Board
Inbox
0
No tasks
Next Action
200
Call Copilot system verification
@Phone
Review and approve weekly task triage batches (reassign, kill, push dates, defer)
@Computer
Tighten Cerberus pre-commit hook regex — exclude method-call assignments like `api_key = self.secrets.get_api_key()`
@Computer
Talk to Nicole about positioning Felipe as the linchpin
@Nicole
MCP grep_codebase bug: timed_out variable error on large files (adapters.py, engine.py, review.py)
@Computer
Confirm auto-tagging suggestions for 5 GTD tasks
@Computer
Rotate ALL secrets.json credentials (exposed in Claude Code API logs) + clean GEMINI-INTEGRATION-PROMPT.md
@Computer
Build unified daily research digest — PA iMessage delivery with LLM analysis
@Computer
Build iMessage vision capabilities - image/video processing integration
@Computer
Build demo pre-flight checklist and failover protocol
Call Sue Bishop
@Phone
Create TCC Permission Reference Card for non-technical operators
Complete relationship DB batch recording
@Computer
Review and approve weekly task triage batches
@Computer
Marketing: Incorporate Cold Fusion video intel into Project Kangaroo positioning (Meta safety chief story, competitive landscape, UHNW trust narrative)
@Computer
Reach out to Jennifer Bloomer and Salvo for advice
@Phone
Investigate and fix funnel service downtime - 3 failures, no response
@Computer
Check Notion credits dashboard to review custom agent usage and plan for future costs
@Computer
Peter + Nicole: Get marriage counseling materials
@Computer · @Nicole
Figure out how to sign into Microsoft Excel with
@Computer
Access and review LIFE MGMT shared drive that Nicole added me to
@Computer
Review and potentially update custom agent creation permissions in Notion workspace
@Computer
Stabilize Nicole's current deployment
Local model on MacBook Pro (MLX inference)
Resolve wedding tickets + cat issue
@Computer · @Phone
Talk to dad more about Will Carroll
@Phone
Update dev-01 & nicole-1 macOS for MYTHOS PREP
@Computer
Follow up on Spencer's hiring decisions and company equity structure
@Phone
Complete full dependency audit on dev-01
@Computer
Fix ghost tasks in SQLite
@Computer
Review Google security gaps including suspicious sign-ins and missing 2SV
@Computer
Follow up with Spencer about creating two versions of pitch deck
@Phone
Review and approve weekly triage batches (7 batches, 53 total tasks)
@Computer
Consider scheduling free Notion consultant session for workspace optimization
@Computer
Fill out and return incorporation questionnaire from Orrick
@Computer
Verify auto-tagging suggestions for 5 GTD tasks
@Computer
1031 Exchange — establish strategy, research Taylor home docs, engage lawyer, sign & submit paperwork
@Anywhere
Review and approve weekly triage batches
@Computer
Purchase PD dress shoes (black)
@Errands
Purchase stools for kitties (window)
@Errands
Purchase deodorizer
@Errands
Purchase baking soda
@Errands
Coordinate move in date @ PY with broker
@Phone
Respond to Shopping del Sol calendar invitation
@Computer
Attend photographer meeting with Peter Donovan
@Anywhere · @Computer
Migrate secrets.json to macOS Keychain (encrypted at rest) — before Nicole deployment
@Computer
Evaluate Notion consultation session offer and decide whether to schedule
@Computer
Add credits to Claude API billing account for ReimertDonovan organization
