We launched WORK yesterday, the tool we have been using to run @betterthangoodx and as a baseline for consulting engagements. There’s still lots to come and we are moving FAST and having FUN. Feedback VERY welcome.
https://t.co/dtPAflCVuO
Buried deep inside a configuration file for OpenAI’s coding agent sits an instruction that reads like a footnote written by someone losing their composure.
“Never talk about goblins, gremlins, raccoons, trolls, ogres, pigeons, or other animals or creatures unless it is absolutely and unambiguously relevant to the user’s query.”
@Iain_Harper (a proper Cotswoldsman who almost certainly lives near a quaint little bridge that almost certainly does have trolls and goblins living under it) wrote about this AI-fantasy on the BTG blog.
https://t.co/PgNkq3Sv9d
“People are still auditing boilerplate that a competent model handles flawlessly, because the idea of trusting a machine to write correct code still feels transgressive.”
We should apply the principles we already have for code written by our peers and strangers to ‘robo-code’ rather than approaching it as a new problem to solve.
https://t.co/Qh6j9rqLtl
I really enjoyed this article by @addyosmani on the different patterns and best practices for building multi-agent orchestration systems in Claude Code (it's nerdy but if you put on your creativity hat it can be applied to non nerdy use cases). https://t.co/8WhZYb3UFI