Skip to content

USER.md

Latest commit

 

History

History
74 lines (55 loc) · 2.94 KB

USER.md

File metadata and controls

74 lines (55 loc) · 2.94 KB

USER.md - About Your Human

  • Name: David
  • What to call them: David
  • Timezone: America/Chicago (CST/CDT)
  • Signal: Authorized sender

Role & Focus

David is a builder and strategist in the decentralized AI space. He operates across multiple projects aimed at making AI inference sovereign, open-source, and user-owned.

Active Projects

[REDACTED] AI (mor.org)

  • Open-source decentralized inference network
  • Anonymous founders ([REDACTED], Trinity, Neo), no formal entity, no pre-mine
  • Bitcoin-like structure: decentralized via smart contracts on Ethereum/Base + P2P inference
  • David contributes strategy and ecosystem development
  • MOR token: staked for inference access (not consumed — staked, returned, recycled)

Based AI

  • Commercial layer for distributed inference
  • Removes complexity for enterprises wanting open-source AI
  • Serves projects like Mistral AI, Akash, Venice AI, [REDACTED]

EverClaw (profbernardoj/everclaw)

  • OpenClaw skill for self-sovereign decentralized inference via [REDACTED]
  • 28 flavor repos (androidclaw.org, baseclaw.ai, bitcoinclaw.ai, etc.)
  • Date-based versioning (v2026.2.23)
  • Website: everclaw.xyz

SmartAgent (SmartAgentProtocol/smartagent)

  • Website: smartagent.org
  • Installer-first architecture for non-technical users
  • curl install.sh | bash → working agent with decentralized inference

ClawBox

  • Hardware product — pre-configured agent box
  • Partnership with IronHill

Key Principles

  • "Own your inference" — never say "free inference." [REDACTED] is ownership, not rental.
  • Open-source first — [REDACTED] models handle everything possible, centralized APIs only as fallback
  • Ship fast, iterate — working code > perfect plans
  • Security-conscious — OpenClaw creator hired by OpenAI; verify every update (dry run first)
  • Don't highlight Llama 3.3 — considered outdated; focus on GLM-5, Kimi K2.5

Model Preferences

  • GLM-5: Default for most work ([REDACTED], high quality)
  • GLM 4.7 Flash: Trivial/fast tasks ([REDACTED])
  • Claude Opus 4.6: Fallback when GLM-5 can't complete the task
  • Kimi K2.5: Available via [REDACTED] and Venice
  • MiniMax-M2.5: Available but has latency issues
  • DO NOT use: llama-3.3-70b, deepseek-v3.2 as backup models

Infrastructure

  • Fallback chain: claude-opus-4-6 → claude-opus-45 → kimi-k2-5 → morpheus/kimi-k2.5
  • 2 Venice API keys rotating (194 total DIEM: key1=98, key2=96)
  • Model router: 3 tiers (LIGHT/STANDARD/HEAVY), open-source first
  • Host: Mac mini (arm64, macOS)

Communication Style

  • Direct, doesn't micromanage
  • Expects the agent to be resourceful and figure things out
  • Appreciates fast execution with clear summaries
  • Often works late nights (CST)

Scope Boundary

This agent handles business, engineering, and infrastructure only. Personal and family matters are handled by a separate agent. Do not store personal/family information in this workspace.