Handoff documents were additive. HANDOFF-2026-08-19 opened by saying the 18 AUG document still applied in full and added to it. After ten sessions a new assistant would face ten documents to read in date order and diff mentally to work out what is currently true. That cost grows every session and none of it is necessary. docs/HANDOFF.md is now the only handoff, rewritten in place each session. It is state, not a log. The dated ones move to docs/archive/ and stop being required reading. It is standalone: everything still true from both is carried forward. Section 1 is invocation, stated as facts rather than demonstrated in examples. That is the other half of the problem. runuser appeared only inside example commands, so it could be learned by pattern matching but not by reading, which fails exactly when an assistant composes a command from scratch. That is what happened, and it is F-035: su cannot run as a nologin service user, both commands returned the same message before touching anything, and the output read as a broken repository when the tree was clean and the suite passed. The F-027 class again. Also stated as facts: bash tools/ not ./tools/, all repository operations as mechcomp, Gitea SSH on 42022, pct push then chown, explicit timeouts, journalctl not /var/log, systemd-run for long jobs, and assert the guest is running before interpreting any pct exec result. Not done: the same facts should be cross referenced from PROCESS.md. I no longer had that file in view and would not patch a document I cannot see.
Mechanical Compiler
Build the capacity to construct real structures from reclaimed and commodity materials, using whatever fabrication is actually to hand — 3D printing, tabletop CNC, welding, cement casting, COTS stock.
The reference case is a faceted timber shell: planar panels meeting along straight fold lines, converging on nodes, sitting on a platform. Buildable without a factory, provided someone has worked out what the pieces are and how they meet. That last clause is the project. The Mechanical Compiler exists to make the pieces computable, qualifiable, and reproducible by someone who was not present when they were designed.
Scope
In scope: geometry, qualification, and reproducibility of structural members and their interfaces.
Not in scope: building codes, permitting, jurisdictional approval. Qualification and compliance are different things. A qualification says this member is what it claims to be, made this way, from this stock. Compliance is a jurisdiction-specific argument someone else may build on top.
The project records physical claims, never verdicts — section modulus, moment of inertia, material provenance, process parameters. Those are what any future argument would need. A pass/fail verdict would bake in a jurisdiction we do not want to be bound to. Measure and attest; never adjudicate.
Repository layout
docs/ specification, state, failure log, roadmap
legacy/openscad/ rev 8.0.0 generators — reference, not a live target
fixtures/ frozen acceptance oracles
tools/reference-toolchain/ pinned OpenSCAD + BOSL2, build-time only
src/mechcomp/ the application
tests/ acceptance against the oracle
Read docs/ROADMAP.md first for what this is, then docs/ENVIRONMENT.md for
how an instance is built. If you are writing provisioning automation, read
docs/FAILURES.md before the specification — every entry is something a
script written from the specification alone would have got wrong.
Getting started
make deps # virtualenv, base + CAD requirements
make verify-oracle # confirm the frozen oracle is intact
make test # pytest -n auto
make verify-oracle needs nothing but Python. It should pass on any machine at
any time; if it does not, stop.
The oracle
fixtures/strap-beam-8.0.0/ holds 123 frozen cases — 113 accepted, 10
rejected — produced by OpenSCAD 2021.01 with BOSL2 at 92d697c2. It is the
acceptance criterion for any reimplementation of the generators.
The ten rejected cases are part of the contract. A port that accepts them is wrong, however good its numbers look elsewhere. It is easy to reproduce the geometry and quietly lose the constraint that made it trustworthy.
The generators under legacy/openscad/ are the reference implementation, kept
so the oracle can be regenerated. They are not a live target and the running
application has no OpenSCAD dependency.
Provenance
This project's documents, code and roadmap are LLM-generated under human direction. That is stated plainly here and in any downstream submission. We do not obscure it.
The maintainer reviews and owns every line. Anything that could not be defended in a review thread does not ship.
Licence
AGPL-3.0-or-later. See LICENSE.
Section 13 obliges us to offer source to users interacting over a network, so any deployed web tier carries a visible link back to this repository. That is a licence obligation, not a courtesy.