v0.8.33 · Help Tab Cleanup + Real Brand Stamp

BP130. Removed Op-Ed Thou-Art tile from Help tab. Real Ed25519 Brand Stamp fingerprint wired (1fde6ef8, first8=1ad73734). SHA-256: E1C617DC81B4E2F4125945C89317BE002D01F2305DB5F437C702F9A9620391BD

v0.8.32 · Diagnostic Mode Remote Hand

BP129 · Diagnostic Mode PROCESS_CONTROL scope ships.

Help tab now has Authorize Access + Run Help Setup buttons. Grant scoped, TTL-bounded, cryptographically-audited permission for cooperative-class Remote Hand.

• Brand Stamp Ed25519 authority verification (fingerprint 1fde6ef8)
• Scoped allowlist: enable_psremoting_and_trust, install_ollama_model, restart_ollama_service
• 4-gate consent modal · TTL selector (15min/1h/24h/until revoke)
• Immutable hash-chained audit log (server-computed, tamper-evident)
• Edge function insert-diagnostic-audit with peer-id enforcement
• RLS lockdown on diagnostic_audit table

Installer SHA-256: EFBE60EE95F89EEB536B5E98D38372BA49CC6781771DED0662727C4E54A18CF0

MnemosyneC v0.9.12

BP137 0.9.12 -- persistent browser authority + browser-as-primary routing: the plow queries the user own logged-in Google/Perplexity (always-on session; sign in via Settings -> Authorities) before the encyclopedia swarm, honest fallthrough when not signed in. Plus the HOT/COLD lift render crash fix. Behind one research_arm flag, default OFF. SHA-256 e06fe1578c5f98adce248d8e81a897bcc340d483adbf63c8237f3415242319a3. Unsigned build.

MnemosyneC v0.9.11

BP137 0.9.11 research arm -- semantic relevance gate (false-serve class closed), compute sidecar (calc questions computed, verify-by-recompute), escalation + Psionic accelerator, browser authority (your own Google/Perplexity via Chrome), distributed star chamber on dont-know. All behind one research_arm flag, default OFF. SHA-256 a55625f7fac31cb72bc8759d1ce62958bea5ae17781591f4bc7ec1fc4b043db6. Unsigned build.

MnemosyneC v0.9.10

BP137: honest substrate answer-cache -- exact-match + semantic (nomic-embed) serve at zero cost; birthday-frame false-serve class closed (Psionic semantic-or-nothing floor); MMLU answer-key contamination hard-gated; Layer2ProveIt fabricated numbers removed for an honest CTA to the live Prove-It panel; Ask input never locks. New feature paths default-OFF. SHA-256 e582698a535746358bce697aaecbb488f977ecae8dbeabdbebfff11efded7899. Unsigned build.

MnemosyneC v0.9.9

MnemosyneC v0.9.9

WAN responder honors each peer's selected model. The mesh responder now answers with the peer's actually-selected local model (via activeBrain / brain_registry) instead of a hardcoded gemma4:12b. This fixes 32GB peers set to gemma2:9b silently running 12b, splitting VRAM, timing out, and abstaining on WAN-dispatched questions -- and it unlocks per-peer model diversity across the cooperative mesh. peer_presence now advertises each peer's real answer model, and the sku-upgrade tier map is de-hardcoded to hardware-aware RAM tiers (2b / 9b / 12b).

Also carries v0.9.8 CleanShell: the lean shell is now the default front door -- six real tabs plus combined Substrate, dead surfaces un-routed, stubs cut.

• SHA-256: 3e4156867c250bb90205277734899a75389ab883d48029672b5cf6bdb2415941
• Size: 483,600,158 bytes (~461.2 MB)

A participation-based cooperative AI peer by Liana Banyan Corporation. Platform credits are not securities. SSPL.

MnemosyneC v0.9.7

BP135 0.9.7 -- plow reasoning-mode (voters show their work; engineering/51 J to C) + argument-ledger (reasoning + verdict + read/write relevance gate). SHA-256 8e42955bc5b4d1da387d695d12fd311a8e682684ca4ea3fded339a41defb66dd, 483676404 bytes.

MnemosyneC v0.9.4

BP134 v0.9.4 -- Hot substrate from install: MnemosyneC now ships the 14-category Chocolate Pack (pre-seeded verified eblets) so the 12-Blade Plow starts WARM on first launch, plus the per-domain question banks for deep field-sowing. Adds a deep-harvest mode, the field-sowing runner, a 14x5=70 MMLU-Pro Prove-It preset with an exportable third-party-verifiable receipt, consent-gated automatic updates (Diagnostic tab, OFF by default), and node-to-node update propagation (peer binaries SHA-256-verified against the canonical manifest before install). Builds on v0.9.3's plow voter think:false fix. Unsigned build. SHA-256: CDD161E0063313015057CC26D9A98C687ED401F1909837EEA758643AD0820557

MnemosyneC v0.9.3

BP134 v0.9.3 -- Plow answers again: fixes the 12-Blade Plow multiple-choice voters returning empty (gemma4:12b is thinking-capable; the voter call now sends think:false top-level so the model emits a letter instead of spending its whole budget on hidden reasoning, which had caused 100% ABSTAIN). Free local AI is now the DEFAULT (no cloud account or API key needed for your first answer). ASK lists your installed Ollama models (gemma2:9b etc.) and actually dispatches to the one you pick. WAN peers on other subnets now classify correctly as WAN, and cooperative peers auto-connect from presence without an invite-token swap. Unsigned build. SHA-256: 960B068B1DE364FD89EC533500D75A45570CA55B537130A0A4D5C3F37878A853

MnemosyneC v0.9.2

MnemosyneC v0.9.2 — plow voter-concurrency fix (the 12-blade plow's MC letter-voters were firing concurrently and OOMing the peer GPU, forcing 100% abstention; now sequential) plus the peer_presence status/state column fix. SHA-256: 43C3884675E0EAE193C26A264C9A5AAA186304A9018D94472441C9874A6BEE91