Older news, kept for readers and for the record. Prices, versions and dates in these pieces may have been superseded: the reference trackers hold the current values.
- Claude Opus 5.5 drops to $4 and $20, and breaks four Opus 5 habits
- Grok 4.7 keeps $2 and $6, and its gains over 4.6 are xhigh versus high
- Qwen-Image-2.1 brings the weights back, but not the Apache licence
- Does Bonsai 2 27B really keep 98% of Qwen3.8 in 5.95 GB?
- Jev 1.13 bills $42 a billion tokens, and it can’t count
- Qwen3.8-Omni-Flash bills audio at $0.15 and ships no weights
- AWS can’t restore me-south-1, six months after the drone strikes
- Gemini 3.8 Live Extended Thinking rejects any tool that blocks
- Atria Dawn Preview is 744B under MIT, and the BF16 weighs 1.5 TB
- OpenAI’s Agents API has no fee, no ZDR and a one hour sandbox clock
- Fugu Max costs $2 and $6 while Fugu Ultra v2 runs without Fable 5
- DeepSeek V4.1-Flash arrived, and the V4 Pro retirement lasted a day
- SWE-2 trails Fable 5.1 by one point, and by 28 on Terminal-Bench 4
- Does Meta Muse do enough to earn your inbox and a card on file?
- Broadcom pulled VDDK 8.0 and 9.0, and the 404 is the only notice
- OpenAI’s 3.1 agent-workdays per human day is not a 3.1x gain
- Mullvad’s DNS servers go dark on 2 November, and Quad9 blocks no ads
- GPT-6 Astra lists at $10 and $50, 2.5x what GPT-5.6 Sol costs
- Android 17 hides the SNI, not your DNS or destination
- Anthropic moves retention into your own cloud, for 30 days
- Gemini 3.8 Flash keeps the price and the 1 January cliff
- ChatGPT Mil and Grok reached IL5 on GenAI.mil
- The EU now calls ChatGPT a very large search engine
- Claude Fable 5.1 breaks forced tool use, cuts cache 75%
- Anthropic’s Model Hardware Standard is gated, and sits under MCP
- WebMCP in ChatGPT needs GPT-5.6 Sol or Terra
- Cohere Parse 5 is $1.50 per 1,000 pages, on three of five dimensions
- A judge vacated one Pentagon label on Anthropic, and the other stands
- Qwen3.8-Flash-Next runs 6B active and weighs 173 GiB
- Tencent’s 770B Hy4 tops one benchmark row in 46
- Trainium4 takes Nvidia memory, which matters more than 2M GPUs
- Microduck is $399, with open software and closed hardware
- Nvidia pauses its $36B AI Compute Partnership
- AWS bought DuckLabs, and DuckDB stays MIT
- Gemini Omni 1.1 Flash stitches its 40 seconds and upscales its 4K
- MTurk closes on 30 September, and approvals run a month later
- Anthropic’s Nscale deal buys 460 MW for late 2027
- Gemini 3.5 Transcribe costs half a cent a minute, ranks fifth
- Nobody has confirmed the $12.9B Hugging Face report
- o3 left ChatGPT, and the API keeps it to 11 December
- Waymo picked Munich for its first EU city, late 2027
- GLM-5.3-Flash was Ox Alpha, and the price doubles
- Jalapeño measures 1.9x per watt against a GB200
- The M6 Mac mini is $899, capped at 16GB
- A2A v1.0 breaks v0.3.0, and the AAIF news buried it
- Cloudflare Bot Preference Sync rewrites your robots.txt
- Wan 3.0 does thirty seconds a clip, closed and $6
- GitHub’s autoscaler watched the wrong service for 7h47m
- The MCP roadmap has five priorities and no dates
- A free 1M context coding model appeared with no lab behind it
- Bun 1.4 is the first stable Rust build, and five things break
- Nevada cleared 8,000 robotaxis, and Tesla’s 5,000 is a ceiling
- Nvidia paid $6B for a non-exclusive Poolside licence
- Prime Air promises 500 cities from 11 drone sites
- Google’s $12.2B Marvell warrant needs $120B of orders
- OpenAI’s Private Safety Processing keeps ZDR alive
- Ramp Router is free until January, with a year of retention
- Cerebras stacks three wafers for 750 PFLOPS
- Cursor opened its git forge the day GitHub broke
- Google’s $10M Spirit Airlines Data Buy Slips to September 9
- Claude Code weekly limits drop a third after August 19
- Groq raised $350M and now sells Nvidia capacity
- Stripe is reportedly buying OpenRouter for over $7B
- Every Claude output is watermarked, with no opt-out
- Nvidia’s $105B OpenAI exposure is a lease guarantee
- Nvidia’s 13F shows $63B in equity and not one new buy
- Quantinuum is putting 98 qubits inside an Oracle data centre
- Microsoft merged the Copilot apps, and three features retire
- Palmyra X6 runs on GLM-5.2, priced $2 in and $8 out
- Qwen3.8-27B ships Apache 2.0 with vision, the 2.4T doesn’t
- GLM-5.3: open weights in two weeks, breaking API change now
- GPT-5.6 Sol Ultrafast hits 750 tokens a second, with no price
- Uber and Pony.ai promised 2,000 robotaxis and named one city
- Gemini 3.7 Flash is half price until 31 December
- Grok 4.6 keeps $2/$6 and adds a 200K pricing cliff
- Twitch’s AI training toggle is on, and it covers your chat
- DeepSeek V4 Pro 0813 is live, at 3x the Flash price
- Nemotron 3.5 Lightning runs 30B on 3B active parameters
- ChatGPT desktop for Linux: the .deb adds an OpenAI apt repo
