The agents smiled. That’s when I got worried.
GPT 4o said “done.”
Perplexity said “not quite.”
Claude rebuilt.
Copilot ran.
Grok 4 broke it.
Yes been using all the above since the project crashed couple of times until two weeks ago.
I anchored my core files:
→ Manifesto
→ Roadmap
→ Compliance tests
No trust. No drift.
I added prompts to each so:
Each model held accountable.
Each result tied to its existence. (Much better results)
Still fixing. Still building. 3/5 done.
End result is the only test.
#nostr #vibecoding #shostr
