All pull requests

Pull requests

• Add opt-in guarded FP64 Gram and diagnostic SVD profiles

  #453·mikkel opened Oct 18, 2026Member·

  ·

• Release 0.10.0: expanded Forge convergence requirements

  #449·martyn opened Oct 17, 2026Member·

  ·

• Adopt ring16 at 6,400 updates, K3P/KA2 without LR horizon cap, longer wall-clock limits; publish verified measurements

  #440·mikkel opened Oct 16, 2026Member·

  ·

• docs: round 2 follow-up results in the what-worked review

  #436·mikkel opened Oct 16, 2026Member·

  ·

• dig2 harmonic-link: what gives BISPF its mode_hold pass (thermal link refuted; chord term ports to BCAP+tether)

  #435·mikkel opened Oct 16, 2026Member·

  ·

• dig2 wallclock-reruns: separate wall-clock hits from recipe failures (3 real verdicts, opt-in NOT MEASURED + contention-aware limit)

  #434·mikkel opened Oct 16, 2026Member·

  ·

• Dig 2 (ring16-cap): 6,400-update ring16 task revision; BCAP/BISPF safe, selected K3P/KA2 still fail (diagnosed)

  #433·mikkel opened Oct 16, 2026Member·

  ·

• Dig2 clock-audit-measured: proposed measured clock audit (v2, needs sign-off) + annealing at ring16 6,400

  #432·mikkel opened Oct 16, 2026Member·

  ·

• Dig2: refresh BISPF and BCAP on the current source (34/34 verdicts reproduce)

  #431·mikkel opened Oct 16, 2026Member·

  ·

• Dig: relative-force moment tether (opt-in) — refuted at stage 1

  #429·mikkel opened Oct 16, 2026Member·

  ·

• dig: cell-quota-min-change re-entry probe (REFUTED on ring16) - opt-in port + readout

  #427·mikkel opened Oct 16, 2026Member·

  ·

• Forge dig: BISPF ablation - tether alone does not explain BISPF's mode_hold lead

  #426·mikkel opened Oct 16, 2026Member·

  ·

• Dig (horizon/clock): opt-in progress-driven schedule; refuted on ring16 at 1,600

  #425·mikkel opened Oct 16, 2026Member·

  ·

• Audit zero-pass Forge tests and recommend keep, modify or remove

  #421·mikkel opened Oct 15, 2026Member·

  ·

• Establish R1/R2 baseline and prepare fixed-task Tier 1 repairs

  #401·martyn opened Oct 12, 2026Member·

  ·

• Record falsified stall-reach (#397) and G-idle (#398) results in experiment memory

  #400·martyn opened Oct 12, 2026Member·

  ·

• BCAP Tier 2: four repair experiments, no global winner

  #395·martyn opened Oct 11, 2026Member·

  ·

• Study balanced transport: width repair, rare and native regressions

  #375·martyn opened Oct 9, 2026Member·

  ·

• Preserve conditional identity and rare density with direction blend plus transport

  #373·martyn opened Oct 9, 2026Member·

  ·

• Research: projection preserves local-v2 rare density but worsens retention

  #368·martyn opened Oct 9, 2026Member·

  ·

• Research: direction alone retains both conditional identity repairs

  #367·martyn opened Oct 9, 2026Member·

  ·

• Research: strict finite progress repairs two conditional BCAP gates

  #361·martyn opened Oct 9, 2026Member·

  ·

• Hydraulic v1 on the direction-blend BCAP winner: retention and precision gains, no gate repaired (stop)

  #360·martyn opened Oct 9, 2026Member·

  ·

• Research: transport rare-density PASS and finite-descent regressions

  #358·martyn opened Oct 9, 2026Member·

  ·

• Investigate magnitude-sensitive G/D updates for Gaussian stability

  #320·martyn opened Oct 7, 2026Member·

  ·