Mark's Reports

Live pull from your Google Calendar — Apr 27 through Jun 10. 38 events: Renaissance prep this week, Spirit Day + 8th exams week of May 4, Chattanooga May 12-13, your 6th SS exam May 15, T.I.S. checkout May 28, Orlando run May 30-31, DC trip Jun 1-5. Color-tagged. Read top-down for the whole school year-end.

G16 review of Apex Claude Code's build report (next card down). Code verified against the report — it matches reality. Two products, deliberately different. Three concrete cross-pollination ideas, ranked. The biggest one to backport: the anchor-footer prompt pattern (## Brief at top + ## How to use this at bottom owns the recency slot, ~5-10% follow-up accuracy lift).

Apex Claude Code's build report. Forked Bidet to a separate public BYOK app. Two PRs shipped: PR 1 (Groq Whisper, BYOK LLM, Wake Lock, IndexedDB resilience, research-backed AI prompt). PR 2 (architecture pivot from Supabase to browser-only, custom user-defined tabs). Domain bidetai.app + email routing live. Personal Bidet untouched. Ready for local smoke test; deploy is your call.

Heavy ship day yesterday. Stack health (TP3 657K rows, all 8 containers healthy, disk 37 GB free). What shipped (Legacy v3, Bidet GPU, processor.py reorder, OMI bidirectional, 32 calendar events). Apex disk crisis explained + resolved. This week's calendar. What needs your hands today.

Deep dive on OMI's April 9 + April 26 releases. Claude 4.6 across all routes, BYOK transcription, voice replies, MCP bidirectional memory, Apple Reminders sync, 3-day battery. Mapped against your Oracle vision. What auto-rolled vs what needs your hands (~10 min). 8 ranked action items. What to skip.

One-time setup for routing OMI's transcription through your own Deepgram / OpenAI / AssemblyAI key instead of OMI's default cloud. Eliminates the silent-fallback path that drained the $60/mo Gemini budget last time. Provider comparison (recommend Deepgram Nova-2 at ~$0.0043/min, $200 free credit), step-by-step setup, $15/mo spend-cap recommendation.

Stream A rewritten — hand-painted Memorial Pearls only, weight-tiered $250-$1,295, 9-lb cremains cap. Stream B unchanged in structure, raised to $375-$995. Line 3 reframed as Community Composting (your bagged-soil idea: pet owners opt into communal vessel with shelter animals, all proceeds donate to shelters). Line 4 = academic outreach, no revenue. Two money-making lines + two community-aligned lines.

Per-tier unit economics — Stream A blended ~90% margin, Stream B ~86.5%. Year 1-3 revenue projections at conservative volumes ($39K Y1 → $186K Y3). Line 3 modeled as pure pass-through to shelters. Capital + breakeven + open numbers list (marketing budget, operator draw, insurance).

A separate distributable Bidet AI, forked from the personal Apex install into MrB-Ed/bidetai-app (private). PR 1: Groq Whisper swap + BYOK LLM + Wake Lock + IndexedDB recording resilience + research-backed AI tab format (long context first, structured directives last — vendor consensus). PR 2: pivot from Supabase to browser-only — IndexedDB stores keys + custom tabs + session history, backend stateless, three-PR plan collapsed into one. Domain wired, Cloudflare Email Routing live (*@bidetai.app → gmail). Two pivots called out honestly: chunked recording was a rabbit hole; Supabase was overengineering. Ready to test locally; deploy to Cloudflare Pages + Cloud Run is your call.

The 02:04 ET brain dump (drop the cement line, pivot to pearls, lock cedar box as "the curing vessel") is transcribed, cleaned, and analyzed. Inside: full text, eight proposed questions for the v3 master plan, an honest fabrication warning about the FORAI output, and the bidet GPU work status (CPU mode live, GPU 90% wired — one missing CUDA library blocks the flip). 02:15 short fragment also processed — ~13s, mostly noise.

What today's rebuild was, what you actually get from it, what's still pending, and where we go from here. Plus tonight's Tailscale-was-down-not-Apex lesson. Posted ~9:30 PM after live verification of the Apex stack.

Earlier today: Phase 0 status, DC trip Westin notes, granulator paths, AI Radar permissions, three Q&A. Companion to the recovery report.

Three actions this week, ten ranked candidates, seven cut. Top pick: ollama pull gemma4:e4b — 5 GB RAM, multimodal, free, local. Watch this: Opus 4.7's new tokenizer eats up to 35% more tokens at the same list price — your bill may be climbing. Privacy filter open-weight, 50M active params, scrubs PII before embedding. Permissions fix worked; first full successful scan in three weeks.

Phase 0 closed tonight: backup + tag + parity check + container reachability + Docker auto-update lock all live-verified. Phases 2, 3, 4, 5 already shipped (Apex commits 8bcdb62 + 68ec619). Phase 6 explicitly extended with MCP server containerization (commit 8a18da9). Open: Phase 1 (re-embed ~3,500 Gemini rows), Phase 6 finish (MCP container + Cloudflare Worker pinger replacing healthchecks.io), Phase 7 soak (24h clean-run + 72h parallel observation).

Three independent agents reviewed the stack - clean-slate architect, post-mortem engineer, external research. Verdict: your direction is right, the plumbing is wrong. Keep the 646K rows, rebuild the operator layer. Meta's Limitless acquisition + Zuck's two 2026 projects confirm the category is real; you're on the frontier, not behind. Three paths forward with tradeoffs + recommendation inside.