Startup Autopsy Laboratory
Your idea has a fatal flaw.
Find it for $5.
Six months from now, you'll have either a product or a lesson. Four AI examiners tell you which — before you build. Verdict in 60 seconds. No account.
Latest verdict
You're racing DispaLoadIQ to lock in 500 owner-operators before they add an SMS layer to a product that already has load board APIs you don't.
Case no.
4471
Examiners
4 agents
On record
1,255
Status
Examination complete
Subject — idea submitted
“An app that reminds you to drink water”
Preliminary scan — agent findings
Market
No moat detected
Tech
Commodity execution
Finance
Negative unit econ
Timing
Market saturation
Cause of death— tap to reveal
“There are 200 of these on the App Store. The good news ends there.”
No account · No signup · Verdict in 60 seconds · $5 flat · 2 min vs. 6 months · 2-min walkthrough
Buying for a team or friend? Get a case bundle →
Live demo
What a roast actually does to an idea
Idea in. Fatal flaw out. Pivot clear.
“An app that reminds you to drink water and tracks your daily intake”
There are 200 of these on the App Store. The top three are free. Apple Health tracks this passively. You're asking someone to change a habit for a problem they've already accepted.
"Daily water tracker for kidney stone patients."
Their doctor told them to drink 3L. They have a medical reason to care. Charge accordingly.
Loops automatically · yours takes 60 seconds
Free preview — no payment needed
How good is your idea, really?
Recent verdicts
View all →“FleetMind — AI dispatch and route optimisation for independent trucking operator…”
You're racing DispaLoadIQ to lock in 500 owner-operators before they add an SMS layer to a product t…
“ShopCompliant — EU regulatory automation for Shopify merchants. The problem: 4M…”
You're selling €20M fine insurance at cookie-banner prices — either raise the price or accept cookie…
“What I'm building: Platform Architecture Authority (PAA) — always-on architect…”
You have 18 months before Microsoft ships native evidence generation — and your MSP sales cycle eats…
The full range
Not every idea gets roasted
The verdict depends on the idea. Here's what each outcome looks like.
“IFTA compliance software for owner-operator truckers — automates fuel receipts, mileage logs, and quarterly state filings. $49/month.”
“Mandatory compliance, captive audience, no modern SaaS incumbent. The regulatory moat is real.”
Market
IFTA compliance costs owner-operators 8–12 hrs/quarter. TruckingOffice and Q7 are legacy desktop. No modern SaaS owns this niche.
Finance
$49/mo with near-zero voluntary churn — compliance software is involuntary. CAC via trucking Facebook groups runs sub-$40.
Why it passes
Regulatory moat, near-zero voluntary churn, reachable niche via trucking communities. No fatal assumption in the analysis — the panel cleared it.
“AI meeting summaries for solo consultants — auto-generates client-ready notes from recordings. $12/month.”
“Right pain, wrong market. Otter.ai owns the generic case. The moat is in the liability, not the transcription.”
Market
Generic meeting notes is saturated. But financial advisors and lawyers need compliance-grade records — a legal audit trail, not just a summary.
Timing
SEC and FCA record-keeping requirements tightened in 2024. Regulated consultants are actively shopping. Window is open.
Pivot
Narrow to compliance-grade records for regulated advisors (financial, legal). Double the price. The audit trail is the product, not the summary.
“A social network for dog owners — share photos, find local playdates, discover dog-friendly venues.”
“BarkHappy, Dogster, and Meetup.com already lost this war. Facebook Groups won by default.”
Market
Five funded attempts at this exist. All pivoted or shut down. Facebook Groups solve the core use case for free with the existing audience already there.
Finance
Niche social networks require critical mass before any monetization. CAC exceeds $80 before the first dollar of revenue. Math does not work.
Full panel — example output
Here's exactly what $5 gets you
A real analysis, shown in full. Every agent finding. Every red flag. Every action. Nothing cut.
Subject — idea submitted
“A subscription app that teaches kids to code through interactive games — $9.99/month”
The verdict
“Code.org has $70M in backing and is free. You're selling what they give away.”
Market is real but brutally crowded. Scratch, Code.org (backed by Bezos and Zuckerberg), Tynker ($130M+ raised), and Khan Academy dominate with free products. Your $9.99/mo is competing against free.
