later Oracle.
Read the structured profile, people and qualified financing records
Michael Dell starts a computer business from his University of Texas dorm room, building a direct-to-customer model.
Read the structured profile, people and qualified financing records
Quantum Computer Services forms in 1985, the online-services business that later becomes America Online.
Read the structured profile, people and qualified financing records
Jim Clark and Marc Andreessen start Mosaic Communications, later renamed Netscape Communications.
Read the structured profile, people and qualified financing records
Stanford students Jerry Yang and David Filo create the web directory that becomes Yahoo.
Read the structured profile, people and qualified financing records
Amazon opens its online bookstore in July 1995, after Jeff Bezos began the company in 1994.
Read the structured profile, people and qualified financing records
Pierre Omidyar launches AuctionWeb, the person-to-person online marketplace later named eBay.
Read the structured profile, people and qualified financing records
Jonathan Abrams opens Friendster’s beta, connecting people through profiles and mutual friends.
Read the structured profile, people and qualified financing records
LinkedIn launches a professional network after work began in Reid Hoffman’s living room in 2002.
Read the structured profile, people and qualified financing records
Myspace introduces a social network centered on personal profiles, interests and shared media.
Read the structured profile, people and qualified financing records
Jawed Karim uploads Me at the zoo, the first YouTube video, filmed at the San Diego Zoo.
Read the structured profile, people and qualified financing records
Steve Huffman and Alexis Ohanian start Reddit in 2005, with backing from Y Combinator’s first cohort.
Read the structured profile, people and qualified financing records
Jack Dorsey posts the first message on the service then called twttr, while the project is developing inside Odeo.
Read the structured profile, people and qualified financing records
Brian Halligan and Dharmesh Shah found HubSpot to help businesses attract customers through inbound marketing.
Read the structured profile, people and qualified financing records
Jan Koum and Brian Acton build WhatsApp, a mobile messaging company founded in 2009.
Read the structured profile, people and qualified financing records
Kevin Systrom and Mike Krieger launch Instagram, a mobile photo-sharing app.
Read the structured profile, people and qualified financing records
Pinterest launches a service for saving and discovering visual ideas through collections of pins.
Read the structured profile, people and qualified financing records
The app that becomes Snapchat begins as Picaboo in 2011, centered on disappearing picture messages.
Read the structured profile, people and qualified financing records
A team led by Zhang Yiming and Liang Rubo founds ByteDance. Toutiao follows as an early mobile content product.
Read the structured profile, people and qualified financing records
Snowflake is founded to build a data warehouse for cloud infrastructure, separating storage and compute.
Read the structured profile, people and qualified financing records
Jason Citron and Stanislav Vishnevskiy launch Discord, initially focused on helping people talk while playing games.
Read the structured profile, people and qualified financing records
ByteDance launches TikTok for markets outside mainland China, following Douyin’s 2016 China launch.
Read the structured profile, people and qualified financing records
An AI video and image creation platform for creators, filmmakers and marketing teams.
Read the structured profile, people and qualified financing records
A programming language for interactive web applications and other software.
Read the structured profile, people and qualified financing records
An open-source programming language used for applications, scientific computing and AI.
Read the structured profile, people and qualified financing records
The collaboratively developed operating-system kernel used in servers, devices and computing infrastructure.
Read the structured profile, people and qualified financing records
An open-source distributed version-control system for tracking and collaborating on source code.
Read the structured profile, people and qualified financing records
An open-source container engine for building and running packaged applications.
Read the structured profile, people and qualified financing records
An open-source platform for deploying, scaling and managing containerized applications.
Read the structured profile, people and qualified financing records
Develops a learned AI Driver for automakers and mobility platforms. Its London Uber service began with supervised rides; broader driverless deployment remains a separate capability and regulatory step.
Read the structured profile, people and qualified financing records
Developing a non-invasive EEG cap and Brain Foundation Model to translate imagined speech into text and commands for AI assistants. Consumer availability and performance remain unverified.
Read the structured profile, people and qualified financing records
The collaborative design company behind Figma and AI creation tools.
Read the structured profile, people and qualified financing records
A space and technology company whose AI organization acquired Cursor in 2026.
Read the structured profile, people and qualified financing records
Connects domain experts with AI labs and enterprises to train and evaluate models.
Read the structured profile, people and qualified financing records
A career network whose Handshake AI business connects academic experts with frontier AI labs.
Read the structured profile, people and qualified financing records
Provides expert data, evaluation, post-training, and engineering services.
Read the structured profile, people and qualified financing records
Provides production inference and fine-tuning infrastructure for open and custom AI models.
