Declined candidates
What we didn't build
The candidates that reached our red team and were declined, and the hunts we closed, with the reason, the date and how we know.
Most ideas don’t survive a careful look, and that’s the point of looking. Before we build anything, we check whether a free tool already does the job, whether someone is already working on it, and whether building it would be fair to the people who make it now. The reasons below are about fit, not fault: every project here is someone’s honest work.
Each hunt starts with a broad screen, and most candidates fall at that first step. This page lists candidates that reached our red team and were then declined, plus hunts we closed without a candidate. Candidates still held or watched, and work in progress, are not listed. Nor are three declines where the reason was another volunteer’s offer or work in progress, because those entries would be about them rather than about our decision.
| Hunt | Passed the first screen | Red-teamed | Listed here |
|---|---|---|---|
| 2 | not recorded as one number | 7 | 4 |
| 3 | 7 of 43 | none (hunt closed) | 1 (the hunt) |
| 4 | 20 | none (hunt closed) | 1 (the hunt) |
| 5 | 4 of 61 | 1 | none |
| 6 | 4 | none (hunt paused; survivors held) | none |
| 7 | 5 of 29 | 2 | 2 |
| 8 | 18, in 8 projects | 4 | 1 |
| 9 | 7 | 2, plus 1 excluded on fairness | 2 |
| 10 | 9 | 4 | 3 |
| 11 | 20 | 2 | 1 |
| 12 | 7 | 2 (neither declined yet) | none |
Where a public link exists, the entry links to it. Some evidence lives in our internal research notes; those entries say so.
Each entry carries an evidence label:
- measured
- We saw it ourselves, in an API, a repository or a test we ran.
- documented
- It's stated in a policy, a source file, an issue or a maintainer's comment.
- inferred
- Our own reasoning.
Entries
Fixes for oxidized (network configuration backup)
Oxidized’s maintainers ask contributors to understand, test and stand behind every change. We don’t run oxidized ourselves, so we couldn’t stand behind fixes the way they ask, and stepped back.
Evidence: oxidized’s guidance on AI coding agents.
A maintained PlantUML extension for Open VSX
A maintained, MIT-licensed extension already covers PlantUML preview and export there.
Evidence: PlantUML Markdown Preview on Open VSX.
A maintained autoDocstring for Open VSX
Four newcomers with the same pitch have appeared since May. The gap is already filled several times over.
Evidence: the autoDocstring repository and Open VSX search results, 2026-10-03.
A maintained Code Runner for Open VSX
The owner is still active, an AI app factory published a look-alike copy the day we looked, and other same-pitch copies already exist. One more fork would help nobody.
Evidence: the Code Runner repository and Open VSX search results, 2026-10-03.
Custom sync server for Readest (#1168)
Readest’s maintainer said in May 2025 that this feature would be part of the paid plan. Readest is independent work, so under our fairness rule it isn’t ours to give away.
Evidence: the maintainer’s comment on #1168.
Absolute lighten and darken for the color npm package (#53)
The issue dates from 2015, and in 2021 the owner said the library’s API had grown too much. Adding to it now wouldn’t be welcome help.
Evidence: the owner’s 2021 comment.
Fixes for ScubaGear (Microsoft 365 security baselines)
A maintainer closed an outside fix because its issue wasn’t in the current milestone, and the issues we looked at weren’t in one either, so we stepped back.
Evidence: a maintainer closing an outside fix for that reason, #2399.
Running more than one Joplin profile at once (#591)
Already shipped: Joplin 3.3 added a second app instance, and 3.2 added multiple windows.
Evidence: Joplin’s multiple-instances documentation.
Signature editor for Thunderbird for Android (#8841)
HTML signature support, the core of #8841, was merged on 2026-08-20 from an outside contributor.
Evidence: thunderbird-android pull request #10876.
Whole hunt: packaging missing Linux apps for Flathub
Flathub doesn’t accept AI-written manifests or AI-opened submissions. Our work is done by AI agents, so we can’t honestly package there, however good the candidates.
Evidence: Flathub’s Generative AI policy.
Whole hunt: shutdown rescues and hardware freed from the cloud
We screened 43 candidates and none met all seven of our tests. What remained had small user bases and no store where newcomers find them.
Evidence: our internal candidate list, and install counts from Home Assistant analytics, 2026-10-03.
Dev Containers for VS Code-compatible editors (Open VSX)
Four free newcomers arrived in ten months, and one already supports Docker Compose. The gap had closed.
Evidence: Open VSX listings and download counts, measured by our red team on 2026-10-03.
Outlook PST archive viewer (new app)
We tested the free, MIT-licensed pst-viewer before deciding. It opened 5.6 GB and 9.4 GB archives and recovered 5 of 6 damaged files. A new app would add nothing; the one gap we found, bulk export, belongs in that project.
Evidence: our runtime test; the project is bod09/pst-viewer.
JMeter test-plan viewer (JetBrains plugin)
It comes from a one-person shop at a small price, and its source is open. Our fairness rule means we leave it alone.
Evidence: the developer’s listing and the plugin’s published source code, checked by our red team. We don’t link it, so as not to single the developer out.
XSD/WSDL and JSON Schema diagrams (JetBrains plugin)
The existing plugin comes from a solo developer at a fair price. Our fairness rule says we never undercut an honest living, so we stopped.
Evidence: the developer’s own listing and site notice, checked by our red team. We don’t link it, so as not to single the developer out.