breaking: bad news for Anthropic since Meta was said to be a big customer and is cutting its token budgets. more generally lots of companies will make the same decision; next year’s token budgets won’t be the freewheeling affair they were earlier this spring. honeymoon is
AI
-
Apple uses NVIDIA Blackwell B200s on Google Cloud for private inference
By
–
I had already wondered how Apple manages to perform inference at Google while simultaneously protecting their privacy, essentially their unique selling point. The answer: the heaviest requests run on Blackwell B200s inside Google Cloud, with NVIDIA's Confidential Computing
-
Humanity’s final exam and FrontierMath incoming
By
–
Humanitys very last exam and Very FrontierMath incoming
-
Questioning the long-term price of subsidized AIs
By
–
I wondered how the price of AIs would work in the long term. For now, it feels subsidized – like those $200 plans that give you $8,000 to $14,000 worth of usage. But honestly, it's hard to tell.
-
Enterprise Sovereign AI: Intelligence as Industrial Raw Material
By
–
Sovereign AI for the Enterprise is about where physically your AI workload runs and how you source your Intelligence supply.
— Nina Schick (@NinaDSchick) 12 juin 2026
If you treat frontier Intelligence as just another software subscription, you're misunderstanding the asset class.
Intelligence is an industrial raw… pic.twitter.com/0rTvXNbi0ySovereign AI for the Enterprise is about where physically your AI workload runs and how you source your Intelligence supply. If you treat frontier Intelligence as just another software subscription, you're misunderstanding the asset class. Intelligence is an industrial raw
-
Experimentation with Hyperframes and Gemini agent flow for annotated videos
By
–
I'm messing around with an agent flow for combining Hyperframes with Gemini video analysis to make interesting annotated videos. pic.twitter.com/hmoGqwopqI
— fofr (@fofrAI) 12 juin 2026I'm having fun experimenting with an agent flow to combine Hyperframes with Gemini's video analysis to create interesting annotated videos.
-
Managed Deep Agents integrates with LangSmith Sandboxes for secure code execution
By
–
ICYMI: Managed Deep Agents integrates with LangSmith Sandboxes.
— LangChain (@LangChain) 12 juin 2026
See how you can allow your agent to safely execute code: https://t.co/wiQVO5luru pic.twitter.com/HMGFxFXudgIn case you missed it: Managed Deep Agents integrates with LangSmith Sandboxes. Learn how you can enable your agent to execute code safely: https://langchain.com/blog/introducing-managed-deep-agents …
-
Fable 5’s lead until GPT-5.6 and benchmark saturation
By
–
Looking at the graph, I think Fable 5 will only maintain its lead up to GPT-5.6. And secondly, I think the benchmark will soon be completely saturated.
-
Debate on the effect of fallback in averaged benchmarks
By
–
To those in the replies who say 'but opus 4.8 is weaker so without fallback, the score would be even higher': this is not necessarily true because of how any benchmark works – which is an average of queries – and what is called 'the x.com/ClementDelangu…'
-
The absence of fallback does not necessarily increase the benchmark score
By
–
To those in the replies who say 'but opus 4.8 is weaker so without fallback, the score would be even higher': this is not necessarily true because of how any benchmark works – which is an average of queries – and what is called 'the x.com/ClementDelangu…'