This is explained in 3.2 How to design good tools?

    This saves the LLM from having to do multiple low level clicking and typing and keeps it on track. Help the poor model out, will ya!?

I'm not sure where this quote is from - it doesn't seem to appear in the linked article.

ahh, sorry, different article :(

putting the review into git notes might have worked better. It's not attached to tje lines directly, but the commit and it can stay as part of the repo

Not at all. Good documentation for humans are working well for models too, but they need so much more details and context to be reliable than humans that it needs a different style of description.

This needs to contain things that you would never write for humans. They also do stupid things which need to be adjusted by these descriptions.

I was thinking about this too, but the problem is that different models need to be prompted differently for better performance.

Claude is the best model for tool calling, you might need to prompt less reliable models differently. Prompt engineering is really hard, a single context for all models will never be the best IMO.

This is why Claude Code is so much better than any other agentic coding tool, because they know the model very well and there is an insane amount of prompt engineering went into it.

I tried GPT-5 with OpenCode thinking that it will be just as good, but it was terrible.

Model-specific prompt engineering makes a huge difference!

Claude can already do this: https://docs.anthropic.com/en/docs/claude-code/github-action...

Yes, I love github actions for claude code! And use it myself for code reviews.

The original title is "Objects should shut the fuck up". I don't like unnecessary cursing either, but it is emphasizing his frustration in this case and cursing objects, not people.

Renaming the title is just losing information for no reason.

Not sure I would consider a user copy-pasting an URL being a bot.

Should curl be considered a bot too? What's the difference?

> Should curl be considered a bot too? What's the difference?

Perplexity definitely does:

    $ curl -sI https://www.perplexity.ai | head -1
    HTTP/2 403

It gets blocked in my setup because bots use this as a workaround.

This implementation is unnecessary complicated. For the step update, you can use Out Of Band update: https://htmx.org/attributes/hx-swap-oob/ which works in a way that you can send multiple HTML fragments which can be anywhere on the page and HTML swaps them out. Good for notifications, step update, breadcrumb update, menu highlight, etc...

I usually solve the second problem by simply saving the state of the individual input fields, you only need a user session. Depending your use-case, you might need to be transactional, but you can still do this saving everything as "partial" and close the "transaction" (whatever it might mean in the given context) at the last step. Much-much simpler than sending form data over and over.

> For the step update, you can use Out Of Band update

I did mention using OOB, but I preferred swapping the entire Stepper because the logic on the backend was just a little bit cleaner, and the Stepper didn't include anything else anyways.

> I usually solve the second problem by simply saving the state of the individual input fields, you only need a user session.

I believe this is exactly what I did in the article, no?