404inc Project 404

[404] not found found.

A free community resurrection program for the projects that almost shipped. Send us your stalled side project, your half-built MVP, your almost-something — we'll spend a weekend with you tearing down the wall it hit, ship something demo-able, and publish the post-mortem so the next builder doesn't repeat it.

47
Projects resurrected
184hr
Volunteer hrs / quarter
9
Now profitable
$0
Cost to participants

A side project doesn't fail because the idea was bad.

It fails because the builder hit a wall they hadn't seen before. The auth refactor that took longer than estimated. The data model that needed re-thinking. The deployment story that turned into six configs and a Cloudflare ticket. The realtime layer that was supposed to be "easy."

Most of those walls aren't genuinely hard — they're hard for the specific person hitting them. There's almost always somebody who's stood in front of that wall before and now finds it boring. Project 404 is the introduction.

"We started this because we kept watching good projects die quietly — and almost every time, two days of pairing would have saved them. So we made the two days a fixture."

— Jordan Hammond, on the founding of Project 404

It runs four times a year. Two days each time. Four projects per cohort. Four to six volunteer engineers, plus the 404inc team, plus the original builder. By Sunday evening there's something demo-able. By the next Friday there's a written post-mortem in the public archive — what stalled it, what fixed it, and what the next builder hitting that wall should know.

The whole thing is free. There is no hidden tier. No "introductory engagement" pitch. No retainer email three weeks later. We do this because the company we run is built on the lessons of side projects that compounded — and we owe a quiet tax to the next ones.

Projects we'll resurrect. Projects we'll politely decline.

We pick four per quarter. Volume means we have to be honest about what fits. Nothing here is moral judgment — it's just where we can actually move the needle in two days.

What fits

  • You've already got something — code, designs, even a working prototype — and it's stuck on something specific.
  • The blocker is technical or architectural, not market or funding.
  • You can articulate, in two paragraphs, what you'd be able to do if the blocker dissolved.
  • The project would benefit other people if it shipped — open-source, indie tooling, civic tech, niche utilities.
  • You're willing to publish a post-mortem with us, including the parts where you got stuck.
  • You can show up for the full weekend — in person if local, or fully remote and synchronous.

What doesn't fit

  • You haven't started yet — the project is an idea, not a 404. We're a resurrection program, not an incubator.
  • The blocker is "I need a co-founder" or "I need funding." We can't help with either, and pretending we can would waste your time.
  • The project is a customer-facing commercial product where the unblock would be real consulting work — that's an Engagement, not a 404.
  • The codebase requires NDAs, security clearances, or regulated handling we can't accommodate in a public weekend.
  • The "wall" turns out to be that you've lost interest. That's fine — sometimes the right move is the eulogy. We can help with that, but it's a different conversation.

Tell us what's stuck.

Six fields. We read every submission within two weeks. The four we pick get a confirmation email and a date; the rest get a paragraph of honest feedback from a senior engineer — that alone is sometimes the unblock.

public link, gist, dropbox, anything we can read without signing in
we read every submission · ~14 day turnaround

Twelve quarters of resurrections.

Every Project 404 weekend ends with a public post-mortem. Below is a slice — twelve of the forty-seven we've done since June 2024. Click any card to read the full retrospective.

