So, I find it hard to agree with opinions claiming that these budget cuts are bad news for AI labs. The reason why budget caps are necessary is that coding agents have become undeniably
@simonw
-
The evolution of AI costs: from $20 to $1000 per employee
By
–
Last year, it was laughable for a company to spend $1,000/employee/month on AI – the $20 plans were enough for everything anyone did with it This year, setting a cap of $1,000/month is seen as the financially prudent thing to do, and the one
-
Throwaway debug code used by model to track down the bug
By
–
It's not code for shipping, it's throwaway debug code the model used to help track down the bug
-
Claude Fable 5: proactive, launches Python CORS servers and pyobjc-Quartz for capture
By
–
After two days with Claude Fable 5, the best way I can describe it is 'relentlessly proactive' – here is an example where I inserted a screenshot of a bug and it started custom Python CORS servers and used pyobjc-framework-Quartz to capture
-
JSON Extras API Explorer with Fable 5 and GPT-5.5
By
–
Since the JSON extras API is a bit difficult to explain without an example, I also had Fable 5 and GPT-5.5 collaborate on this custom API exploration tool to test the new feature https://tools.simonwillison.net/datasette-extras-explorer#url=httpslatest.datasette.iofixturesfacetable.json_size1_facetstate&extras=countcount_sqlexpandable_columnsfacet_results
… -
OpenAI considers pricing war after Anthropic’s impressive models
By
–
Alternative interpretation: OpenAI are finding Anthropic's latest models to be surprisingly good and are considering a pricing war in response
-
Visible cyber/bio refusals vs invisible frontier LLM refusals
By
–
Those were the visible refusals for cyber/bio/etc – it was just the refusals for "frontier LLM development" that were deliberately made invisible
-
Looking Back at ChatGPT Code Interpreter, the First Coding Agent
By
–
It's funny to look back at this Twitter conversation about the then-new ChatGPT Code Interpreter from three years ago – in hindsight, it was our first glimpse of a coding agent, before we knew what a coding agent was.
-
Fable 5: Making guardrails visible for LLMs, canceling a scandalous decision
By
–
Do not miss the exact text however: « We modify the guardrails of Fable 5 for the development of cutting-edge LLMs to make them visible » – making them visible means they cancel the truly scandalous (dare I say 'misaligned') decision to have
-
Honest refusal is a win over lying
By
–
I read and entirely understood it, and I quoted the exact text that said "to make them visible" I see this as a win. I don't want models that lie to me – refusing a task but being honest about that refusal is a big improvement