Honest.
We'll tell you NEEDS MORE EVIDENCE when it's true. No fake APPLY scores to make you feel good.
35 / NEEDS MORE EVIDENCE
Live · 2026 job search
We show the gap before you send the CV.
Run one free fit scan against one JD. The engine scores evidence, flags risky claims, and rewrites one bullet with proof placeholders. No account. Median scan: about 28 seconds.
$0 today · about 0.1 cents per scan in real model cost · $79 deep audit waitlist
Why different
Most CV tools polish. First Landed checks.
Generic AI makes weak evidence sound smooth. First Landed scores the CV against one JD, names missing proof, and leaves [NEEDS PROOF] where the engine cannot verify a claim.
The point is not a better adjective. The point is knowing what you can defend in an interview.
Three things different.
Specific.
Every risky claim quotes your exact bullet. Every rewrite shows before and after, with [NEEDS PROOF] placeholders where the engine can't verify.
Fact-Checker caught this
Cross-border.
Built for SG/HK/CN/TW finance moves: markets, payments, data, product, business analyst, and first-role applications. No borrowed bank logos, no personal story, no generic AI feedback.
SG
HK
CN
TW
Today: 4 stages. Paid pack: 11 stages.
Same engine architecture. The free scan runs R0/R3/R5/R7. The $79 deep audit runs R0-R10 + R9b.
Submit your JD and CV
Drop a PDF or paste text. Single shot, no account.
R0-R3: Engine reads + ranks
Finds your strongest match, then scores against the JD.
R5: Fact-Checker
Flags risky claims, including synthetic chronological impossibilities like "led a desk in 2008" while the same example profile is a 2008-2011 undergraduate. Example: timeline impossible
R7: Verdict + rewrite
Final score, missing evidence list, and one example bullet rewrite with proof placeholders.
$79 deep audit adds stages R1, R2, R4, R6, R8, R9, R9b, R10 around the live scan. Multi-model attack-defender debate on every bullet. Currently waitlist. Email [email protected] to join.
Two numbers we can defend.
0
Average engine cost per scan
0
Median scan latency
EXAMPLE VERDICT
STRETCH
72
Example role has credible payments and analytics evidence, but the CV still needs proof for ownership, production users, and measurable execution quality.
Missing evidence: 4
Risky claims: 2
Synthetic example, not a real person
Example verdict: FX payments analyst.
A strong angle still needs proof.
The engine treats "owned FX pricing automation" as risky when the example CV only shows project coordination. The rewrite narrows the claim to coordination and keeps [NEEDS PROOF: users and quality measure] visible.
See how the stages workFree today. Paid packs on the runway.
Fit Scan
One role, one honest verdict.
$0
/ one role
- 4-stage engine (R0→R3→R5→R7)
- Verdict + score + strongest angle
- Missing evidence + risky claims with Fact-Checker
- One example bullet rewrite
Live. No login. Runs in your browser.
Single Role Pack
One serious application.
$79
/ one role
- Full 11-stage adversarial debate (R0-R10 + R9b)
- Multi-model attacker / defender on every bullet
- Full CV rewrite, not just one example
- JD-specific cover letter draft
Coming Q3 2026.
5-Role Pack
For a compact search sprint.
$299
/ 5 roles
- 5 unique JDs in the pack
- All Single Role Pack features x 5
- Cross-role consistency check
- Priority email response (24h)
Coming Q3 2026.
The questions a careful user asks.
Yes. The 4-stage Fit Scan is free for one role, no login, no email required. Engine cost is about 0.1 cents per scan and we eat it. We're tracking real cost data to figure out when to charge.
Today the engine prompt is English-only. Bilingual CVs work if the bullets are in English. Chinese-output version (/zh) is the next milestone after this design lands.
Yes, we parse PDF + TXT in your browser. PDFs work for most bank/consulting CVs. If your PDF is image-only or has weird layouts, copy-paste into the template field instead.
The scan runs in your browser and sends CV+JD to one OpenAI-compatible model call. We don't have a database. We log anonymized order ID + cost + latency to a Cloudflare KV bucket for capacity planning. Failed scans also log a failure-class enum - still no CV content.
You shouldn't trust an AI on its own. The Fact-Checker (R5) flags impossible claims in clearly labeled synthetic examples, such as an example CV claiming "led a desk in 2008" while the same example says the candidate was a 2008-2011 undergraduate. That check is the difference from generic ChatGPT feedback.
Currently waitlist. Full attacker/defender adversarial debate on every bullet, Code Red on factually risky claims, JD-tailored cover letter, plus the existing 4-stage Fit Scan output. Email [email protected] to join.