Risk register
Why we might die
We ran WhittleOS's own one-liner through WhittleOS's own kill-gate — the same analysis you'd pay for. It did not flatter us. This page is the full verdict, every failure reason, and what we're doing about each — updated as the status changes.
“Founder-tooling for founders: a notoriously crowded, low-trust segment… this stacks multiple hard problems at once: credibility, distribution, retention, and monetization.” — our adversarial mode, on us.
The same one-liner, elsewhere
A popular AI idea validator scored it 72/100 — “PROMISING” with a rocket emoji and “you're among the top ideas we've seen” — and added +3 points for “live demand signals” that included a search-interest score of 0 out of 100.
“If your own product said KILL, why are you building?”
Fair question — it's the first thing our gate teaches: a KILL is not “never”. It means don't build further until these specific holes are closed, and every verdict ships with the ladder of what would need to be true. This page is us climbing that ladder in public. If we can't close these rows, you shouldn't pay us — and we'll have proven the product works.
The register — 10 ways this dies
1 open · 7 testing · 2 closed
01Founder tooling for founders is a crowded, low-trust market
TestingGate said: “The buyer is already surrounded by free content, AI chats and public frameworks. Nothing here proves your judgment is more trustworthy than the free alternatives.”
Our move: The bet: a kill-screen is not advice — it's a decision instrument with a visible rubric. We publish a paying-customer counter; if it doesn't move, this risk wins.
02No durable distribution channel
OpenGate said: “A skeptical, online-native audience that is easy to reach and hard to convert — and the founder won't do sales calls.”
Our move: Build in public in the exact communities we serve. Every public kill — including this page — is the content engine. Channel performance gets measured, not assumed.
03Demand is not proven beyond our own pain
TestingGate said: “Without written evidence that founders already try to pay for this and are dissatisfied, this is a clever framing, not validated pain.”
Our move: Following our own gate's ladder: 20 written statements of the pain, then 5–10 prepaid commitments at the real price. Numbers will be published here.
04Subscription fights episodic usage
TestingGate said: “Evaluate a few ideas, pick one, churn. The discovery feature patches retention instead of proving it.”
Our move: Our own adversarial mode suggested the fix: evaluation packs (credits) tested against subscription. Whichever wins, retention numbers go public.
05Scope too broad for one person
ClosedGate said: “Kill ideas, rank survivors, and discover candidates — three jobs with different failure modes, stacked on a nights-and-weekends budget.”
Our move: Closed by sequencing: the kill-screen is the wedge. Ranking and discovery ship behind it, not beside it.
06One wrong KILL destroys trust
TestingGate said: “If the system kills a user's favorite idea and is wrong, credibility dies instantly; if it's too cautious, it becomes generic.”
Our move: An outcome journal: we track what happens to killed and approved ideas and publish the record. Every verdict ships with its reasons, so you can attack the specifics, not the vibe.
07AI costs eat the margin at a low price point
ClosedGate said: “Repeated evaluations on inference can erase margin — and if the product is mostly deterministic, users may not see subscription value.”
Our move: Closed by architecture: scores are computed by a fixed rubric in code; the model only gathers and grounds evidence. A full check costs ~$0.07 of inference today — published.
08Built from self-recognition, not external demand
TestingGate said: “The founder is the target user — familiarity, but also the classic trap of building introspection tools other founders won't pay for.”
Our move: External users only count if they're strangers: we report usage from outside our own circle.
09Payments friction (no direct Stripe; merchant-of-record instead)
TestingGate said: “Global self-serve buyers expect frictionless checkout; payment friction suppresses conversion at low ticket sizes.”
Our move: Solved via Polar as Merchant of Record — checkout, tax and invoices run through it. The row stays until real buyers have completed real (non-test) purchases at volume. Yes, we publish even this one.
10A kill-biased product invites disputes and support load
TestingGate said: “Founders will challenge verdicts, demand explanations, and ask for refunds — against a founder with low support tolerance.”
Our move: Every verdict is explainable line by line: which filter fired and why. Disputes get answered with the rubric, not customer-support theater.
“Maybe your gate is just cruel?” — we checked
We disguised 10 real products that solo founders later took to sustainable revenue and 10 documented failures — each described as it looked pre-launch, names changed so the model can't recognize them — and ran every one through this same gate twice.
15%
of future-winner runs were killed — 3 of 20: a community-data directory both runs (on founder-fit, KF-08), plus the deliberate hard case (an AI wrapper that won on timing) once. The rest: “pilot first” at 58–78/100, the honest verdict for an unproven idea.
90%
of documented-failure runs were killed — and the gate's reasons matched the autopsies: no willingness to pay, ops-heavy marketplaces, free products with no model.
2/2
a probe shaped exactly like an “AI idea validator” — paste an idea, get a score and advice — was killed both runs: remove the language model and nothing remains.
The gate isn't cruel. It's specific — it cleared 85% of the winners it had every excuse to kill (most as “pilot first”), and it killed us for reasons we have to answer. Full run data ships with the repo.
Receipts
This isn't staged. The run is a repeatable script in our repository: same input the external validator got, three of our own analysis modes, ~48 seconds, $0.07 of inference (re-run 2026-06-24 on the current gate). The verdict, the score and every reason on this page come from that output, unedited. We re-run it as the product evolves — the day our own gate upgrades us from KILL, that becomes the headline here, with the diff.
Now imagine what it says about your idea.
The same gate, the same honesty, bound to your founder profile. No calls, no coaching, no AI flattery.