The most common questions about ThesisGap. Not covered? Email yan@thesisgap.com.
One full thesis analysis: keyword extraction, Polymarket market search, AI decomposition into sub-claims, direction labeling (supports / contradicts / neutral) for each matched market, Sonnet fallback analysis for unmatched claims, and the final Gap score computation. Typical runtime is 30-90 seconds.
No. Buy once, use anytime. If we change pricing rules in the future, existing credits are grandfathered under the terms you bought them — no "balance expires in 60 days" reset.
The credit is refunded automatically. If AI decomposition fails, produces zero valid sub-claims, or the background task exceeds 90 seconds without responding, the credit returns to your balance. You only pay for analyses that actually produce results.
No. One-time Stripe checkout. Top up when you need to. No recurring charges, no cancellation flow, no "free trial then auto-renew" trap.
Within 14 days of purchase — email yan@thesisgap.com with the Stripe receipt. Beyond 14 days, no refund, but the credits never expire so you can just keep them for later.
Decomposition uses the primary LLM at temperature 0.1 for stable output. Classification runs two independent LLMs as an ensemble — they agree on ~82% of pairs. Manual sampling shows the direction label is correct on roughly 92-95% of (sub-claim, market) pairs. The analysis page lets you manually override any label.
For sub-claims with zero matching markets, a separate LLM writes a standalone analysis rendered as an amber "AI-only signal, not market-backed" card. You decide whether to include it in the Gap calculation. Polymarket covers US politics, crypto, and parts of macro densely, but small-cap equities, China macro, and private markets are thinly covered.
ThesisGap itself (reading market data for research) is compliant in most regions. But Polymarket trading is CFTC-restricted in the US, inaccessible from mainland China, and of undetermined legality in India and parts of the Middle East. We don't recommend trading on Polymarket from those regions, but using ThesisGap to analyze the public market data is typically fine. ThesisGap itself does not involve any crypto trading or custody of funds.
No. Your theses and analysis results are stored in Supabase and gated by row-level security (RLS) so only you can read them. We don't publish your content, don't sell to data brokers, and don't use it for external model training. A small number of explicitly opted-in examples appear on /examples in anonymized form.
Content pages (home, pricing, methodology, examples, about, FAQ) ship in all three languages: Simplified Chinese at the root (/), Traditional Chinese at /zh-Hant, and English at /en. The product itself (thesis input, analysis results, dashboard) still ships mostly Simplified-Chinese UI copy — Traditional and English product-UI translation is rolling out. Use the language switcher in the top right to toggle; it preserves your current page.
Starter $9 / 20 analyses (~$0.45 each) is the trial tier — enough to run 2-3 theses and see if you like it. Pro $29 / 80 (~$0.36 each) is the sweet spot for most users — about 10-15 theses a month. Power $79 / 250 (~$0.32 each) is for heavy researchers or content creators. All tiers share one credit balance; multiple purchases stack automatically.
Best fit: theses with explicit time windows and verifiable thresholds — "the Fed cuts no more than 50bps in 2026", "BTC breaks $120K by Q3 2026", "OpenAI valuation exceeds $500B by year-end 2026". Less well-suited: purely qualitative macro takes ("US-China relations will deteriorate" — no time window, no numeric threshold). The more concrete and verifiable your thesis, the more sub-claims end up matched to real markets.
Yes. Each re-run costs 1 credit but uses the latest market probabilities and pool for that day. Running the same thesis 2-3 times in a week is reasonable (market conditions drift); re-running every day usually doesn't show meaningful change. The dashboard keeps a history of every thesis you've submitted.
A Gap score is neither good nor bad on its own — it quantifies the size of your disagreement with the market consensus. |Gap| < 10 (green) means you're broadly aligned with the market, which is itself a signal that your independent thinking matches the $1B consensus. |Gap| 10-25 (amber) is meaningful disagreement, worth re-examining your assumptions or checking market liquidity. |Gap| > 25 (red) is material disagreement — either you see something the market doesn't, or your assumptions have systematic bias.
For now, you can save the analysis page from your browser (Cmd+S / Ctrl+S) or print it to PDF. Structured JSON / CSV export is on the roadmap for Q3 2026. If you have an urgent API-export need, email us with your use case — high-demand features get prioritized.
ThesisGap is a solo project by Yan — based in Shenzhen, born 1993, full-stack engineering background, active investor personally. No funding, no co-founders, no board. Which means product decisions only answer to one thing: helping users see where they and the market disagree. Full background at /en/about.
Email yan@thesisgap.com — the founder personally reads and replies to every message. Product feedback, bug reports, methodology discussions, partnerships all welcome. No Discord or group chat yet — email one-on-one is faster.
Didn't find your answer?