Buy the platform once. Pay for AI as you use it.
The expensive, variable part is the AI — so we don't pretend it's free. Platform access is a one-time purchase. Credits are separate, and every action's price is published below.
Attacking an idea and grading a spec are free and need no account. Signing up adds starter credits, saved results and watchlists. Generating a build spec needs membership. You can cancel a subscription any time — it stops at the end of the period you already paid for.
Founding 100
$79$159one time
Seat #3 of 100 · 98 left at this price
The price rises with each tier as the database grows. This is a real ladder, not a countdown that resets.
- Every idea with its full evidence chain
- Counter-evidence engine — the case against building it
- All 30 platform adapters (ZIP + copy-paste)
- Unlimited watchlists and alerts
- 1 idea lock per month — 90-day exclusivity
- MCP access for Claude Code, Cursor and Codex
- Every future Pro feature, forever
- 300 starter credits (they don't expire)
- Lifetime BYOK — use your own API key, no monthly fee
Not included: unlimited AI. When the starter credits run out you either top up or connect your own API key. We'd rather say this here than surprise you later.
What needs an account, and what does not
One table, generated from the same rules the app enforces. If a row here is wrong, the app is wrong.
| Capability | Needs |
|---|---|
| Attack an idea and get a verdictThree per hour without an account. | nothing |
| Grade a spec you already wroteRuns in your browser; the text is never sent to us. | nothing |
| Read the open ideas in fullEvidence chain, counter-evidence, verdict and every source link. | nothing |
| See the thresholds behind every verdict | nothing |
| Keep your attack results | a free account |
| Starter credits on sign-upEnough to try a full attack and one spec. | a free account |
| Watch a problem area for new signals | a free account |
| Every idea in the catalogue | membership |
| Full attack — complete evidence chain and validation planCosts credits; the price is shown before you spend. | membership |
| Generate a build specOne Universal Core, then any platform format for less. | membership |
| Connect a coding agent over MCP | membership |
What a credit actually buys
Most tools hide this. Here is the whole list. Search costs nothing — it's a database query, not a model call, and charging for it would be dishonest.
| Deeper research on an existing run | free |
| Quick idea attack | free |
| Idea search / filtering | free |
| Watchlist weekly digest | 1cr · $0.10 |
| Refresh spec from existing core | 2cr · $0.20 |
| Founder-fit scoring (20 ideas) | 2cr · $0.20 |
| Chat message | 2cr · $0.20 |
| Quick Attack | 5cr · $0.50 |
| Platform adapter package | 5cr · $0.50 |
| Deep analysis / counter-evidence | 6cr · $0.60 |
| Full idea attack | 12cr · $1.20 |
| Full validation report | 18cr · $1.80 |
| Lock an idea for 90 days | 25cr · $2.50 |
| Universal Core spec | 32cr · $3.20 |
| Full Attack | 40cr · $4.00 |
| Deep Research | 80cr · $8.00 |
Top-up packs — never expire
100 credits
$0.10 each
500 credits
$0.096 each
1,200 credits
$0.095 each
3,000 credits
$0.095 each
Straight answers
- What can I do without paying?
- Attack an idea and read the verdict, its confidence, its counter-evidence and every source behind it — three times an hour, no account. Grade a spec you already wrote, which runs entirely in your browser. Read the open ideas in full. Signing up adds starter credits, saved results and watchlists.
- Then what actually costs money?
- Generating a build spec, and running a full attack rather than the quick one. Both call a deep model, and a model call has a price we pay per request. The credit cost is shown before you spend anything, and a failed generation is not charged.
- What happens to the starter credits?
- They sit in your balance and do not expire. They are enough to run a full attack and generate one spec, which is the point — you should be able to see the paid output before deciding whether it is worth paying for.
- Is lifetime access really one payment?
- Yes. Platform access is bought once and does not renew, expire or convert into a subscription. Credits for model calls are separate and always have been — we would rather say that plainly than hide a running cost inside the word lifetime.
- Can I use my own API key?
- Yes. Bring an Anthropic, OpenAI or OpenRouter key and the credit system switches off for your account — you pay your provider directly and see your own token counter. The key is encrypted at rest and never appears in logs or error reports. The batch pipeline that builds the database still runs on our account, not yours.
- Is the counter-evidence real, or just a section header?
- It is generated from the same records as the supporting case: shutdown competitors, free alternatives already solving the complaint, incumbents who shipped adjacent features, and unit economics that don't close. If an idea has no falsifying evidence, we say that means thin data — not a safe bet.
- Where does the data come from, and do you scrape?
- No scraping. Hacker News, GitHub, GitLab, the Stack Exchange network, Product Hunt, App Store reviews, Discourse community forums, EU procurement notices and TrustMRR — all official APIs. We keep the derived signal, a short attributed excerpt and a link to the original; the full text is discarded once the signal is extracted.
- What if the verdict on my idea is wrong?
- Then argue with the threshold rather than the tone. Every rule that produces a verdict is published on the methodology page with the reasoning behind the number, and every claim in the result opens the record it came from. If a source says something other than what we concluded from it, that is a bug and we want to hear about it.