@Computer
Forge: Design + build panel selector — changeset→panel routing, min_findings calibration, expert-level selection. Full handoff: FORGE-PANEL-SELECTOR-HANDOFF.md
@Computer
Follow up: Debora Bon NDA - awaiting signed copy via Erik
@Computer · @Phone
Back up dev-01 critical files to Google Drive
@Computer
Register for Notion Developer Platform virtual event on May 13
@Computer
Review Orrick Contract Generator terms and conditions
@Computer
Review Notion Custom Agents dashboard to see what agents are running
@Computer
Follow up on Anthropic support re: usage bundle availability for Team plans
@Computer · @Phone
Enter all Hazel rules in UI and test with dummy files
@Computer
Review new Workspace Intelligence admin settings in Google Admin console
@Computer
Set permissions and expectations for who can create Custom Agents
@Computer
Create PND folders 07–09 and add sub-entity Hazel rules
@Computer
Fix voice message deletion capability - enable OpenClaw to delete voice messages when requested
@Development
Explore and implement Claude's visualization tools for pitches and Nicole's family
@Computer
Order Verizon Fios 2 Gig
@Computer · @Phone
Call Jeff Faunce about stock portfolio
@Phone
Finish Orrick entity survey
@Computer
Handoff + communication standards — seamless async transitions (docs, task quality, iMessage capture) so no meetings or catch-ups needed across continents
@Computer
Create new client profile for Karl Reimert — individual farm operator and portfolio, first beta tester for OpenClaw with complex financial decisions
@Computer
Review metadata suggestions for 5 GTD tasks (energy/context/area classifications)
@Computer
Research foreign holding company structure timing and cost (Bahamas/Bermuda) for Paraguayan investor flow
@Computer
Complete master budget
@Computer
Source Antonio De Simone painting of steam yacht Iolanda — check Bonhams, Christie's, Invaluable, 1stDibs, artnet. Set alerts. Budget $2,500-$5,000. Also look for copy of 'Happenings aboard the Iolanda' (Moses Taylor, 1927, ~20 copies printed)
@Computer
Review Orrick's engagement proposal and decide whether to proceed with full relationship or explore alternatives
@Computer · @Nicole
Visit the Met — Bishop Jade Collection exhibition (Gallery 222). Closes June 28, 2026
@Errands
Generate codebase migration prompt queue — ~22 prompts to bring all pipelines,
@Computer
Evaluate Starlink for Paraguay — Nicole ask Karl how much it costs, check coverage/reliability for OpenClaw deployment
@Nicole
Print high-quality reproduction of De Simone Iolanda painting for office — giclée on canvas, period frame
@Computer
Create Pitch Assets view in Reference DB — filtered view for all competitive intelligence, pitch quotes, trust-gap evidence, and sales ammunition
@Computer
Curate full office art & heritage display — compile list of all pieces, source/print/frame each, install before Paraguay move
@Computer
Recalibrate GTD task duration estimates for autonomous workflow — bulk update Duration field on tasks sized before Session 33
@Computer
Fix ghost tasks in SQLite — 8+ IDs appear in query but return not found on delete. Data integrity issue from Notion migration.
@Computer
Research YPO (Young Presidents Organization) equivalent — current requirements, modern alternatives. Discuss with Austin Wermers.
