I think "It is always more complicated than it seems" is a good rule of thumb for life.

This is your periodic reminder that @Google has the worst privacy policy of any individually-available plan. Want to refer to old chats? Then you have opted into 3-year retention of your data and training on your interactions.

So glad Anthropic is valiantly protecting us against the erosion of it's business model.

This matters legally too. The law asks what a "reasonable person" would do and makes the application certain laws dependent on internal mental states, because we recognize each human as a moral actor. If we accept the argument that AI systems are moral actors, then what legally follows is the assignment of responsibility for harms to the AI system itself, not to those who created it or directed it to do the harmful thing.

Any lawyers using cowork - pay attention to whether your practice guidelines will allow this. They probably should - it is not meaningfully different than using Sharepoint on Microsoft 365 - but client policies around AI tend to be much more restrictive.

Word to the wise: @Intuit QuickBooks online does not use an idempotency key when submitting direct deposit requests. Evidently their developers all skipped the day when they learned about how to safely handle transactions.

That plus the worldwide watermarking. @anthropic doesn't care. They are enshittifying at AI speed, propped up only by the fact that their underlying tech is genuinely good. But they are doing everything they can to make their models as frustrating and patronizing as possible.

Or doing legal analysis of different AI system risks: You mentioned prompt injection! NO FABLE FOR YOU!

Doing code review on a process orchestration app - oops! You mentioned "sandbox" and "SIGKILL"! You must be trying to attack someone!

Claude code with Fable is actively frustrating to use. Together with the other shenanigans that @anthropic keeps pulling, it seems not worth it anymore.

If you are trying to do anything even remotely like trying to make a system safe, Fable's "safeguards" actively sabotage you.

So I genuinely like Fable better than Opus. But will I reach for it when it is not part of the subscription plan? Probably not very often. GPT-5.5/Opus-4.8 is working, and Fable's restrictions are annoying and require workarounds. Plus, people underestimate Grok. It's coming up.

Shorter @AnthropicAI: "Now that we have made significant advances in building a market, it is now time to enact regulatory barriers that will prevent others from following us and reducing our market share."

Fable is as capable as Codex 5.5 (or more). However, it triggers ridiculous limits verifying basic bug fixes. Evidently fixing bugs now requires restricted cybersecurity access.

I packaged up my agentic coding assistance system for anyone to use: github.com/VanL/agent-guidan…. What this is: I have been doing a lot of AI-assisted development recently, and I have found that I frequently get better results than some others. I think that part of it is the structured system I use for managing agent context and guiding agents away from common pitfalls. Agent-guidance makes agents use a more highly-structured specification-plan-implement cycle than the basic planning mode will get you. In combination with github.com/VanL/agent-mcp, plans and implementations get a competitive review from a different agent, helping eliminate blind spots. If you are a user of @garrytan's gstack, @every's compound engineering, and @obrajesse's superpowers, agent-guidance will work with those skill stacks to make implementation easier and higher-quality.

Pity the four of us who used em-dashes naturally. Now everyone thinks we're ChatGPTing everything.