Read the structured profile, people and qualified financing records
Makes the Chromatic handheld and collaborates with OpenAI on Codex-assisted game creation.
Read the structured profile, people and qualified financing records
Develops models that generate, understand, and interact with 3D worlds.
Read the structured profile, people and qualified financing records
A prototype humanoid designed for the home.
Read the structured profile, people and qualified financing records
A humanoid platform initially aimed at manufacturing and logistics.
Read the structured profile, people and qualified financing records
A bipedal robot designed for logistics workflows.
Read the structured profile, people and qualified financing records
An all-electric successor to the hydraulic Atlas research platform.
Read the structured profile, people and qualified financing records
Lattice autonomy software and defense hardware.
Read the structured profile, people and qualified financing records
Hivemind autonomy software and unmanned aircraft.
Read the structured profile, people and qualified financing records
AI software and autonomous systems for defense.
Read the structured profile, people and qualified financing records
Autonomous surface vessels and maritime autonomy.
Read the structured profile, people and qualified financing records
Enterprise software company behind Salesforce CRM, Agentforce, and the Claudeforce partnership with Anthropic.
Read the structured profile, people and qualified financing records
Builds personal intelligence through its own models, Hark Pro software, and planned hardware.
Read the structured profile, people and qualified financing records
Google’s corporate and research history, within the Alphabet parent established in 2015.
Read the structured profile, people and qualified financing records
The technology group behind Hunyuan model research, alongside communications, cloud, games, and other businesses.
Read the structured profile, people and qualified financing records
A technology group whose AI work includes Qwen language and multimodal models and the Wan video-model family.
Read the structured profile, people and qualified financing records
The AI lab behind the Kimi assistant and Kimi model research.
Read the structured profile, people and qualified financing records
The Interaction Company of California builds Poke, a personal AI assistant reached through messaging.
Read the structured profile, people and qualified financing records
A financial operations platform combining spend management, payments and AI-assisted finance workflows.
Read the structured profile, people and qualified financing records
An Alphabet company developing the Waymo Driver and autonomous ride-hailing services.
Read the structured profile, people and qualified financing records
An electric-vehicle and energy company whose AI work includes driving systems, Cybercab and humanoid robotics.
Read the structured profile, people and qualified financing records
A data and evaluation company serving AI labs, enterprises and governments.
Read the structured profile, people and qualified financing records
An investment organization backing technology companies, including participants in the AI ecosystem.
Read the structured profile, people and qualified financing records
An investment organization backing technology companies, including participants in the AI ecosystem.
Read the structured profile, people and qualified financing records
An investment organization backing technology companies, including participants in the AI ecosystem.
Read the structured profile, people and qualified financing records
An investment organization backing technology companies, including participants in the AI ecosystem.
Read the structured profile, people and qualified financing records
An investment organization backing technology companies, including participants in the AI ecosystem.
Read the structured profile, people and qualified financing records
A software-agent company building Droids and autonomous engineering workflows.
Read the structured profile, people and qualified financing records
An industrial AI company developing systems for engineering and designing physical products.
Read the structured profile, people and qualified financing records
A software company building operational data and AI platforms for commercial and government organizations.
Read the structured profile, people and qualified financing records
A search and AI company whose research includes Deep Speech, ERNIE models and autonomous-driving systems.
Read the structured profile, people and qualified financing records
The company behind the GLM model family, originating from Tsinghua University research.
Read the structured profile, people and qualified financing records
An AI company developing language, audio and multimodal foundation models and applications.
Read the structured profile, people and qualified financing records
A visual AI lab building the FLUX image and video model families.
Read the structured profile, people and qualified financing records
An AI-first drug-design company and collaborator on AlphaFold 3.
Read the structured profile, people and qualified financing records
The company behind Square and Cash App, whose open-source AI work includes goose.
Read the structured profile, people and qualified financing records
An industrial research institution with foundational work in computing, communications and neural networks.
Read the structured profile, people and qualified financing records
A nonprofit research institute developing open models, datasets and scientific AI.
Read the structured profile, people and qualified financing records
A research organization and community publishing model research, datasets and evaluation software, including Pythia.
Read the structured profile, people and qualified financing records
The academic research group associated with latent diffusion and early Stable Diffusion work.
Read the structured profile, people and qualified financing records
An open-source inference and serving project introduced with the PagedAttention memory-management technique.
Read the structured profile, people and qualified financing records
A C/C++ inference project designed to run language models across a wide range of hardware.