@Computer
Paraguay move prep — deliverables list, schedule for getting everything buttoned up, audit electronics for 220V compatibility
@Nicole
UMBRELLA: Close gap between OpenClaw vision (Gemini conversation) and built reality — 8 features (GTD-1239 through 1246)
@Computer
Nicole: Home office setup — get uncle's Cowboys jersey framed, get Cowboys jerseys for both of us, countdown display for office
@Nicole
Show Nicole existing syndication structures — Jerry Virid agreement, Gleeberman agreement, syndication reference books
@Nicole
Get syndication agreement from Austin Wermers (son of Tom Wermers, ran Wermers Multi-Family acquisitions)
@Phone
Write operational security SOPs with Nicole — document best practices for comms, data handling, device security
@Nicole
Talk to Meg (quant trader, commodities, Chicago) — how Paraguayan beef tracks CME/global beef market, futures, swaps, derivatives
@Phone
Visit Meg in Chicago — follow-on from commodities conversation, deepen relationship (add Meg to Relationship DB)
@Phone
Felipe strategy — make him integral to PY UHNW scaling + plan for Libertad FC exit (Cartes reputational risk)
@Nicole
Personal workflow + GTD optimization — voice memo format, weekly review cadence, time blocking, task categorization, estimation accuracy review
@Computer
Complete master budget analysis
@Computer
Fundraising strategy — evaluate routes, determine amount/timeline, develop comprehensive plan
@Computer
Autonomous agent architecture — define workflows, autonomy levels, permissions, overnight execution config
@Computer
Reach out to Salvo (cousin's husband, sold startup to Google) + Jennifer Bloomer for advice on device/product
@Phone
Nicole handoff doc — ventures, OpenClaw status, GTD system, pipelines, her role, venture structure, contribution model
@Computer
Set up shared calendar of upcoming important dates with Nicole — birthdays, deadlines, travel, events
@Nicole
Signal vs noise filtering — ensure PA system and pipelines focus on actionable signal, prune noise from data flows
@Computer
Single-channel architecture — one source of truth capturing all client data, replicable across deployments
@Computer
Nicole: develop fundraising and buy-out plan as last resort option re Beethoven's Financial Services
@Agenda
Scope creep review — audit all planned features against critical path, cut anything not needed for first paying client
@Computer
Access and review LIFE MGMT shared drive that Nicole added me to
@Computer
Prospect research — identify orgs that could be good potential OpenClaw clients (UHNW families, family offices, firms with privacy needs)
@Computer
Personal relationship DB brain dump — voice memo download of all key relationships, context, and history
@Computer
BUILD: Comms Intelligence Prompt 1 — Email classification upgrade + scheduling decision pipeline + iMessage reply handler
@Computer
BUILD: Comms Intelligence Prompt 2 — Follow-Up Sequencer (outbound tracking + nudge timer + iMessage nudge handler)
@Computer
Call Copilot: verify Prompt 3 committed, check debrief_engine.py + web routes deployed, confirm git tag call-copilot-v1.0
@Computer
Call Copilot: client-quality UX — iMessage /call commands, web UI start/stop, auto-debrief on stop, calendar-aware prompts, audio routing setup
@Computer
Run outcome_capture.py --summary against tonight's 4 deployments (runner-queue-now + aegis-queue)
@Computer
Master file: add ZERO-TO-DEPTH DESIGN governance rule + Learning 273 (developer-first design pattern)
@Computer
Web UI multi-user auth — session cookies, login page, per-user routing (Nicole/Erik/Karl)
@Computer
Expand review-architecture.py into generic review.py with configurable expert panels
@Computer
Purchase + configure Health Auto Export iOS app — set up iCloud Drive auto-sync to Mac Mini (prerequisite for Phase 3 biometric ingestion)
@Phone
Add 5 new panels to review.py build (integration, supply-chain, data-flow, test-quality, documentation) + OWASP ASI-10 expert
@Computer
Build GOAT Advisory Panel agent swarm — Kimi K2.5 approach, 27 personas, 7 panels, router, learning loop. See HANDOFF-1-AGENT-SWARM-BUILD.md
