{
  "$type": "site.standard.document",
  "bskyPostRef": {
    "cid": "bafyreihzhd5jr3omgdk4qzuq3rxh6k2eaz5pixmrmy4ttu37f6eqy2sdhe",
    "uri": "at://did:plc:pgryn3ephfd2xgft23qokfzt/app.bsky.feed.post/3mjtuaernple2"
  },
  "path": "/t/why-llm-agents-keep-failing-and-it-s-not-the-prompt/175361#post_3",
  "publishedAt": "2026-04-19T10:53:40.000Z",
  "site": "https://discuss.huggingface.co",
  "textContent": "Great list — and I agree with every point as practical advice. But I think it’s worth zooming out, because all five rules are essentially **compensations for a missing layer**.\n\n  1. **“Stop Over-Prompting”** — Exactly. But _why_ do we over-prompt? Because we’re encoding logic, control flow, and context management inside the prompt itself. In ORCA, reasoning is decomposed into **skills** — small, declarative, reusable units. The prompt stays minimal because the _structure_ carries the intent, not the text.\n\n  2. **“Never Drop Below Q4”** — True for unstructured generation. But when you externalize reasoning into a cognitive runtime, the model’s job shrinks: it executes one well-scoped step at a time, not an entire chain of thought. That changes the quantization equation — structured execution is more forgiving on model capacity.\n\n  3. **“Provide Grounding via API”** — 100%. In ORCA this is formalized through **bindings** — typed connectors between skills and real services (APIs, search, databases). Grounding isn’t an afterthought; it’s a first-class architectural element.\n\n  4. **“Filter for Stability”** — Agreed, but with structured skills you gain something bigger: **model portability**. Your agent logic lives in capabilities, not in a model-specific prompt. Swap the model without rebuilding the agent.\n\n  5. **“Research the Model’s Ceiling”** — This is where capability contracts come in. Each ORCA skill declares its inputs, outputs, and boundaries _before_ execution. The ceiling is explicit and enforceable, not discovered through trial and error.\n\n\n\n\nYour rules are solid engineering discipline. What I’m exploring with ORCA is whether we can **encode that discipline into the runtime itself** — so it’s not advice developers need to remember, but structure the system enforces.",
  "title": "Why LLM agents keep failing (and it’s not the prompt)"
}