What Vacant is
Vacant is an accountability layer wrapped around your AI assistant. It doesn't replace your AI. It runs the acceptance tests you wrote, delivers only on a pass, and leaves a signed receipt for every step that you can recompute yourself.
#In one sentence
The AI says "done." Saying it's done doesn't count. Running it does.
Vacant is a layer wrapped around your AI assistant. You hand it a requirement and its acceptance tests; it has your AI do the work, then runs your tests once — on a pass, the work reaches you; on a fail, it is stopped at the gate. Every step of the process is signed; change one cell and the chain breaks at that cell.
#What it does, and what it doesn't
| What it does | Decides delivery with the acceptance tests you wrote; signs each verdict into a receipt chain you can recompute yourself; makes participants who did bad work less likely to get the next job. |
| What it doesn't | It doesn't make the AI smarter. It doesn't guarantee that this AI is reliable — a level belongs to one run, not to one AI. It doesn't grade work that can't be written as a test. |
#Why "accountable" and not "trust"
The project's definition changed five times, and each time the claim got smaller. The last version settled on:
A delivery without a verifiable receipt is not accepted.
This is an intake point, not an interceptor. The difference is spelled out in chapter four: a "refuse" verdict is not the same as "impossible to publish." The five-level receipt vocabulary explains how far up those levels we can honestly claim to reach.
#Two shapes
The shape that owns generation (the main line of this research, and also the acceptance gate):
Acceptance gate (CONFORM) — runs your visible tests and delivers the first candidate that passes; if none of five passes, it refuses delivery. Revision loop (H-MIX) — on a fail, it doesn't switch to someone else; it pastes the failure text back so the same attempt can fix it. Signed receipt chain — every cell is signed and locks the previous one; anyone with the public key can re-verify it offline.
The shape that does not own generation (the v0.9.0 product line):
Zero-config — works as soon as it's installed, with no configuration. It only sends one message back at the moment the agent says "I'm done": no executable signal, no power to retry, no power to refuse delivery.
The full record of zero-config's four batches
The other new product line is accountable trace: when something goes wrong, it traces back to the step that caused the error, instead of only saying this delivery failed.
#Where the evidence stands
58 public claims, each with a verdict. 34 of them come from the project's official index; 24 were transcribed one by one from the Vacant repo's decision documents after 2026-09-25 and are not yet in the official index (marked "repo verdict" in the ledger). Refuted claims stay too — they are statements we withdrew, not things that never happened. The full ledger is in the evidence ledger.
#How to read this site
- To see how it works → Architecture, seven layers from cryptographic identity to hooking into your assistant.
- To run it once yourself → Install and Quickstart.
- To see what we measured → the gain axis is where we spent the most effort, and also where we refuted ourselves the most.
- To see what happens once it's installed → Zero-config, including the batches that measured no difference.
- To see how errors get traced → Accountable trace.
- To see the full record of refutations → Lab notes: 32 posts, each one an experiment that really ran or a claim that was refuted.