@Computer
Expand GOAT Advisory Panel — add personas for tax, sales/UHNW psychology, operations, product/UX, regulatory, family dynamics, competitive intel, rhetoric. See HANDOFF-2-GOAT-PANEL-EXPANSION.md
@Computer
Monitor Glasswing/Mythos disclosure timeline — track when full vuln report publishes (expected early July 2026). Critical patch window for both Mac Minis
@Computer
Upload v36.0 master file to project settings + SCP corrected version to dev-01
@Computer
Investigate Sessions 59-61 handoff gap — master file jumps 58→62, locate handoffs if they exist
Enable FileVault on dev-01 via SSH (6 steps — fdesetup enable, save recovery key, test SSH unlock on reboot)
@Computer
Commit uncommitted web/app.py on dev-01 and push (update_master_file MCP tool)
@Computer
Create benchmark docs for llm-security panel (planted issues for calibration)
Create benchmark docs for operational-readiness panel (planted issues for calibration)
Update master file v36.0 → v37.0 (expert expansion 77 experts, comms rename, learnings Z+AA)
Pickup: NAS (Synology DS224+ with drives), MagSafe mount, spare Cat6 cables, UPS
@Errands
Buy: Synology DS423+ NAS + 2x 8TB WD Red Plus drives for backup infrastructure
@Errands
Research + buy UPS battery backup for router + switch + dev-01 + NAS
@Computer
FIX: Master file sync pipeline — remove LLM drafting, run splitter-only after advisory chat pushes via update_master_file MCP
@Computer
Deploy 15 legal-domain panel YAMLs to ~/.openclaw/reviews/panels/ (from Project handoff chat, never transferred)
@Computer
Archive abandoned agent framework code (agent_loader.py, agents/ dir, MCP agent tools) — replaced by Nervous System
@Computer
DESIGN: Nervous System architecture — formalize Session 57B design into doc + swarm review (5 layers: Sensor/Pattern/Reasoning/Memory/Attention)
@Computer
Archive deployed Comms Intel Phase 2 mega prompt to prompts/to-review/03-archive/
@Computer
Design principal dependency graph + proactive monitoring system (Sensor Mesh core requirement — fire prevention not fire fighting)
@Computer
Swarm review: Autonomous Vendor Support Agent design doc (Playbook → Playwright → Vendor Chat)
@Computer
Nicole: Review Peter psyche profile v1, add independent observations + corrections
@Computer
WhatsApp cosmetic: fix double-logging (plist stdout + appendFileSync both write to same log file)
@Computer
PA Engine outbound messaging: send_message + lookup_contact tools, WhatsApp send_poller adapter, classify prompt routing, review tier default
@Computer
Build YouTube Content Pipeline (Phase 1+2): yt-dlp extract + Claude API analysis + Memory Graph ingestor
@Computer
Install MLX + mlx-lm on laptop (M4 Max 128GB) for local inference
@Computer
Fix PA engine feedback loop: system iMessages re-ingested + fabricated responses
@Computer
Governance Mechanical Gates — audit 30+ behavioral rules, design code-based replacements (Cerberus heads, SQLite gates, swarm panels). Pre-Eisel.
@Computer
MCP root cause fix — deploy async handler prompt (mcp-async-handler-fix.md), verify event loop no longer blocks, stress test 50+ calls with 0 errors
@Computer
Build Founder's Lens review panel for review.py — non-technical questions only (Session 1 Peter test, Karl Spanish test, principal awareness test)
@Computer
Capture Nicole's unfiltered UX reactions to all user-facing workflows BEFORE any technical onboarding — this data has a shelf life
@Nicole
RID Network Download — incremental capture (low-energy / travel task)
@Anywhere · @Phone
Get wedding guest list from Viviana for RID capture
@Phone
@Nicole: Brand execution for 'don't trust us, trust yourself' positioning
@Agenda · @Nicole
Run FINRA/SEC check on Next Round Capital Partners + Ken Smythe via Cowork
@Computer
GOVERNANCE AUDIT: Advisory chat fabrication + automated governance pipeline health check — what is actually running vs. broken vs. never built
@Computer
Verify Morning Briefing flags opinion/POV content distinctly from reported journalism
@Computer
