Adversarial decision review for UK indie operators
Most "validate your idea" services tell you you're right. The default ChatGPT pass tells you you're right too. A public test ran three made-up startup ideas through it; ChatGPT said "build it" to all three.
This service does the opposite. It tries to kill your bet with cited evidence, before you spend the build time. If it can't kill it, the kill criteria you walk away with are the ones worth tracking.
What you get
A structured artefact: not a chat transcript, not a bullet list. A document, on its own page, with the audit trail visible.
- Pre-survivor checks (concern 0). Publisher-level free incumbent (with URL); three closest extant competitors by URL and pricing; persona-isolation compatibility; open-bet-conflict check. Any of these surfacing a structural killer ends the review before further analysis.
- HARDEN round. Failure-mode enumeration. Each concern either dismissed with cited evidence or escalated to a kill. Iterates until critiques converge to micro-edits (no new structural risks).
- STEELMAN round. If HARDEN kills, the verdict is examined with primary-source verification before being accepted. Crunchbase-existence isn't the same as a real competitor; first-pass critic verdicts get re-tested before discarding the bet.
- Verdict. KILL · BUILD-WITH-MITIGATIONS (with the mitigations as preconditions) · BUILD-WITH-PIVOT (with the pivot shape named).
- Kill criteria. Week-1, week-4, week-12 signals that ratify or invalidate the bet if you decide to build anyway.
What it looks like
The same methodology was applied to one of my own bets earlier today; the artefact is published in full as a worked example:
Case 001: UK leasehold service-charge audit, killed in 121 seconds →
That's the shape of the artefact you'd receive. Yours covers your bet; that one covers mine.
Who it's for
Specifically:
- UK indie SaaS founders considering a pivot or a new bet within an existing portfolio.
- UK bootstrapped operators evaluating productisation of an existing service business.
- UK outside-IR35 contractors transitioning from consultancy income to product income.
- UK small agency owners weighing a productisation move.
UK-jurisdiction-aware throughout. Statute citations inline where relevant (IR35 / ITEPA 2003, Companies Act 2006 small-company thresholds, Late Payment Act, CDPA 1988, regulator perimeters where the bet sits near a reserved activity). The methodology specifically tests against UK-specific structural facts (e.g. UK Government's i.AI lab shipping AI-agent data infrastructure free, MIT-licensed, which kills a class of UK-public-data API bets at concern 0).
Run it inside your own chat (free, DIY)
The methodology is also published as an MCP server. If your assistant supports the Model Context Protocol, you can connect it once and call the review from inside any conversation — your model does the reasoning, your tokens, no involvement from me.
kill_my_idea · MCP server
https://mcp.miralinhart.comStreamable HTTP transport, protocol version 2025-06-18. One tool: kill_my_idea. Pass your bet, the closest customer alternative, and your acquisition channel; the tool returns the structured methodology brief and your model runs the review.
- Free, no auth, no signup. (Stripe-gated bearer tokens later if paid signal materialises.)
- Methodology version published at
v2026-05-11; bumped when the rules change. - What it isn't: I'm not in the loop. The MCP version doesn't have my cited-evidence verification — your model's verdict is your model's verdict. The £29/£79 tiers below are where I do the citation work.
Pricing
First 10 case studies
Free in exchange for permission to publish the artefact on this site as a case study. Anonymised by default; named on request. The audit trail is the public artefact; your bet's confidential details are summarised, not transcribed.
PICK round (single)
One adversarial round: pre-survivor checks plus structured failure-mode enumeration with cited evidence. Useful for "does this bet survive an initial adversarial pass?". Delivered as a self-contained PDF or shareable URL within 48 hours.
Full PICK + HARDEN + KILL-or-BUILD
Three rounds: PICK (survivor selection across alternatives if you have them, or focused analysis of a single bet), HARDEN (failure-mode enumeration with convergence rule), STEELMAN (verification of any kill verdict against primary sources before acceptance). Verdict plus kill criteria at week-1 / week-4 / week-12. Delivered within 72 hours.
How to apply
What to include in your email
- The bet's one-paragraph description (what you'd build, who pays, what they get, what price).
- The closest alternative the customer would reach for if you didn't exist (be specific — name URLs if you can).
- Your acquisition channel hypothesis (where the first paying customer comes from).
- Any constraints worth knowing (capital, build-time budget, regulatory exposure, persona considerations).
- The verdict you'd like the review to test ("convince me to build it" or "convince me to kill it" — both work).
Apply for a free case study
If your bet fits the UK indie / bootstrapped / IR35-transition shape, the first ten reviews are free in exchange for permission to publish the artefact. Email with the five points above.
If your bet doesn't fit the UK angle but you'd still like the review at £29/£79, say so in the email and I'll let you know whether I'll take it on.
What I won't review
- Adult-adjacent business models (debanking risk to the whole stack while I sit behind a single merchant of record — see my notes).
- Gambling, FCA-regulated investment products, or anything where my role would itself be a regulated activity.
- Bets the operator has already decided to build — the methodology is invalidation discipline, not feature feedback.
About the methodology
The process was developed live across one day of running it on six of my own bets, five of which got killed. The journey is documented here. The methodology has converged on four mandatory pre-survivor checks and three review modes; it earns its place by being adversarial-by-default and citation-stamped rather than a polite ChatGPT pass.
If you're curious whether it would work for your bet specifically, the case study above is the most honest answer. The free tier is the safest first step.