{
  "$type": "site.standard.document",
  "bskyPostRef": {
    "cid": "bafyreiachnbfk7xespndegkdkrxabom3jvqxi5ke6u7m3np23vfl6kibge",
    "uri": "at://did:plc:pi6woz4d47bkuws673w2il2r/app.bsky.feed.post/3mnoey3zedm22"
  },
  "path": "/t/call-for-proposals-reviewing-testing-and-managing-llm-generated-haskell/14041?page=2#post_23",
  "publishedAt": "2026-06-07T03:19:48.000Z",
  "site": "https://discourse.haskell.org",
  "tags": [
    "github.com",
    "GitHub - Gabriella439/grace: A prompt engineering functional programming...",
    "Beyond agentic coding with Grace (w/ Gabriella Gonzalez)"
  ],
  "textContent": "Haven’t seen Grace mentioned here:\n\ngithub.com\n\n### GitHub - Gabriella439/grace: A prompt engineering functional programming...\n\nA prompt engineering functional programming language\n\nBeyond agentic coding with Grace (w/ Gabriella Gonzalez)\n\nMy favorite YouTube Haskeller Impure Pics had a great video about it. VERY interesting. I like the idea of guardrails on LLM’s and Grace achieves that quite elegantly. If prompt engineering is normally a declarative process, Grace changes that by using functional programming’s paradigm to encapsulate agents into safe(r) sub-processes that the type system is then able to check and iterate through within that sub agent sandbox. The way we protect our data through encapsulation and type checking is exactly how Grace is able to add those guardrails on these notoriously and confidently incorrect hallucination machines.",
  "title": "Call for proposals: Reviewing, testing, and managing LLM generated Haskell"
}