It ensures that the agent does a good job
Image Text:
Github overview of a pull request for the file AGENTS.md.
Comments regarding line "# Repository Guidelines" in the file:
zanieb: "What's the point of this title? Seems like wasted tokens?"
zanieb: "(certainly inconsequential in the big scheme, but I don't see it adding value)"
charliemarsh: "It ensures that the agent does a good job."
7 replies
I’m proud of the fact that I don’t understand this.
You and me both.
It wastes tokens which is good. Every penny my boss spends on AI brings the company closer to bankruptcy.
My company has just implemented a token leaderboard. I find that asking particularly insane questions burns more tokens, since it won't have previous results to fall back on.
Kind sir, could you share those questions? I'm in great need!
If the agent can't tell a contextual comment from an actionable code line, then it is indeed a wasted token. Not because of the header, but because of the "AI" being a useless POS.
Which, to be fair, is the case in pretty much all scenarios.
Wasn't a comment, it was a header in a markdown prompt for ai agents to slurp into their context.