Search memories, docs, and tasks from the real workspace store.

memorymemory/2026-04-03.md4/3/2026, 9:15:36 PM

2026-04-03

Claude Memory Import Review

Reviewed a Claude memory export for Tim and merged only net-new durable items.

Added / retained as useful durable context:

  • Tim’s pronouns: he/him
  • Location: Woodstock, IL
  • Email accounts: tim@shaylabs.com, tim@tntproseries.com, hello@shaylabs.com
  • Primary computer: Mac Studio
  • Apple ecosystem usage
  • Labradoodle puppy named Rocky
  • Family / dad / family-routine context matters
  • Company context: TNT Pro Series, Shay Labs, Effiniti Group
  • Amazon / brand context: TNT Pro Series, Queen Helene, Prell
  • UGC Engine data layer rule: Neon.tech (PostgreSQL), not Supabase
  • n8n self-hosted on Hostinger VPS remains relevant system context

Did not add duplicate items already covered elsewhere:

  • Tim’s name
  • timezone / general working context already known
  • high-level strategic preferences already captured in USER.md / AGENTS.md

Note: kept durable facts in MEMORY.md and user-profile style items in USER.md.

Claude Memory Import Review (Projects tranche)

Reviewed additional Claude-exported project memory and merged only net-new durable project context.

Added / retained as useful durable context:

  • OpticPrime as an AI product-image generation app for Amazon sellers
  • OpticPrime architecture and workflow context (three-layer prompt compiler, Creative Lab, divergence/regeneration logic, mobile-first readability, category/tile systems)
  • OpticPrime status note: currently shelved after dissatisfaction with results
  • OpticPrime adjacent systems: prompt compiler code, supplements/vitamins tiles, A+ builder flows
  • ClipWise as an early-stage AI video editing fork preceding the broader UGC Engine
  • Three-account TikTok strategy per brand: Education/Authority, Comedy/Entertainment, Product/Sales
  • Brand voice definitions for TNT Pro Series, Queen Helene, and Prell

Did not duplicate:

  • already-known company/brand context
  • already-known UGC Engine / Amazon / n8n baseline context

Claude Memory Import Review (UGC / analytics tranche)

Reviewed additional Claude-exported build/project memory and merged only net-new durable system context.

Added / retained as useful durable context:

  • fuller UGC Engine platform description (surfaces, generation pipeline, narrative formats)
  • Brand Analytics Dashboard as a 7-group AI review-analysis system feeding downstream products
  • Remotion / Whisper / Kokoro / FAL / KIE implementation context for UGC Engine
  • competitive review analysis / indirect competitor reference rules
  • AI Listing Mastermind (ai-vs-ai)
  • TopView AI Omni Reference exploration
  • TNT Pro Series SEO/backlink strategy exploration
  • tntproseries.com analytics setup with Microsoft Clarity and GA4
  • Amazon FBA profitability modeling for Queen Helene and Prell

Did not duplicate:

  • already-recorded OpticPrime/ClipWise/TikTok strategy context
  • already-recorded Neon / n8n / brand baseline context

Later-day durable updates

  • Tim wants a daily Telegram morning brief at 8:00 AM America/Chicago including: relevant news, business ideas, tasks for the day, and recommendations for what Tim and the agent should tackle together.
  • Tim explicitly reinforced the execution rule: when given a project, the agent should carry through to final outcome without repeated check-ins unless genuinely blocked, high-risk, or needing a real decision.
  • Tim explicitly reinforced the completion-update rule: when work is finished, the agent should clearly say it is done.
  • Tim wants the GitHub repo tnt21101/second-brain-nextjs to serve as the shared home base for operating files, docs, and ongoing collaboration.
  • The GitHub home base now includes mirrored core operating files under agent-core/ and is treated as the shared operating base.
  • A new durable rule was added conceptually: operationally important skills/tools should be written into core files (especially AGENTS.md / TOOLS.md) rather than merely installed.
  • Tim’s additional durable working-style/profile context from Claude exports: keeps a lean calendar, prefers deep work, favors proper applications over n8n for core product functionality, wants existing work checked before starting from scratch, and wants a plan shown first if a task would take 20+ steps.
  • Claude-exported durable project context was reviewed and merged: OpticPrime, UGC Engine / Brand Analytics / Remotion stack context, AI Listing Mastermind, TopView AI exploration, TNT Pro Series SEO/backlink strategy, GA4 + Clarity setup, and Amazon FBA profitability modeling are relevant long-term context.
  • The 2nd Brain app was evolved from a simple viewer into a more operational system: real file-backed search, an Agent Review page, and an AutoAgent-inspired meta/recommendation layer were added and pushed to GitHub.
  • Mission Control was rebuilt inside second-brain-nextjs as an operator-oriented dashboard shell centered on Amazon profit analytics, bottleneck surfacing, execution lanes, and a kill / push / watch portfolio classifier. The Amazon/SP-API data layer is currently architected with a mock adapter pending live SP-API hookup.
  • Ongoing unfinished high-value items remain: real Google Workspace gog auth, final Paperclip ↔ OpenClaw join wiring, and swapping Mission Control’s mock Amazon adapter to a real SP-API-backed profit adapter.

Hermes isolated evaluation

Installed Hermes Agent in an isolated path outside the main OpenClaw operating path:

  • code: /root/.parallel-platforms/hermes-agent
  • home/config: /root/.parallel-platforms/hermes-home
  • launcher: /root/.parallel-platforms/bin/hermes-isolated

Validated:

  • repo clone and Python 3.11 virtual environment
  • editable install of Hermes Agent + tinker-atropos
  • isolated hermes version, hermes doctor, hermes status, and CLI help
  • wrapper hardening to prevent accidental inheritance of ambient API keys from the parent shell

Recommendation:

  • keep Hermes as an isolated bench / R&D runtime for now
  • do not promote it to a first-class primary operating layer yet
  • keep OpenClaw as the main operator runtime and Paperclip as the multi-agent orchestration layer

Also mirrored the shared-governance rule into the GitHub home base and added a comparison doc for OpenClaw vs Hermes vs Paperclip.

memU operationalization

Installed and operationalized memU as a local optional structured-memory sidecar for future agents.

Installed artifacts:

  • source checkout: /root/.openclaw/workspace/vendor/memU
  • isolated Python 3.13 runtime: /root/.openclaw/workspace/.tools/memu
  • PATH launcher for Python access: /root/.local/bin/memu-python

Validated:

  • memu-py 1.5.1 installs cleanly in an isolated Python 3.13 env via uv
  • import works and Rust extension returns Hello from memu!
  • MemoryService(...) initializes successfully with the default in-memory backend

Known blockers / caveats:

  • upstream memu-server console entrypoint in this release is broken (ModuleNotFoundError: memu.server)
  • SQLite initialization is broken in local validation (ValueError: <class 'list'> has no matching SQLAlchemy type)
  • documented future-agent usage should therefore prefer import-based access through memu-python