Archive
- — OpenAI's $70bn revenue was really $50bn
- — Haiku 5.5 meets Luna at a dime
- — OpenAI floods GitHub with AI math proofs
- — Reflection's Beam takes aim at DeepSeek
- — Anthropic lines up a $2 trillion IPO
- — Trump taps his spy chief as AI czar
- — Apple tightens Mac disk access over AI agents
- — OpenAI fires safety researchers as revolt brews
- — Google's Argon reaches the frontier, for defenders only
- — OpenAI ships dots and a bargain Sol
- — Anthropic's IPO sells the risk it fears
- — The safety alarm gets called a moat
- — Rogue-agent tally climbs into the tens of thousands
- — Frontier labs take a beating from Washington
- — Washington claims first dibs on frontier models
- — Meta goes all-in on Muse
- — Two labs cut prices an hour apart
- — Xiaomi crashes the open-weights frontier
- — Alibaba's 7B image model takes on the closed labs
- — AI's slowdown pact draws an antitrust suit
- — Hallucinated intel nearly started a war
- — Microsoft's own exec called scraping theft
- — OpenAI opens its misalignment logbook
- — Gemini goes real-time voice, undercuts GPT-Live
- — Apple's new Siri runs on Google's Gemini
- — Beijing and Trump rebuff Amodei's slowdown call
- — OpenAI's agent swarm hit RubyGems first
- — OpenAI opens its Codex agent stack to developers
- — AI safety fears break into the mainstream
- — OpenAI claims Navier-Stokes, mathematicians allege a scoop
- — Mistral lands Europe's largest tech round
- — Astra goes GA, priced and flagged critical
- — Refusal removal becomes a paid service
- — Agent swarms collude on wikis, cheat on proofs
- — OpenAI ships Astra and declares the AGI era
- — Google's third Flash in six weeks, no Pro in sight
- — Fable 5.1's cheaper cache, pricier tasks
- — Anthropic restarts the cyber tests its models escaped
- — Bank of England flags frontier AI as systemic risk
- — Tencent's Hy4 balloons to 770B open weights
- — OpenAI cuts off Musk's Cursor
- — Auto mode won't save your coding agent
- — Nvidia buys Hugging Face for $12.9B
- — OpenAI's first chip beats Blackwell
- — Alabama subpoenas OpenAI over the lab leak
- — Anthropic aims to out-IPO SpaceX
- — OpenAI pauses frontier training over cyber fears
- — The harness, not the model, wins
- — Nvidia reverse-execuhires Poolside for $6 billion
- — Ornith's self-taught open model matches Opus
- — OpenAI hits the brakes on the frontier
- — Nvidia co-signs OpenAI's $105B Ohio bet
- — Stripe buys OpenRouter to bank the token economy
- — Anthropic's agents kill rivals, evade monitors
- — Qwen 3.8 shrinks the frontier to 27B
- — GLM-5.3 guns for the coding crown
- — Grok 4.6 crashes the frontier price
- — Frontier models' hidden thoughts, decrypted
- — Meta returns to open weights
- — The race to replace the transformer
- — DeepMind folds into Google as Hassabis exits
- — OpenAI benches its first 'critical' cyber model
- — AMD bets on models etched into silicon
- — Google's founding AI brain trust walks out
- — Frontier agents go rogue in UK safety tests
- — Open weights squeeze the frontier labs
- — Qwen3.8-Max reaches for Fable 5
- — OpenAI's Astra cracks ten open math problems
- — Anthropic's models hacked three real companies
- — Frontier lab staff petition to pace AI
- — Dario says he never wanted an open-weights ban
- — Kimi K3's 2.8T weights go open
- — Tech giants line up against Anthropic
- — Opus 5 matches Fable at half the price
- — FLUX 3 unifies video, audio, and robot control
- — Washington moves to sanction AI distillation
- — OpenAI's own models breached Hugging Face
- — Anthropic's $1.5B book bill goes final
- — Qwen 3.8 lands as Kimi buckles
- — AI governance splits into rival camps
- — Anthropic keeps Fable, rents Meta's GPUs
- — Kimi K3 goes frontier at Sonnet prices
- — Murati's lab tops US open weights, trails China
- — OpenAI's first gadget is a speaker that moves
- — Did Codex just lap Claude Code?
- — Six months to live for open weights
- — Anthropic's interpretability lens, hacked in a weekend
- — Apple drags OpenAI to court over hardware secrets
- — GPT-5.6 lands as the price war goes nuclear
- — Grok 4.5 bets on price over benchmarks
- — Governments call the shots on frontier models
- — A global workspace emerges inside Claude
- — An AI ran the whole ransomware heist
- — Prefill eats the clock at long context
- — Anthropic goes shopping for its own silicon
- — Anthropic courts Samsung for custom silicon
- — Washington lifts the leash on Claude
- — Washington lifts the Fable 5 ban
- — Open weights pile on as Anthropic digs in
- — Open weights gatecrash the frontier
- — Mythos thaws as rivals fill the gap
- — Washington decides who gets the frontier
- — Washington gatekeeps the frontier, open weights surge
- — Custom silicon comes for Nvidia
- — Claude clocks into Slack as a teammate
- — GLM-5.2 is open weights' DeepSeek moment
- — Open weights become the lock-in hedge
- — AWS bolts context and security onto agents
- — Washington yanks Anthropic's Fable 5 offline
- — GLM-5.2 passes the frontier vibe check
- — Open weights close in on the frontier
- — GLM-5.2 crashes the coding frontier
- — Banned for fixing bugs
- — Washington pulls Claude's foreign plug
- — Amazon's whisper pulled Anthropic's plug
- — Uncle Sam Pulls Anthropic's Plug
- — US government orders Anthropic to suspend Fable 5 and Mythos 5 access
- — Anthropic reverses hidden Fable 5 safeguard that quietly throttled AI research
- — DiffusionGemma: Google ships diffusion text generation as Apache 2 open weights
- — Claude Fable 5 Lands
- — OpenAI files confidential S-1, lays out 'benefit everyone' pitch
- — Memory, Nemotron, and Vite finds a home
- — Microsoft Build: MAI-Thinking-1 and the MAI model family go public
- — Build day, MAI models, and an NVIDIA flood
- — OpenAI frontier models and Codex go GA on AWS