orchard-cms
2026.04.12
SHIPPED
Self-hosted CMS for small newsrooms. Spent 14 months stuck on an auth refactor that kept growing. We rewrote the session layer around signed cookies + a single Postgres table, dropped JWT entirely, shipped Sunday evening.
Wall: authTime to unblock: 9hrsNow used by: 6 papers
hexa.fm
2026.03.08
SHIPPED
Open-source podcast host. Builder hit a transcoding wall — couldn't get FFmpeg pipelines to scale beyond a single VPS. We moved transcode to Cloudflare Stream and rewrote the upload-to-publish flow as a worker chain. ~340 episodes hosted in the first month.
Wall: infraTime to unblock: 11hrsEpisodes hosted: 340+
blueprint-cli
2026.02.24
SHIPPED
Architecture-decision-record CLI for distributed teams. Solo-dev burnout in mid-2024 left it half-built. Two volunteers picked it up, finished the template engine, shipped a working v1 with Markdown + GitHub integration.
Wall: burnoutTime to unblock: 13hrsGitHub stars since: 1.2k
canopy-mesh
2026.02.10
ARCHIVED
P2P file sync for journalists in low-bandwidth environments. We agreed with the builder it had been overtaken by simpler approaches — wrote a thoughtful eulogy together explaining the architectural trade-offs, archived the repo with dignity. Sometimes the resurrection is the funeral.
Wall: overtakenOutcome: archivedEulogy length: 2,400 words
latte-stats
2026.01.18
SHIPPED
Coffee-shop POS analytics. Builder had an unworkable Postgres schema — fact tables joined to dimensions joined to a query layer nobody could reason about. We dropped to a star schema, added DuckDB on top for the dashboards, rebuilt the front-end. Now a paid product.
Wall: data modelTime to unblock: 14hrsPaying customers: 12
ferrum-id
2025.12.06
HANDOFF
Identity proxy for indie SaaS. Original builder had moved on but didn't want to abandon it. We did a knowledge-transfer weekend, matched it with a maintainer from the community, and helped set up the governance docs. Now actively developed.
Wall: ownershipOutcome: handed offNew maintainer: @kshen
riverdesk
2025.11.15
SHIPPED
Async help-desk for tiny teams. Stalled on a realtime layer the builder had written in raw websockets and couldn't keep alive in production. We refactored to Phoenix LiveView in 36hrs. Public beta opened the following week.
Wall: realtimeTime to unblock: 15hrsBeta signups: 480
paperback-search
2025.10.22
SHIPPED
Federated search for independent bookstores. Stuck on an embedding pipeline that cost too much per query. We replaced the cloud LLM with a quantized local model + pgvector and dropped per-query cost by 96%. Now indexing 41 stores.
Wall: ML costTime to unblock: 12hrsStores indexed: 41
trail-companion
2025.09.08
SHIPPED
Offline-first hiking app for volunteer trail crews. Stuck on map-tile syncing. We rebuilt the sync layer around CRDTs and a tile budget per device. Now used by three regional trail associations.
Wall: offline syncTime to unblock: 10hrsTrail miles covered: 1,840
civic-budget
2025.08.17
HANDOFF
Open-source municipal budget visualizer. Original builder needed to step away after a year. We helped find a civic-tech non-profit to adopt it, did the full transition over a weekend including infrastructure, secrets rotation, and contributor docs.
Wall: ownershipOutcome: handed offAdopted by: CivicCo
coral-grader
2025.07.04
SHIPPED
Community ML tool for grading coral reef photos. Stuck on a training pipeline that wouldn't converge. We rewrote the data-loader, fixed a class-imbalance bug, retrained from scratch in 8 hours of GPU time. Used by reef-monitoring volunteers across three countries.
Wall: ML trainingTime to unblock: 16hrsReefs monitored: 23
tinker-token
2024.11.30
ARCHIVED
Hobbyist crypto experiment. We helped the builder realize the project's premise had aged out and the maintenance cost was a tax on better ideas. Wrote the post-mortem on what the original instinct was right about and what it wasn't.
Wall: premiseOutcome: archivedBuilder's verdict: "relieved"

Spend a weekend in someone else's mess.

Volunteering is the heart of Project 404. The 404inc team can lead a resurrection, but four projects per quarter is too many people for two of us. We need engineers who'll show up for two days, dig into a codebase they didn't write, and pair with someone who's been staring at the same wall for too long.

You don't need a senior title. You need two days, working knowledge of at least one part of the modern stack, and the willingness to read someone else's code without judgment. The matchmaking handles the rest.

Here's what the weekend looks like:

/ FRI EVENING
Kickoff dinner. Builder, volunteers, and the 404inc engineer meet. We read the project together, agree on the wall, and split into pairs.
/ SAT MORNING
Architecture sketch. One hour, whiteboard. Then everyone codes. Espresso provided. We don't track hours.
/ SAT EVENING
Demo & re-plan. Whatever's working gets shown. Whatever's broken gets re-scoped. Sometimes the wall moves; sometimes we move around it.
/ SUN AFTERNOON
Ship & retrospective. We push something demo-able. Then we draft the post-mortem together. Builder gets the keys back at 6pm.
/ FOLLOWING WEEK
Post-mortem published. One of the volunteers usually writes it; the 404inc team edits. It goes live in the public archive.
pick all that apply · we match by complement, not by overlap
we'll match you by Jun 1

Honest FAQ.

Is this really free?
Yes. We don't take equity. We don't bill afterwards. We don't add you to a "newsletter." We don't do an introductory engagement call three weeks later. The whole point is that this is unconditional. If we ever change that, we'd rename it.
Why are you doing it, then?
Because the 404inc team learned by being unblocked by people who didn't owe us anything. This is a quiet tax we pay back. It's also genuinely fun — you spend most of your week reasoning about your own architecture. A weekend in someone else's is restorative.
My project isn't open source. Can I still submit?
Yes, with one condition: the post-mortem has to be public. We don't need your code to be public, but we do need the lessons to be — that's the deliverable. If you can't make the lessons public for legal or competitive reasons, this isn't the right program for you.
Do volunteers own what they write that weekend?
No — the original builder owns the project, including any code that gets contributed. Volunteers know this going in. We do credit every volunteer in the post-mortem, and many cite the work in their portfolios. If a volunteer wants joint ownership, that's a private conversation between them and the builder; we stay out of it.
What if you can't fix my project in two days?
Sometimes that happens. The deliverable is never "the project is finished" — it's "the wall is gone, you can see the next mile." We've had projects where the weekend was just enough to get the build working again, or to migrate one critical service. The post-mortem is honest about what got done and what didn't.
I want to fund this. Can I sponsor a cohort?
No. We deliberately don't take sponsorship, because the moment we do, we're optimizing for the sponsor's brand visibility instead of for the projects. If you want to help, volunteer. If you're an organization that wants to send engineers as volunteers, we'll happily make that work.
Can I attend without participating, just to watch?
No. The weekends are small and intense; observers change the dynamic. The post-mortems are public, the archive is public, and we sometimes do public talks on the lessons — those are the right channels if you're curious without being committed.

The wall isn't yours.

It just has your name on it because you got there first. Send it over — we'll spend a weekend tearing it down with you.