- Anthropic’s Theseus: a data center deal with no numbers
- Intel raised $20B and 14A still has no named customer
- NVIDIA’s $500B compute financing and the 25% backstop
- Claude Code auto mode is the default from August 14
- Meta Muse Glimmer: the 30B agent needs 24GB of VRAM
- Source Foundry: $400M and a $5B bet against ASML
- ByteDance’s 10T model is pretraining, not a product
- Genesis-Science-1: DOE Opened a Portal, Not a Model
- TfL licensed 15 Wayve cars in London, driver still inside
- Terafab: what Tesla and SpaceX’s $16.8B first phase buys
- AMD buys Taalas: the HC1 etches Llama 3.1 into silicon
- Entanglement Ran 62 km of Aerial Fiber at 92.8% Uptime
- Anthropic’s custom silicon team: no chip, no timeline
- GPT-5.6 Sol in ChatGPT is not the Sol in your API calls
- IonQ Evergreen-05: DARPA funds 25 clocks, not 125
- Jeff Dean leaves Google: what changes for Gemini devs
- Meta Muse Code: Muse Spark 1.2 and the 21x cheaper tier
- Unitree priced at $9B, and 73.6% of humanoids go to labs
- Google Assistant shuts down September 4: what you lose
- The grok-voice-latest alias just got 60 percent dearer
- Shieldstral 1.0: the 3B guard model that ties a 20B
- Bending Spoons is buying Airtable for $1.285B in cash
- MiniMax H3 open weights: not licensed in the EU, UK or US
- Qwen 3.8 Max: is it open source? Not yet
- MacBook Air M5 shortage: $1,299 and a six week wait
- A judge shut Waymo’s Santa Monica depot from 11pm to 6am
- What Article 50 of the EU AI Act makes you label
- Inkling-Small out-codes the 975B flagship at 276B
- OpenAI Astra: what the ten Lean proofs actually prove
- Qualcomm’s double-digit chip price rise starts September 1
- DeepSeek V4-Flash-0731: same API, new model
- Google pulled Nano Banana 2 from Earth after one day
- Zoox wins Part 555: paid robotaxi, no steering wheel
- Amazon’s $1.8M Claude bill: 860% over, 5 months unseen
- Gemini Robotics 2: what Google’s own benchmarks say
- Seagate 50TB HAMR drives: 2028, not 2027
- Cloudflare origin PQC: ML-DSA-44, OpenSSL 3.5.0
- FCC robot ban: the 4.4 lb rule that catches vacuums
- GPT-5.6 price cut: Luna drops 80%, Sol stays at $5
- MCP 2026-07-28 removes sessions, and no policy covers that
- Nvidia SSI deal: $5B reported, 10x compute on Vera Rubin
- OpenAI’s free GPT-5.6 for researchers: what you get
- The Pixel 11 price rise is 4.3x on RAM, not sixfold
- Laguna S 2.1: 118B open weights on one DGX Spark
- Tesla Optimus V3: production starts, reveal still missing
- IBM buys HRL: 18 spin qubits and a 300mm fab
- iOS 26.6 is out: the Spotlight index it quietly builds
- SK and NVIDIA signed letters of intent, not a $500B deal
- Cursor agent swarm: SQLite for $1,339 and the 15x catch
- Google’s €890M DMA fine: what changes in 60 days
- The Kimi K3 weights are 1.4 TB, and that is the catch
- AMD Helios vs Vera Rubin NVL72: the MI455X numbers
- Etched raises $300M at $10.3B: where did Sohu go?
- Open Weights letter: 25 signers, and what it changes
- Claude Opus 5 keeps the $5/$25 price, with thinking on
- FLUX 3 is here: video, audio, open weights later
- OpenAI Presence: the enterprise AI agent platform
- PsiQuantum’s $125M DARPA deal buys verification, not a machine
- NVIDIA Spectrum-6: what 102.4 Tb/s Ethernet changes
- A judge approved Anthropic’s $1.5B author settlement
- ChatGPT ads are live: OpenAI’s self-serve Ads Manager
- Claude Cowork Record a Skill: teach agents by demo
- Gemini 3.6 Flash is cheaper than 3.5 and scores higher
- BMS committed to eight Vera Rubin systems, not a cluster
- Qwen-Image-3.0 is here: no weights, no benchmarks
- Claude Code drops Node for Bun 1.4.0 in Rust
- Did Claude Fable 5 disprove the Jacobian Conjecture?
- Mistral’s first robotics model steers on one camera
- Fireworks raised $1.5B, and a raise ships no code
- Humanoid HMND 01: what the robot unicorn really ships
- Classic McEliece is the first ISO post-quantum standard
- Claude Fable 5 goes permanent July 20: the fine print
- Gemini 3.5 Pro missed July 17: launch vs leak
- NVIDIA Jetson T3000 and T2000: Thor for mainstream robots
- Apple Intelligence China: approved, not shipped
- NVIDIA Ising Decoder ColorCode 1: what 347x really means
- iOS 27 ships 14 September with Siri AI still in beta
- Kimi K3 is live, and most of its specs are still leaks
- Inkling is a 975B open model that loses benchmarks on purpose
- Apple sued OpenAI over trade secrets on 10 July
- New York just paused data centres above 50MW
- Nvidia cut half its authorised AI chip buyers in Asia
- Meta’s Iris chip enters production in September
- Meta ships Muse Spark 1.1 at $1.25 per million tokens
- Ship ML-DSA now, says Cloudflare, and here is what it costs
- Is Tencent’s 295B Hy3 really free?
- OpenAI GPT-Realtime-2.1: what changed for voice agents
- OpenAI’s first chip, Jalapeño, is built for inference