Files
LithosAnanake/experiments
Robert Allan JamesandClaude Sonnet 5 cb32e6632b
Build / build-amd64-iso (push) Waiting to run
Build / build-aarch64-iso (push) Waiting to run
Build / build-riscv64-img (push) Waiting to run
Root-cause the std79 lockdown gap: likely never engages for any locked identity (FABRIC-3.md §XXV follow-up)
Following up on the messaging-words-not-denied finding: 'MSG-STATUS
ACL-STD79-ALLOWED? .' sent into rajames came back UNKNOWN WORD, not
"denied" -- acl-std79.4th's own supporting words were never compiled
into the VM's dictionary at all.

Root cause, confirmed via the raw boot log: "WARN: block 4998 exceeds
1KB, truncating" fires during every std79-locked identity's own
birth. capsule_exec_payload() parses "Block N" content as everything
from that header to the next "Block N" header or end-of-payload,
truncating at 1024 bytes for both storage and execution.
MINT_RESTRICTED_PERSONALITY is a short (~200 byte) string written
into a zero-padded 4096-byte devblock at mint time; capsule_runcap_
birth() reads back the entire reserved multi-devblock region with no
second "Block N" header anywhere in it to terminate "block 4998"
early, so the parser treats the whole mostly-padding region as one
oversized block.

Confirmed universal, not one identity's quirk: the warning fires
exactly 8 times in a full 9-identity boot -- once per std79-locked
identity (rajames, 00-06; zuse runs natively on Hera, never through
this path) -- and dates back to at least 2026-09-10 in this repo's
own logs, well before this session. The std79 lockdown has likely
never actually engaged, for any of the 8 locked identities, since it
was built.

Not fully closed to the byte: the real content sits at the start of
the oversized block, inside the surviving truncated slice, so
truncating the tail shouldn't by itself stop the head from executing
-- the exact remaining mechanical step (stale leftover data, a
line-boundary artifact, or something else) isn't nailed down yet.

Explicitly not fixed -- documented per Bob's own "stop here, document
it, and continue" call. Security-relevant (an intended lockdown
restricting nothing) and deserves its own deliberate fix with real
test coverage, not a bolt-on to an unrelated change.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01EXieurDfDSsDFdnSyusuWo
2026-09-12 22:59:54 -04:00
..

experiments/

Design of Experiments (DoE) campaign data and analysis for the LithosAnanke kernel.

  • bare_metal/ — the bare-metal DoE experiment: blind full-factorial 2⁴ runs measuring the L8 Jacquard mode selector's effect on the Steady-State Machine across amd64/aarch64/riscv64. See experiments/bare_metal/README.md for the full directory layout (runs/, latest/, analysis/{charts,tables,report}).

Hosted-VM DoE campaigns (factorial/heartbeat/ablation studies) live under docs/working/experiments/ instead — this top-level directory is specifically the kernel/bare-metal side.