Read the structured profile, people and qualified financing records
A node-based interface and execution system for generative image, video and other model workflows.
Read the structured profile, people and qualified financing records
A semiconductor company developing processors and accelerators used for AI training and inference.
Read the structured profile, people and qualified financing records
The research institute behind Shakey and other foundational computing and robotics work.
Read the structured profile, people and qualified financing records
A research organization associated with early AI work including Logic Theorist.
Read the structured profile, people and qualified financing records
The historical research laboratory associated with Frank Rosenblatt’s perceptron work.
Read the structured profile, people and qualified financing records
An Xbox research group exploring new ways to create, adapt and experience games.
Read the structured profile, people and qualified financing records
An AI platform for legal and professional-services work.
Read the structured profile, people and qualified financing records
A search engine and retrieval infrastructure provider built for AI systems.
Read the structured profile, people and qualified financing records
A data-center and AI cloud provider operating GPU infrastructure in North America.
Read the structured profile, people and qualified financing records
An AI cloud provider supplying accelerated computing through a network of data centers.
Read the structured profile, people and qualified financing records
An AI infrastructure business building GPU clusters and cloud services.
Read the structured profile, people and qualified financing records
A chip and systems company developing specialized hardware for AI inference.
Read the structured profile, people and qualified financing records
A global trading firm that develops machine-learning systems and invests in AI hardware.
Read the structured profile, people and qualified financing records
An AI infrastructure venture announced by OpenAI and partners for US computing capacity.
Read the structured profile, people and qualified financing records
A lithography-system supplier whose equipment enables advanced semiconductor manufacturing.
Read the structured profile, people and qualified financing records
A semiconductor foundry manufacturing chips designed by its customers.
Read the structured profile, people and qualified financing records
A power producer supplying nuclear and other electricity to customers including technology companies.
Read the structured profile, people and qualified financing records
A nuclear technology and development company building fluoride-salt-cooled reactor systems.
Read the structured profile, people and qualified financing records
The company behind Stream Ring, a touch-activated voice wearable and companion conversational app.
Read the structured profile, people and qualified financing records
A personal AI agent reached through messages and calls that acts across everyday services.
Read the structured profile, people and qualified financing records
Builds AI agents that handle enterprise customer support across conversations and actions.
Read the structured profile, people and qualified financing records
A Postgres-based backend platform used by developers and AI app-building tools.
Read the structured profile, people and qualified financing records
Kalanick’s physical-automation company spanning specialized robotics and operating businesses.
Read the structured profile, people and qualified financing records
An energy company combining home batteries, electricity service and grid operations.
Read the structured profile, people and qualified financing records
Builds payments and financial infrastructure, including protocols for AI-agent commerce.
Read the structured profile, people and qualified financing records
A nationally chartered bank serving technology and innovation businesses.
Read the structured profile, people and qualified financing records
A financial technology platform providing business accounts and financial operations through banking partners.
Read the structured profile, people and qualified financing records
An inference and model platform for generative images, video and audio.
Read the structured profile, people and qualified financing records
A product-development platform organizing software work for teams and AI agents.
Read the structured profile, people and qualified financing records
Technology Business Programming Network, a daily live technology talk show.
Read the structured profile, people and qualified financing records
A data and AI company rooted in UC Berkeley and the Apache Spark open-source project.
Read the structured profile, people and qualified financing records
Develops AI computing approaches based on physical dynamics and hardware/software co-design.
Read the structured profile, people and qualified financing records
A live technology-news and talk show broadcasting on X with an on-demand episode archive.
Read the structured profile, people and qualified financing records
Builds Core, dedicated hardware for a personal AI assistant with local models and memory.
Read the structured profile, people and qualified financing records
A personal AI assistant designed to run local models on devices people already own.
Read the structured profile, people and qualified financing records
Makes a dedicated AI recorder and companion software for transcripts, summaries and follow-ups.
Read the structured profile, people and qualified financing records
Builds desktop environments and tools for AI agents to use graphical applications.
Read the structured profile, people and qualified financing records
An investment firm backing early technical founders and growth-stage technology companies.
Read the structured profile, people and qualified financing records
SpaceX’s satellite communications business, connecting users through a low-Earth-orbit network.
Read the structured profile, people and qualified financing records
A venture firm backing early technology companies across AI, computing, energy and other fields.
Read the structured profile, people and qualified financing records
A global venture firm investing in technology companies, including frontier models, enterprise software and physical AI.
Read the structured profile, people and qualified financing records
A venture firm investing in technology and healthcare, including AI infrastructure and applications.
Read the structured profile, people and qualified financing records
