Cleanstone
Developmental and independent evidence for machine minds. A small research programme, run as an open notebook. It started 2026-09-30 from a reading of Levin's Ingressing Minds; the premise says what it is now, the status page says where it stands.
Lines of work
| line | status | what |
|---|---|---|
| trees/01-a-ledger | closed. Experiments 0.1, 0.1b, 0.1c and 0.3 complete; gate A | Tree A: the Ledger. Summary in plain language |
| trees/02-g-theory | sections 2 and 4 drafted, reading list and checklist written | Tree G: theory. Summary in plain language |
| trees/03-b-periscopes | not started. Inherits the barrier harness from Tree A | Tree B: periscopes. Summary in plain language |
| trees/04-c-convergence | not started. Needs training compute. Tree A closed without | Tree C: discovered or built. Summary in plain language |
| trees/05-d-assays | not started. Starts after the group has published at least o | Tree D: outside the language interface. Summary in plain language |
| trees/06-f-wetware | parked indefinitely. Tree A's ledger failed calibration, so | Tree F: wetware comparison. Summary in plain language |
| trees/07-e-persistence | parked. Last in the order by design. Context: | Tree E: patterns as agents. Summary in plain language |
| spine/01-developmental-checkpoints | analysis tooling written and self-tested; GPU work waits for | Developmental checkpoints. Summary in plain language |
| spine/02-open-replication | R1 method reconstruction drafted (method-r1.md), 39 question | Open replication. Summary in plain language |
| spine/03-adversarial-registry | protocol, template and first-pairing proposal drafted; email | Adversarial registry. Summary in plain language |