Technical execution is straightforward — no moat in the stack. A two-person team at any of the 30+ existing edtech players could replicate your MVP in a sprint. Interactive coding games are a solved UI problem.
CAC for children's apps through paid social runs $35–80. At $9.99/mo, payback is 4–8 months. Churn in kids' apps is brutal — content requirements are constant and parents cancel when interest wanes after week three.
Edtech had its window during COVID lockdowns. The spike that launched Duolingo's explosive growth and pushed parents to pay for home learning tools has largely closed. School districts are now actively pushing back on screen time.
Competitors found during analysis
Live dataCode.org
$70M+ raised
Free. Backed by Bezos, Zuckerberg, Gates. 70M+ students. Your primary obstacle.
Tynker
$130M raised
Paid curriculum. Already in 100,000 schools. Direct competitor with a decade head start.
Scratch (MIT)
Non-profit raised
Free forever. 100M registered users. Sets the price expectation: zero.
Cause of death
Free competitors with massive backing
The market leader charges nothing. Overcoming free requires a 10x experience advantage or a different distribution channel entirely.
Post-COVID edtech churn
Retention data from the 2021–2023 edtech boom shows median churn of 40–60% after month two once novelty wears off. The TAM is real; the LTV is not.
Wrong decision-maker
Children don't pay. Parents pay — and they are already overwhelmed with subscription fatigue. The gap isn't better coding games; it's trust and habit formation.
⚠ Blind spot
You're building for the child but selling to the parent. The parent's actual fear isn't "will my kid learn to code" — it's "will they actually use this after day four." No amount of game design solves the abandonment problem without a retention loop aimed at the parent, not the kid.
Recommended intervention
Call 5 after-school program directors this week — not to sell, to ask what they spend on curriculum and what problem they can't solve. You need 3 "yes, that's real" signals before building anything.
Price one district pilot at $5k for 90 days, all-in. Remove the budget objection so you can test the actual product.
Kill the parent-facing subscription page. Every hour spent on B2C conversion is a distraction from the channel that can actually scale.
Map the procurement cycle for one specific district type (e.g. Title I schools — federal funding available, underserved, motivated). That's your wedge.
Sign up for Duolingo for Schools and spend 30 minutes understanding exactly what they don't offer. That gap is your pitch.
Or see a free sample first
Case files on record
Recent examinations
Subject
“FleetMind — AI dispatch and route optimisation for independent trucking operators (1–10 trucks). The problem: 1.9 million trucking companies in the US operate 1–5 trucks. They lose 15–25% of potential revenue to empty return miles (deadhead). Enterprise load boards (DAT, Truckstop.com) exist but charge $150–300/month, require annual contracts, and are designed for brokers and large fleets — not a 3-truck owner-operator who needs to know "what load can I grab on the way home." What we're building: A WhatsApp + SMS-native AI dispatcher that: 1. Monitors live load boards and texts the driver when a backhaul load appears on their route home 2. Handles rate negotiation boilerplate with brokers via automated message templates 3. Generates a simple daily P&L (fuel cost vs load revenue) texted each evening 4. No app to download. Works on any phone. Works offline (SMS fallback). Pricing: $79/month flat. No contracts. Free 30-day trial. Stage: 6 paying customers, $474 MRR. Found via Facebook groups for truckers. Target market: Independent owner-operators, 1–5 trucks, US only for now. Business model: bootstrapped”
Cause of death
“You're racing DispaLoadIQ to lock in 500 owner-operators before they add an SMS layer to a product that already has load board APIs you don't.”
Subject
“ShopCompliant — EU regulatory automation for Shopify merchants. The problem: 4M+ Shopify stores globally, a significant portion selling to EU customers. EU compliance is a maze — GDPR cookie consent, DSA (Digital Services Act) seller disclosure requirements, and GDPR data subject request handling. Most Shopify merchants (SMBs) are non-compliant because OneTrust charges $15K+/year (enterprise-only) and generic cookie banners don't cover DSA or DSAR handling. Fines are €20M or 4% of global revenue. What we're building: A Shopify app (one-click install) that: 1. Auto-scans the store for compliance gaps across GDPR + DSA 2. Injects a compliant consent banner (not generic — dynamically adjusts per visitor country) 3. Generates and hosts the legally required DSA seller information page 4. Routes and responds to GDPR data subject access requests automatically Pricing: $49/month. No setup fee. No contracts. Stage: Pre-revenue. 200 email signups from a waitlist page. Target market: Shopify merchants in the US, UK, Canada selling to EU customers. Business model: B2B SaaS, Shopify app store distribution.”
