Validation
Preamble
Validation is where Founder Loop turns evidence-backed wedges into pass, loop, or kill. It has to improve founder judgment without letting a polished verdict become a substitute for proof.
A Verdict Can Become Theater
Validation is the first place where a weak idea can be made to sound investable. The input arrives with evidence, categories, claims, clusters, uncertainty, and recommended next action. A model can arrange that material into a confident memo. It can name the ICP, describe the pain, identify the wedge, and produce a pass verdict that feels like momentum. The verdict matters when it changes the next contact with reality.
Founder Loop is also where founder judgment can be quietly displaced. If the model decides what passes, the founder becomes an executor of a verdict. If the founder decides everything from instinct, the Factory becomes decoration around taste.
The useful middle is harder. The loop should force the founder to see what is being accepted, what is being postponed, what would disconfirm the idea, and which consequence the next test is allowed to create.
Pass, Loop, Kill
Founder Loop produces one of three outcomes. Pass means the packet has enough support for a same-evening market probe. Loop means the idea has promise but needs missing proof, sharper ICP, disconfirmation, or better source coverage. Kill means the evidence has failed the bar for more attention.
The system writes back why. A pass creates validation artifacts. A loop creates source-plan work. A kill preserves the reason so the same false positive does not return under a new name.
The Judge Needs Calibration
LLM judgment can help pressure the packet, but it cannot become silent authority. The validation harness needs a founder packet contract, unsupported-claim checks, source-fit checks, domain-chip deltas, disconfirmation state, and judge/human calibration.
The model may propose a verdict. The system has to show what the verdict rests on and where the evidence is missing. The useful loop produces calibrated refusal, repair, or market contact. The public claim stays narrow until the loop proves it can make a founder slower in the right place and faster after the evidence is strong enough.
The First Real Proof
Validation lands when a passed wedge creates a structured handoff that can be tested that evening without translation work. The proof is a market probe, a kill decision, or a loop that sends Discovery back for specific missing evidence. A verdict earns its voice when it changes what happens next.