{
  "id": 5899425,
  "title": "99.7% Rejected in 84ms: Why I Stopped Making the Generator Smarter",
  "url": "https://urgent.news/2026/09/06/99-7-rejected-in-84ms-why-i-stopped-making-the-generator-smarter",
  "topic": "tech",
  "section": "Tech",
  "published": "2026-09-06T03:34:11.000Z",
  "source": {
    "name": "Dev.to",
    "slug": "dev-to",
    "url": "https://dev.to/hidetzu/997-rejected-in-84ms-why-i-stopped-making-the-generator-smarter-5em0"
  },
  "original_language": "en",
  "account": "In developing a puzzle generator, I found that 99.7% of its outputs were rejected. This rejection rate of 99.7% means that only 0.26% of generated puzzles were accepted. The generator does not guarantee correctness and relies on a separate verifier to check each solution. The generator produces plausible-looking letter combinations, while the verifier exhaustively searches for solutions and confirms their uniqueness. Various reasons for rejection include no unique solution, lack of a solution, having more than 10 distinct letters, no leading zeros, and exceeding three letters. A significant portion of rejections occur before the search begins due to these criteria. The verifier's role is crucial, as it ensures the generator's outputs are valid. The generator's efficiency is not a concern; instead, the focus is on the verifier's ability to confirm the solutions. The design decisions that maintain the rejection rates include the verifier's exhaustive search, the guarantee of a unique solution, and the exclusion of puzzles with more than 10 distinct letters. The separation of concerns between the generator and verifier allows for a trade-off where the generator can be less intelligent, and the verifier can afford to run longer. This approach is applied across various domains, including code generation and structured output generation, where verification remains the minimum requirement.",
  "summary": "I wrote a puzzle generator whose acceptance rate is 0.26% . It throws away 99.7% of everything it produces, and that is the design working as intended, not failing. Generating five valid puzzles takes 1,947 attempts and 84 milliseconds. The point is not the puzzles. The point is that the generator makes no correctness guarantee at all, and a verifier makes every one of them. Once you split those…",
  "key_points": [
    "99.7% rejection rate means only 0.26% of puzzles accepted",
    "Generator produces plausible letter combinations, verifier checks solutions",
    "Verifier's exhaustive search and unique solution guarantee maintain rejection rates"
  ],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}