Cause of death
“You're selling €20M fine insurance at cookie-banner prices — either raise the price or accept cookie-banner margins against SWEDev.”
Subject
“What I'm building: Platform Architecture Authority (PAA) — always-on architecture intelligence for Microsoft environments. Azure + M365 + Zero Trust in one platform. 723 automated checks, AI synthesis that turns raw data into consultant-grade findings, IaC generation, ADRs, and board-ready evidence. Who it's for: Microsoft Gold Partners (MSPs) who need to deliver architecture intelligence continuously across their client base — without hiring a senior architect per client. Secondary: mid-market Azure tenants (50–500 employees) facing NIS2 or ISO 27001 pressure with no dedicated security team. Where I stand: Built and running. I spent 10 years at Microsoft in the Azure Engineering org doing exactly these engagements manually. PAA started as personal tooling — cut my own delivery time from 2 weeks to 2 days. Now productised. Why it might be stupid: Microsoft gives away WAF assessments for free. Defender for Cloud exists. Every CSPM vendor in the world is bolting AI onto their scanner and calling it intelligence. And I'm one person trying to sell into a partner ecosystem that moves slowly and buys on relationships. Why I think it isn't: Nobody starts from architecture. Every competitor starts from security and works backwards. The WAF free tool is a questionnaire — it doesn't touch your infrastructure. Defender finds what's wrong, not what to do. The MSP white-label model means the product sells through partners who already have the client relationships. And the 2-weeks-to-2-days proof exists — I lived it.”
Cause of death
“You have 18 months before Microsoft ships native evidence generation — and your MSP sales cycle eats 4 of them before a single contract closes.”
0 cases on record
Standard examination protocol
How it works
INTAKE
01
Submit the specimen
Describe your idea — what you're building, who it's for, where you stand. Rough is fine. The agents do the research.
EXAMINATION
02
Four agents convene
Market pulls live competitor data. Tech assesses execution risk. Finance runs the unit economics. Timing checks if the window is open.
VERDICT
03
Cause of death filed
A survival tier, a one-line verdict, the strongest case for your idea, the top kill shots, the blind spot you missed, founder fit assessment, and 3–5 specific actions to take in the next 7 days. Permanent and shareable.
Our forensic specialists
Why not a
generalist AI?
A general-purpose model is optimized to keep you engaged — it will validate your idea. Our four-agent panel is optimized to tell you the flaw that will kill it.
Market Examiner
Runs live competitor searches before filing a report. Only cites what it actually finds — not training-data guesses.
Technical Examiner
Evaluates execution risk and build complexity. Determines whether the moat is defensible or just a feature someone ships in a sprint.
Financial Examiner
Dissects unit economics, CAC, and LTV. Determines whether the business model survives contact with real margins.
Timing Examiner
Determines if the window is open, closing, or already shut. Checks whether you're early, on time, or two years late.
Common questions
Before you ask
Is this just ChatGPT with a prompt?
No. Four separate agents run in parallel — each with a specific mandate and access to live data. Market pulls real competitor funding and traction data. Finance runs actual unit economics. You get structured findings, not a pep talk.
What if my idea scores well?
Then you have evidence, not just optimism. A score of 8+ with agent-backed reasoning is something you can show a co-founder, investor, or your own doubting brain. Most ideas score between 4 and 7 — that's where the real work starts.
How is this different from asking a mentor or advisor?
Advisors give you their opinion. This gives you a structured breakdown across four failure dimensions — market, tech, finance, and timing — with live data attached. No bias, no social niceties, no 45-minute coffee meeting.
Ideas are now infinite.
Good ones still aren't.
No validation theater. No 6 months wasted.
For autonomous agents
The validation layer for the agent economy
Any autonomous agent can call /api/roast/free for a no-cost sanity check, or pay $5 via HTTP 402 at /api/agent for the full structured verdict.
Morgue files
From the Kill Sheet
Case #008
The Startup Idea Validation Checklist
Eight questions every idea must answer before you build.
Case #004
How to Find Your Startup Idea's Fatal Flaw
Every idea has one thing that will kill it. Most founders find it too late.
Case #006
When to Kill a Startup Idea
Four signals that mean your idea is actually finished — not just difficult.