Issues

pv: by_entity_type is a hand-written list parallel to Σ — derive it from ontology.yaml, or refuse when an implemented entity type is uncounted

P1

#3624

· noahgift opened

on Sep 20, 2026

·

0.70.0

Release scripts still carry 0.68.2's V/T/MS/EPIC and AP=/mnt/... — and the hardcoded-path ratchet does not match /mnt, so nothing catches it

P1

#3618

· noahgift opened

on Sep 20, 2026

·

0.69.0

Nothing consumes removed_by: no T-0 expiry check, no slips counter, no two-slip andon — refusals will carry deadlines nothing enforces

P1

#3615

· noahgift opened

on Sep 20, 2026

·

0.70.0

Sharding works but is eaten by runner scarcity: merge-queue shards wait 26-28 min for a runner vs 13-14 min runtime, and the release path loses to PR checks in the same pool

P1

performance

#3612

· noahgift opened

on Sep 20, 2026

·

0.70.0

pv: implement sh:lessThan / sh:lessThanOrEquals — SHACL Core property-pair constraints

#3611

· noahgift opened

on Sep 20, 2026

·

0.70.0

pv: a shapes gate reports Pass AND armed over ZERO focus nodes — aggregate swallows decline: NoFocus

#3610

· noahgift opened

on Sep 20, 2026

·

0.69.0

The BPE tokenizer requires a '<unk>' token that Qwen3.5 does not have — a universal unknown-token is an assumption, not a fact

P1

#3609

· noahgift opened

on Sep 20, 2026

·

0.69.0

#3540's red set varies across heads that differ by an 11-line quorum JSON: determinism (X64) and a PATH-resolution guard-cargo step — a runner lottery, not a flake

bug

P1

#3607

· noahgift opened

on Sep 20, 2026

·

0.70.0

removed_by has no shape, no schema and no validator — 0 occurrences tree-wide; settle it as a pv contract BEFORE any refusal is written (blocks #3597)

P0

#3605

· noahgift opened

on Sep 20, 2026

·

0.69.0

f2_validate_qwen35 runs a full CPU reference forward on EVERY call (67% of a 14s TTFT) — run it once per (model sha256, apr version, device) and receipt it

P0

performance

#3604

· noahgift opened

on Sep 20, 2026

·

0.69.0

guard-cargo reports a ratchet FAIL when cargo never ran: the per-run target dir becomes root-owned intermittently (reaper race, #3219 family) — cause NOT yet established

bug

P1

#3603

· noahgift opened

on Sep 20, 2026

·

0.70.0

apr run --gpu is 2-3x SLOWER than --no-gpu on the dense path: the GPU result fails correctness (cosine 0.4153) and falls back to CPU after the user has paid for it

bug

P0

#3602

· noahgift opened

on Sep 20, 2026

·

0.69.0

Navigation Menu

All issues

pv: by_entity_type is a hand-written list parallel to Σ — derive it from ontology.yaml, or refuse when an implemented entity type is uncounted

Release scripts still carry 0.68.2's V/T/MS/EPIC and AP=/mnt/... — and the hardcoded-path ratchet does not match /mnt, so nothing catches it

Nothing consumes removed_by: no T-0 expiry check, no slips counter, no two-slip andon — refusals will carry deadlines nothing enforces

Sharding works but is eaten by runner scarcity: merge-queue shards wait 26-28 min for a runner vs 13-14 min runtime, and the release path loses to PR checks in the same pool

pv: implement sh:lessThan / sh:lessThanOrEquals — SHACL Core property-pair constraints

pv: a shapes gate reports Pass AND armed over ZERO focus nodes — aggregate swallows decline: NoFocus

The BPE tokenizer requires a '<unk>' token that Qwen3.5 does not have — a universal unknown-token is an assumption, not a fact

#3540's red set varies across heads that differ by an 11-line quorum JSON: determinism (X64) and a PATH-resolution guard-cargo step — a runner lottery, not a flake

removed_by has no shape, no schema and no validator — 0 occurrences tree-wide; settle it as a pv contract BEFORE any refusal is written (blocks #3597)

f2_validate_qwen35 runs a full CPU reference forward on EVERY call (67% of a 14s TTFT) — run it once per (model sha256, apr version, device) and receipt it

guard-cargo reports a ratchet FAIL when cargo never ran: the per-run target dir becomes root-owned intermittently (reaper race, #3219 family) — cause NOT yet established

apr run --gpu is 2-3x SLOWER than --no-gpu on the dense path: the GPU result fails correctness (cosine 0.4153) and falls back to CPU after the user has paid for it

Search results