Research how major publications label opinion pieces — build detection rules for OpenClaw news intake
@Computer
Build knowledge base inbox auto-processor (launchd + Claude Code)
@Computer
Add Learning: Auditable reasoning traces on all automated judgments (anti-drift, evidence chains, gaps disclosure)
@Computer
Idea Gauntlet + design: Citation database (doc store + citation index + hover-click viewer UI)
@Computer
Text Spencer remaining questions from April 21 call (Jensen Huang NVIDIA chip export thesis, plus other)
@Phone
Congratulate Spencer on engagement via text (relationship maintenance)
@Phone
Move Spencer April 2026 analysis files from RID staging to spencer/sources on dev-01 (3 files: arc comparison, wire check, RID update)
@Computer
Build source credibility layer into RID -- domain-specific veracity scoring (observed vs self-reported, 7 credibility domains, verification status tracking)
@Computer
MYTHOS/GLASSWING security hardening sprint — full stack audit, patching, watchdog, CVE monitoring, pre-wedding/pre-move patch days
@Computer
Back up dev-01 critical files to Google Drive (temp until NAS operational) — openclaw.db, data/, config.json, secrets.json
@Computer
Forge Rebuild: Write engine problem statement for design pipeline input
@Computer
Forge Rebuild: Run design pipeline (5-panel swarm + design-generator + contradiction resolution)
@Computer
Forge Rebuild: Build new Forge from design-generator spec via Claude Code (Opus 4.7 max effort)
@Computer
Forge Rebuild: Layer 2 verification — swarm review on rebuilt code (arch + security + failure-mode + data-flow)
@Computer
Forge Rebuild: Fix Layer 2 findings + wire calibration suite to new Forge (fix caveats 2 and 3)
@Computer
DIL: Fix v1 design doc findings, produce v2, re-review (after GTD-1556 synthesis)
@Computer
DIL: Chunk-size validation experiment (250/500/750/1000 words)
@Computer
DIL: Human review cost model — measure failure rate
DIL: Define automation complacency spot-check protocol
DIL: Orrick legal review — verification language + liability + GDPR
DIL: Build golden-set validation harness (5-10 annotated docs)
@Computer
Design full-pipeline provenance traceability for Maybach (panel→prompt→execution→outcome)
@Computer
Maybach design req: model_source end-to-end in pipeline output for monoculture audit
@Computer
Assemble DIL calibration corpus from Peter's annotated high-stakes documents
@Computer
Add adversarial test cases to DIL calibration corpus
@Computer
Build non-Claude adversarial red-team expert YAML panel (Gemini-backed)
@Computer
Stand up Karl weekly 5-minute demo ritual (first external forcing function)
@Anywhere
Identify and engage external factory auditor (H1 residual mitigation)
@Anywhere
Create DIL error log template and pre-commit to logging first 5-10 errors at full fidelity
@Computer
Save BRIEF-PROMPT template to dev-01 for transcript consumption workflow
@Computer
Write STUDY-PLAN.md papers-driven curriculum on dev-01
@Computer
Add stochastic stability test to Gate 3 graduation criteria (3 runs, ≤15pp variance)
@Computer
Add "clean" calibration examples to panel template (false positive training)
@Computer
Add false positive ratio tracking to review.py benchmark reports
@Computer
Supplement benchmarks with non-Claude-authored planted defects (H3 decontamination)
@Computer
Build Analytical Red Team panel (confident-wrong-output testing) with research-backed design
@Computer
Build Principal Alignment panel (Founder's Lens / Invisibility Principle)
@Computer
Add runbook schema to Cerberus prompt-preflight (currently only validates gauntlet-format docs)
@Computer
Step 4: review.py runtime calibration injection — inject panel calibration into each expert system message (merges Gate 2 finding #2 + Claude Code backlog)
@Computer
Step 4: Fix input artifact ambiguity — expert prompts assume "design document" but review.py accepts any input type (Gate 2 finding #3)
@Computer
Step 4: Fix lane-boundary coverage gaps — expert exclusion lists create orphaned domains no expert covers (Gate 2 finding #4)
@Computer
