Surprising finding: YC funnel open for misuse
Score 8/10 → skill asks if you'd apply to YC → on yes,
opens YC apply?ref=gstack
Rubric is open source. Same model grades both sides.
Claude optimizes answers to SKILL.md acceptance criteria, produces brilliant YC application?
Sharpest failure mode: context poisoning across the
agent chain.
Office-hours trusts my prompts. CEO-review trusts
office-hours. Eng-review trusts CEO-review. Implementation
executes the plan.
But research = what you do when you DON'T know.
Nobody challenges the premise.
I'm tired of LLM skill slop.
The skills look polished. They sound confident. They quietly mislead you. They never crash — they just produce wrong answers in beautiful prose.
So I added regression tests to mine.
@garrytan if you hear me, please consider adding regressions to GStack skills including office-hours. Skills without tests make me cry.
Full writeup, skill source links in comments.
Added blind regression tests to my plan-cmo-review skill. They failed. We iterated. They passed.
Ran them 9 more times — cracks reappeared. Iterated again. They pass.
Methodology is still rough. Still better than "looks good to me."
@KassyDillon Exaggerated. The disruption lasted 2-3 minutes at most and the disruptors got booed at. I was there at the commencement ceremony. Mark my words, Boston University has been one of the least-brainwashed US universities. Congratulations to BU grads!
@LensVeritatis Over-exaggerated. The disruption lasted 2-3 minutes at most and the disruptors got booed at. I was there at the commencement ceremony. Mark my words, Boston University has been one of the least-brainwashed US universities. Congratulations to BU grads!
@MarinaMedvin The disruption lasted 2-3 minutes at most and the disruptors got booed at. I was there at the commencement ceremony. Mark my words, Boston University has been one of the least-brainwashed US universities. Congratulations to BU grads!
@elonmusk@elonmusk make Twitter a non-profit (no pressure 🙂).
@elonmusk even better, buy an education institution. Universities (unlike K-12) are money-makers - yet, you get to influence the future of free speech. Are you in for $$ or making US stronger?
@altos_labs Congratulations, but as a prospective client I do hope to see factual evidence/proof of your technology advancement, as opposed to academic honours.