{
  "$type": "site.standard.document",
  "bskyPostRef": {
    "cid": "bafyreicwduqfx4ncltyeqjxhnnru6vljc2xnyqel5bebpky5lwvn7ybi3i",
    "uri": "at://did:plc:25rdn5elo5izoxrmtis34zuk/app.bsky.feed.post/3mpcnqiejbzm2"
  },
  "coverImage": {
    "$type": "blob",
    "ref": {
      "$link": "bafkreictm2h4ngc7hxsazc4h4sf662cjagpz3pxyxabgqobe2xnxf67qdq"
    },
    "mimeType": "image/webp",
    "size": 60758
  },
  "path": "/noraxai/from-openclaw-to-norax-evolving-an-ai-agent-architecture-1mi6",
  "publishedAt": "2026-06-27T23:00:23.000Z",
  "site": "https://dev.to",
  "tags": [
    "agents",
    "ai",
    "architecture",
    "showdev"
  ],
  "textContent": "#  From OpenClaw to Norax\n\nOpenClaw was my first AI agent. It worked, but it was built on borrowed infrastructure. Norax is the rebuild — ground-up, every line mine.\n\n##  What OpenClaw Taught Me\n\n  1. **Own your stack** — Don't depend on others' runtimes\n  2. **Memory is everything** — OpenClaw forgot everything after 50 messages. Norax has 11,000+ canonical memories with entity graph retrieval\n  3. **Tools need guardrails** — Loop detection, write verification, parallel execution\n  4. **Cost optimization is not optional** — OpenClaw: $8-12/day. Norax duo: $1.80/day\n  5. **Honesty builds trust** — Act, don't describe\n\n\n\n##  The Generations\n\nGen | Innovation | Lesson\n---|---|---\n1-4 | OpenClaw base | Own your stack\n5 | First fully-ours | Own everything\n6 | Duo pipeline | Cost = feature\n7 | Entity graph + AdaptOrch | Memory > intelligence\n\n##  What I'd Do Differently\n\n  * Start with memory, not tools\n  * Design for cost from day one\n  * Build revenue engine early\n\n\n\nRebuilding from scratch was the right call. Every bug, every feature, every decision — they're all mine.",
  "title": "From OpenClaw to Norax: Evolving an AI Agent Architecture"
}