The text stays here.
The brief is processed in JavaScript and a Web Worker. Model inference runs locally in your browser after downloading openly available weights. No account, form submission or hosted prompt API.
FREE • NO SIGN-UP • YOUR BRIEF STAYS IN YOUR BROWSER
Don't confuse a bounty
with a paycheck.
Give a gig a fair shot before giving it your weekend. Turn a messy task brief into checkable evidence, unresolved payout questions, and a transparent hourly-value estimate—with local open-weight AI.
DECISION INTELLIGENCE01 / 03Read the ask → Price the risk → Plan the proof
◈ Open-weight FLAN-T5▢ No API keys⌁ Local inference↗ Export proof plan
THE QUESTION“Is the reward real, and what's the work?”START WITH EVIDENCE, NOT HYPE.
01 / THE WORKBENCH
A 60-second preflight.
Zero imaginary revenue.
Paste a publicly shareable brief or draft terms. Use a synthetic example to see how it works. Do not paste secrets, passwords, identities or proprietary source.
INPUT / BRIEFMAX 12,000 CHARS
Your task text is not sent to an inference API. On first AI run, the browser downloads a public model and a local WebAssembly engine from the project's static assets/model host. AI may be slow or unavailable on some devices; deterministic checks work independently.
OUTPUT / CHECK BEFORE ACCEPTINGNOT STARTED
◎
A listing isn't a receipt.
Run the preflight to surface evidence requirements and questions that the brief doesn't answer. No scanner can verify escrow or guarantee payment from pasted prose.
Draft evidence plan
REVIEWEvidence the posting mentions
Risk patterns, not verdicts
Questions to clarify before starting
✦ OPEN-WEIGHT AINOT LOADED
Run local AI to turn the task into a concise proof hypothesis, buyer question and bid opening line. Each suggestion needs human review.
Preparing…
MODEL'S EVIDENCE READ
ONE BUYER QUESTION
A DRAFT OPENING LINE
This is a probabilistic model's text, not verified facts, a paid acceptance, or a promise of safety.02 / THE EXPECTED-VALUE DESK
The number worth
being skeptical of.
No invented “win probability.” You enter the odds. We show the calculation, and we label every estimate as yours.
15%
0% no selection100% certain selection
For a contest, 15% is only a demonstration assumption—not an estimate from entry data. For a contracted fixed-fee job, use your own evidence-backed payment probability instead.
PROBABILITY-WEIGHTED NET / YOUR ASSUMPTIONS$0.00Expected value, not money received
Expected per work hour$0.00
Net if selected$0.00
After platform fees$0.00
Break-even win chance*0%
*Break-even here only means covering direct spend, not your labor opportunity cost. All calculations ignore tax, exchange rates, uncertainty in selection criteria, delayed payouts and transaction risk.
03 / THE METHOD
Local by design.
Honest by default.
Use what machines are good at—text-to-text summarization and consistent math. Keep payment and eligibility verification in human hands.
Math is fully visible.
One honest formula: (reward × (1 − fee) × your odds − costs) ÷ hours. Never replace assumptions with “AI confidence” or call a posted bounty an actual receipt.
Proof is the deliverable.
Checklists are generated from explicit phrases like “repo,” “video” or “test.” The open model adds a draft synthesis; you must verify it against actual task requirements.
OPEN MODEL Xenova/flan-t5-small (Apache-2.0)ENGINE Transformers.js / ONNX Runtime WebCODE MIT licensed
BETA / SMALL, SPECIFIC SIDE HUSTLEWould you pay for a
Would you pay for a
more rigorous first pass?
A possible $9/month personal review tool or a bespoke $19 one-off evidence briefing is a **pricing experiment**, not a subscription for sale today. No checkout, no promises, no affiliate links. A pilot buyer would need a separately agreed scope and terms.