About & maintenance

We Can Just Build Things is a verified, values-enforced catalog and guided build flow for shipping Nostr, AT Protocol, and general open-source tools with AI agents. It exists to give the person defining what their community needs a way to deliver a professional-grade product without first having to become a developer — and to make the result accountable to the people who depend on it.

It is built to be shared infrastructure, not a walled garden. The methodology is open, the catalog is open, and the enforcement engine is open.

How it stays current

Maintenance runs in three layers, increasing in cost as you go up. The whole design assumes humans are scarce and automation is cheap — anything checkable by tooling is checked by tooling.

Automated, weekly, free

A scheduled GitHub Action runs every Monday and:

• license-watcher hashes each upstream LICENSE, compares it to the recorded license + commit, and scans recent commits for relicensing keywords (relicens, now under, BSL, SSPL, …) — three independent signals, so it catches a quiet relicense before it’s a surprise.
• maintenance-checker pings each entry’s release feed and reclassifies status (active < 90d, minimal < 12mo, dormant < 36mo, abandoned beyond).
• The three-layer enforcement engine re-runs against the full tree, so a transitive dependency that newly pulls in an excluded provider is caught even when no manifest changed.

Each opens a PR with the proposed diff. No human is needed until the PR exists.

Human review, weekly, low effort

A maintainer glances at the bot PRs — typically ~15 minutes a week. New-tool contributions arrive via issue template and CMS, both of which run the enforcement engine before a PR is mergeable, so review is mostly “does this fit, is the description accurate” rather than “is this clean.” Review SLA: 7 business days to first response.

Periodic, larger, scheduled

Twice a year, a full re-verification pass refreshes every license commit SHA, revalidates primary-source links, and regenerates adoption metadata. Because the tooling does the heavy lifting, this is a half-day of reviewing diffs, not an audit from scratch.

The honest part

A catalog of this scale maintained by hand would collapse within six months — every comparable directory that’s tried has. This one works only because the verification is automated. Humans spend their attention on what tooling genuinely can’t decide: editorial judgment, new-category structure, and contested exclusions.

Designed to outlast its first maintainer

If the original maintainer disappears on day fifteen, someone else can fork the repository, run the same workflows, and the catalog keeps verifying itself. The access-transfer checklist in the repo covers moving the hosting, DNS, and analytics ownership cleanly.

Lineage and attribution

• The build flow adapts the open PIE Cookbook from the Portland Incubator Experiment.
• The catalog is grounded in what the And Other Stuff collective actually uses to ship freedom-tech projects.
• The provider-lockdown recipe model is anchored to Shakespeare’s BYOK design.

Want to help?

See Contribute to add a tool, fix an entry, or adapt the flow for your own program.