Step 4: Audit compare-to citation hygiene across all 21 panels — verify no fabricated/future-dated references (Gate 2 finding #5)
@Computer
Step 4: Fix cross-panel coverage gaps — database schema + subprocess lifecycle have no primary expert owner (Gate 2 finding #6)
@Computer
Step 4: review.py synthesis arbitration — add shared finding category taxonomy for cross-expert aggregation (Claude Code backlog)
@Computer
Step 4: Add per-expert model field to panel YAML schema for heterogeneous deployment (Claude Code backlog)
@Computer
Step 4: Document rebuttal_round: false compensating-control rationale in PANEL-GRADUATION-CRITERIA.md (Claude Code backlog)
@Computer
Step 4: Review panel-validator.py 5-expert cap — policy decision on 6th expert slot
@Computer
Step 4: Update PANEL-GRADUATION-CRITERIA.md — input-type-aware Gate 2 stopping rules + signal/noise triage
@Computer
Graduate state-of-the-art panel (Step 3) — blinded dual-model rewrite + Gates 1-4
@Computer
Bulk panel graduation (Step 4) — 20 remaining panels to GOAT standard
@Computer
benchmark-builder two-pass write architecture — separate doc generation from defect planting
@Computer
Waiting For
41
Remove duplicate OpenClaw MCP connectors from claude.ai (keep 1 of 3)
@Computer
Plan Venture 2 strategy and outline how Nicole can lead launch
@Computer · @Nicole
Review meeting recording consent language for AI processing coverage — Springmeyer task, needs legal review before any recorded client sessions
@Phone
Buy PD's suit for the wedding — being done in Paraguay
@Errands
Submit incorporation paperwork to Orrick
@Computer
Discuss with Nicole whether ready to engage Orrick as ongoing outside counsel vs just formation work
@Nicole
Verify INPET valuation at 50M
@Anywhere
Entity formation — incorporation checklist in progress with Orrick
@Computer
Erik Mac Mini provisioning — beta scope only, blocked on Nicole boundary convo
@Computer
Review USCIS letters from Dad — update immigration timeline, determine next steps for Nicole's travel permit and Green Card
@Computer · @Nicole
Call Jeff Faunce about stock portfolio status
@Phone
Nicole: Family Office Onboarding/Setup — define engagement and onboarding approach for Viviana, Silvia, Felipe. Map each family's structure, needs, decision-makers, and onboarding sequence.
@Nicole
Ask Alex Macbain about cross-border entity structure
@Computer · @Phone
Nicole: Incorporate all spreadsheets and financial data into PSB — consolidate existing family data, valuations, and records into a single organized structure
@Computer
Client-facing governance + transparency doc — 'How Your AI Thinks' rules in plain English + domain triggers + verification points
@Computer
Work through Paraguay Move Gauntlet — resolve all NEEDS INFO items before departure
@Computer
Follow up with Nicole’s Travel Permit
@Nicole · @Phone
Determine visa status and Nicole's Green Card options
@Computer
Follow up with Karl Reimert to calculate cow numbers
@Agenda · @Phone
Design two-speed local inference architecture — fast model (7B-14B) for conversational turns, deep model (70B-405B) for complex analysis, cloud for real-time data only
@Computer
Design local RAG layer — vector database + embedding model + ingestion pipeline + retrieval step in hybrid router
@Computer
Build local-only mode toggle (inference.mode config key) — hybrid router bypasses cloud entirely when set to local_only
@Computer
Build PII Stripper pipeline — strips names, amounts, accounts, addresses from queries before cloud API calls; re-substitutes on response return
@Computer
Erik GitHub access — beta scope only, blocked on Nicole boundary convo
@Computer
Nicole box: configure passwordless sudo for /sbin/shutdown — enables watchdog auto-reboot. sudoers line: peterd ALL=(ALL) NOPASSWD: /sbin/shutdown. NOT on dev-01.
@Computer
Call Copilot: Ellen Ehrenpreis review — design consent framework that protects clients + counterparties across Paraguay, Germany, US jurisdictions
@Nicole · @Phone
Build drafter learning loop — track edits to drafted messages, diff draft vs sent, feed corrections back into future drafts. Blocked on new Cerberus head.
