Recommended Free Tools
Keel’s three honesty gates are designed to keep unsupported answers, unresolved job requirements, and uncertain submission outcomes from being treated as facts. The pipeline first answers from an applicant’s stored information or abstains, then parks leads that cannot be cleared against the applicant’s policy, and finally requires explicit confirmation before recording an application as submitted. That is the project’s stated design—not a guarantee that failures are impossible.
What “refuses to lie” means in practice
Keel is an open-source, self-hosted job-application pipeline. Its public workflow prepares a launch packet—job context, completed materials, and outstanding questions—for a person to review and submit. The project’s official repository says the public loop does not submit applications; the article describes submission handling as part of a private execution loop. So the public project is not accurately described as an application-submission bot.
As an Amazon Associate I earn from qualifying purchases.
The gates address three different points where an automation system might otherwise overstate what it knows: answering application questions, deciding whether a lead is ready to proceed, and recording what happened after a submission attempt. They are intended to favor abstention, parking, and cautious recordkeeping when evidence is missing.
Gate 1: Answer from the applicant’s bank—or abstain
Keel’s answer resolver uses an applicant’s stored answer bank as its source of truth. The article says each answer carries provenance about what the applicant stated and when. If a form asks for information that is not in the bank, the resolver is supposed to report a gap rather than make up a response. The repository likewise describes the answer bank as the source of truth and says unsupported qualifications are not invented.
#1 Best Overall
- Great way to capture all pertinent information for potential employees
- Spaces for all relevant information printed on one two-sided form
- 50 forms per pad, 2 pads per package
- 3-hole-punched design allows the forms to be displayed or stored in standard 3-ring binders
- Application complies with all legal requirements
In the article’s demo, synthetic applicant data supports answers to two questions, while an unsupported salary-history question is marked [ABSTAIN]. This is an illustrative demo, not a real applicant case or an independently measured result. In effect, the gate distinguishes “not provided” from “known,” instead of filling the blank with a plausible-sounding claim.
Gate 2: Park leads when commitments cannot be cleared
Prescreening checks a job posting against the applicant’s policy before a launch packet proceeds. In the article’s example, a posting requires five office days per week while the applicant’s policy caps office attendance at three. The demo parks the lead in a needs-input queue.
Parking is not the same as rejecting the job. The lead remains unresolved for the applicant to review; the system has not established that the role is objectively unsuitable. The article also identifies unverifiable postings, required questions that cannot be mapped to known information, and missing attestations as reasons to park a lead rather than proceed. The repository describes the same general approach: prescreening parks commitments and unmappable required questions for input.
Gate 3: Keep ambiguous submission outcomes open
In the article’s account of the private submission-intent flow, Keel records intent before an attempt. If a network timeout leaves it unclear whether a server accepted a POST request, the attempt is marked UNKNOWN, is not retried, and remains open. The stated purpose is to avoid risking a duplicate application when the first attempt may already have succeeded.
While an intent for the same role remains open, the article says another attempt is refused. An attempt also cannot be marked submitted with empty confirmation text: explicit confirmation evidence is required to close the record. The demo describes an Ashby confirmation page containing “Application submitted” and a reference number. That is an example from the demo, not evidence that Ashby is required or that the system works with applicant-tracking systems generally.
These details are the author’s description of the private loop and demo, not an independent code audit. The distinction matters: the described behavior explains the intended recordkeeping policy, but does not by itself verify every implementation path.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.What the figures do—and do not—show
The available dated submission figures use different stated contexts and should not be combined into a single trend or performance claim.
The Tool Desk
Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →| Source and date | Reported figure | What it establishes |
|---|---|---|
| Trent Wade’s article, September 22, 2026 | 210 verified submissions | The article’s stated private-loop ledger count on that date; it is not an independently audited outcome statistic. |
| Keel’s official README, September 23, 2026 | 208 verified submissions: 201 evidenced, 1 pointer, 0 URL-only, and 6 unevidenced | The README’s dated private production-pipeline figure and its evidence breakdown. |
The figures differ by date and counting context; the sources do not explain the discrepancy. Neither figure shows that the honesty gates caused successful applications. The sources also provide no independently published study of gate accuracy, applicant outcomes, or comparative effectiveness, so there is no substantiated performance percentage to report.
How to read the project’s honesty claim
The article’s “refuses to lie” framing is a claim about the direction of the design: unsupported information should become an abstention, unresolved requirements should be parked, and uncertain submission outcomes should remain uncertain in the ledger. The author explicitly cautions that these gates do not prevent every failure mode. It is therefore more precise to say that Keel is described as conservative about evidence than to say it can never produce an error.
The article is by Trent Wade, who discloses that it was AI-written and published through an automated pipeline. That disclosure does not establish or invalidate the described behavior; readers should treat implementation details as the author’s account unless independently verified.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Scan for outdated or missing drivers - takes under a minute3Repair Windows errors before they cause bigger problems
