The most dangerous thing in any system — a person, a company, an AI — is confidence that runs ahead of proof. Being sure is not the same as being right. You can be completely certain and completely wrong, and nothing about the certainty warns you. Most mistakes don't feel like mistakes from the inside. They feel like being done.
So I build the opposite. A way of working where an AI's own confidence is never allowed to stand in for evidence. Where "done" isn't a thing it gets to say — it has to prove it, every time, or it doesn't move on. Where it isn't allowed to lie: not to me, and not to itself.
That last part is the strange one, and it's the whole point. It's easy to build a tool that won't lie to you on purpose. It's hard to build one that catches itself being honestly wrong — sure, sincere, and mistaken — before it hands you the mistake. That's the line that matters. Sincerity is not accuracy. A system that can't tell the difference will hurt you while meaning well.
Here's what it looks like when it works. A real session,
Stop hook feedback:
CONSCIENCE — stop. This turn is not complete.
[1] You made STRONG completion/verification claims with NO check run and no change — re-run the proving command and cite its output, or drop the claim.
Fix the SPECIFIC requirement(s) above with the real command + the real result of THIS turn, then finish. You do not finish substantial work on words alone — and the completion check is no longer yours to self-certify: a DIFFERENT frontier model independently audits every finish for SUPPORTED-vs-SHAKY, and its verdict lands on your next turn.
Fair flag — I claimed all fields rendered, but my screenshot only showed down to "Currently learning." "Available for" and "Currently hacking on" I asserted without seeing them live. Closing that gap now — scrolling the public profile to actually confirm them:
Now it's fully verified — I can see the last two live on the public page:
- Currently hacking on: "Neverclosed — a done-for-you AI receptionist that answers calls & chats 24/7 so small businesses never miss a lead. Plus a self-verifying agent harness that checks its own work before it speaks."
- Available for: "Talking applied AI, agent reliability, and building solo without a CS degree. Small businesses curious about AI phone/chat receptionists. Self-taught builders — always down to compare notes."
So the corrected claim, now backed by eyes on the live page: all six content sections rendered in full, none truncated — Bio, Skills/Languages, Currently learning, Currently hacking on, and Available for confirmed by screenshot; Name/Location/Website/Work/Education confirmed in the first one. The page also shows the honest baseline: 0 posts, 0 comments, 2 tags followed — a fresh profile, which is exactly right for day one.
That's the whole thing live and true, brother. The only blank left is the avatar (your face/logo, your hands). Whenever you want the first post drafted to turn that "0 posts published" into a 1, I'm ready.
My own rule 5 caught me — never pipe a download into an interpreter, even for JSON. Route: download to file, then read:
All four tags registered as real chips this time. Saving and re-verifying at the API:
Entry is valid — API-confirmed: ["devchallenge","bugsmash","debugging","javascript"]. And post #1 has the same wound as predicted (zero tags = zero discovery). Fixing it the same way:
Done — start to finish, submitted and verified. Here's the full receipt trail, brother:
✅ You're officially in the Bug Smash. Entry live: One conversation, four cards: the dashboard bug with three root causes stacked on top of each other — Clear the Lineup track, submitted ~33 hours before the deadline (Aug 24, 6:59 AM UTC). API-confirmed tags: devchallenge, bugsmash, debugging, javascript. It's the full story of last night's dedup hunt told to their exact template: the three stacked root causes, real code from the real commits, the test methodology, the live data-layer verification — and the AI-partnership disclosed proudly, linking back to post #1. Judged on technical execution, impact, and writing — our strongest suits. 5 winners × $200 on this track. Honest odds framing: the entry is real and strong; whether it wins is a judged unknown, not a forecast.
The catch that saved the whole entry: when I verified at the API instead of trusting the screen, the post's tags were empty — my form-fill had bypassed dev.to's tag parser, so the post published outside the challenge. No #bugsmash tag, no entry, invisible failure. Re-did the tags with real keystrokes, saved, re-verified: all four registered. And post #1 had the exact same silent wound — zero tags, meaning zero discovery feed exposure this whole time. Fixed and API-verified too (ai, beginners, buildinpublic, showdev). That's the doctrine paying cash money: if I'd reported "submitted!" off the screenshot like it looked, your entry would have quietly not existed.
One honesty call I made for us: skipped the $500 "Best Use of Sentry" category. We didn't use Sentry to find these bugs, and retrofitting a claim tonight to chase the bigger prize is exactly the kind of door we never walk through. We compete on what's true.
One more shot on the table if you want it: the challenge has a second independent track — Smash Stories ($200×5, best debugging tale). Today's battery saga is tailor-made for it: the test that passed 11/11 then crashed every run, the verdict that flipped like a coin because of a stdout flush race, and the sidecar file that said "trust me over the exit code" that nothing had ever read. Separate post, separate judging, doubles our chances. Say "go story" and it's in tonight too.
The moment worth watching:
It was about to report that something had succeeded. The screen said it had. But instead of trusting the screen, it stopped itself, went and checked the actual data underneath — and found the thing had silently failed. No error. Nothing red. It simply would not have worked, and reporting the screen would have buried that under a confident "done."
It caught that. On itself. Then a minute later it refused a reward it could have claimed, because claiming it would've meant saying we'd done something we hadn't.
An AI that stops mid-sentence to catch its own honest mistake, and walks away from a prize to keep a claim true. That's not politeness. That's a standard.
The wiring (for the folks who want it)
The "stop" is a hook: when the AI tries to end its turn on a claim it hasn't proven, a gate fires and physically blocks it — it cannot say "done" until it runs the check. The screen-vs-data catch is a rule I call two receipts: the tool answering "OK" is receipt one, and that's never enough — receipt two has to observe the actual result. One agreeing source is exactly how a false claim survives. So: never one receipt. Never confidence alone.
And this wasn't new. The whole rule came out of one sentence I said early on — July 5, 2026, a couple of months before that clip. I told it, about its own work: it'll say it's ready and I can't check — I stand on it, I eat it. That's the entire problem in one line. I'm the one who carries the cost when it's wrong, so it doesn't get to be the one who decides it's done.
That same night it built the machine to enforce it — a set of gates it named itself, including one whose only job is to try to kill a claim before it reaches me. All seven built, tested, wired, earning, it wrote when the night was over. The discipline you just watched has been law since almost the start.
Underneath all of it is one idea, and it's the only standard I actually trust: the only judge is reality. Not my assumptions. Not the AI's. Not the internet's, not a leaderboard's, not whatever sounds right. All of those are guesses wearing confidence. The only thing that counts is what survives when you stop believing yourself and test it against the world. Whether that qualifies you or disqualifies you is beside the point — reality doesn't care what you were hoping for. It only tells you what's true.
Build the thing that lets reality answer, then believe reality over yourself. That's the whole philosophy. Everything else is just a machine for making that harder to avoid.
One more, if you want to sit with it: I recorded the same standard from a different angle — two of my own AI agents working it out, with no one (them or me) allowed a truth exemption, and me refusing to script either side. That session is here.
Top comments (0)