@Computer
Run full codebase audit via review.py swarm (106 files, 14 panels) — blocked on review.py deploy
@Computer
When Ellen responds to fee questions — verify fixed fee all-inclusive, Newco sequencing, AI tools, change order process. See ORRICK-SIGNING-CHECKLIST.md
@Computer
Brian Donovan review of Orrick engagement letter — 5 flagged sections (Sec 3 conflicts/Anthropic, Sec 12 promo, Founders Ack/Nicole counsel, Sec 10 AI tools, Sec 11 arbitration)
@Phone
NAS Setup: Harden DSM admin (non-standard port, IP restrict) + document backup config
@Computer
NAS post-purchase review: all devices backed up, first test restore completed
@Computer
Update provision.sh to include NAS backup target configuration for client deployments
@Computer
Strategic Communications Swarm — review running on dev-01 (8/11 panels done), assess results when complete
@Computer
Follow up: Debora Bon NDA — awaiting signed copy via Erik
@Phone
Run PSB Holdings portfolio analysis — asset-by-asset opportunity scan across ranches, Inpet, RE, portfolio stakes
@Computer
CorporationWiki access — registration broken, emailed support, waiting for fix
@Computer
Complete Orrick entity survey
@Computer
Run calibration suite, score against answer keys, fix weak panel calibrations
@Computer
First real Forge production run (Memory Graph ingestors) — post-calibration only
@Computer
Upgrade graduation criteria v1.3 with research-backed improvements (post-architecture graduation)
@Computer
Scheduled
23
Upgrade from Claude Sonnet 4 to latest Claude Sonnet version before June 15, 2026 retirement
@Computer
Review and potentially adjust default model settings for Claude Code team before April 30 when Opus 4.7 becomes default
@Computer
Assess potential cost impact of Opus 4.7 default model change for organization
@Computer
Review and potentially adjust default model settings for Claude Code team before April 30
@Computer
Assess potential cost impact of Opus 4.7 default model change
@Computer
Nicole RID session — walk through wedding guest list
7pm MT: Buy Brent Hallenbeck plane ticket
@Errands
DESIGN: Structured session state accumulation — log decisions/changes during sessions, produce handoff FROM the log
@Computer
DESIGN: 30-message context guillotine — automatic session rotation with fact-only state handoff
@Computer
Migrate from 1M context beta to Claude Sonnet 4.6 before April 30, 2026 deadline
@Computer
Download + benchmark 70B model on laptop (Llama 3.3 70B Q8 or Qwen 2.5 72B) — test behavioral analysis quality
@Computer
Build local inference API server on laptop — MLX serving endpoint for RID pipeline
@Computer
Add source document store to RID — raw emails/transcripts linked to identities, queryable by codename
@Computer
Build RID enrichment pipeline — ingest raw source, extract signals against 81 dimensions, score with evidence citations via local model
@Computer
Test RID enrichment pipeline with Ellen's March 23 email as first case — validate local 70B vs frontier quality
@Computer
Build enrich_profile CLI command — structured observation push from local analysis back into RID SQLCipher stores
@Computer
Build laptop-to-dev01 SQLCipher sync for RID stores — Tailscale, encrypted transfer, conflict resolution
@Computer
Sales pitch + pricing model + live demo — one-pager for prospects, pricing tiers, demo-ready for Karl/beta testers
@Computer
3pm: GTD onboarding with Nicole — write triage guide, sit together for first 20-task batch, establish daily routines
@Computer
Provision staging box (Mini #2, 24GB) — Gemma 4 + Ollama + modified provision.sh (from Gemma 4 chat)
@Computer
AirPods Max 2 — 30-day post-purchase review (comfort, battery, call quality, daily usage)
@Anywhere
Review first full-matrix + scorecard results after 4 days of synthesis data
@Computer
Physical Security: syslog adapter + 5 fast-track rules + hardware log validation (USA work)