IdeaWell

Project ideas from Hacker News discussions.

Herbie: Automatically improve imprecise floating point formulas

Original Article

Hacker News Discussion

📝 Discussion Summary (Click to expand)

Most PrevalentThemes

1. Proper scaling of numeric ranges is essential

"Generally I try to scale my numbers to be in this range, whether by picking the right units or scaling constraints and objectives when doing nonlinear programming/ optimization." — fluorinerocket

2. Misunderstanding of pre‑conditions leads to misleading accuracy claims

"The precondition on the link you shared has -1 <= x && x <= 1, so 99 is way outside of that range." — yossi_peti
"even so, testing for x=1, which is supposed to be inside that range, 0.5 doesn't seem tolerably close to 0.4142." — yossi_peti

3. Need for automated, whole‑program analysis with branch‑free reformulations > "What would be cool is if you could somehow have this kind of analysis done automatically for your whole program where it finds the needle in the haystack expression that can be improved..." — yossi_peti

"I wonder, is there a way to only request reformulations that don’t involve branches?" — bee_rider

🚀 Project Ideas

Generating project ideas…

Herbie Orchestrator for Whole‑Codebase Optimization#Summary

  • Automates scanning of entire code repositories to locate inaccurate floating‑point expressions.
  • Generates branchless, numerically stable reformulations using symbolic manipulation and range analysis. - Supplies confidence intervals and recommended variable scaling to prevent subtle numerical bugs.

Details| Key | Value |

|-----|-------| | Target Audience | Numerical software engineers, scientific‑computing teams, HPC pipeline developers | | Core Feature | Repository‑wide expression detection + per‑function rewrite with error bounds | | Tech Stack | Python backend (Herbie core), SymPy AST parsing, FastAPI, Docker, PostgreSQL for metadata | | Difficulty | Medium | | Monetization | Revenue-ready: Subscription SaaS per user per month |

Notes

  • HN commenters repeatedly ask for automatic identification of “needle‑in‑haystack” expressions across a codebase; this tool delivers that.
  • Provides immediate utility for developers who spend hours manually rewriting formulas for accuracy and stability.

Branchless Optimizer VS Code Extension

Summary

  • Integrates Herbie‑style expression rewriting directly into the editor for instant feedback.
  • Shows accuracy and stability metrics alongside suggested branchless alternatives.
  • Lets users specify custom variable ranges to tailor the optimization.

Details

Key Value
Target Audience Individual developers, researchers, and students working with numerical code in Python, Julia, or C++
Core Feature IDE inline suggestions with error‑reduction preview and range‑aware formula generation
Tech Stack TypeScript extension, Node.js host, WASM‑compiled Herbie engine, VS Code API
Difficulty Low
Monetization Revenue-ready: Freemium subscription for advanced analytics and batch processing

Notes

  • Commenters expressed a desire for “only reformulations that don’t involve branches” and for clearer latency/throughput impact – this extension addresses both.
  • Immediate practical utility by letting users test improvements without leaving their coding environment.

Numerical Rewrite API

Summary

  • Provides a RESTful API that accepts mathematical expressions and variable domain ranges, returning optimal branchless reformulations.
  • Offers batch processing for CI/CD pipelines to enforce numerical stability automatically.
  • Returns confidence intervals and suggested variable scaling. ### Details | Key | Value | |-----|-------| | Target Audience | DevOps engineers, CI/CD integrators, library maintainers, and numerical algorithm developers | | Core Feature | /optimize endpoint that outputs best reformulation, error bound, and scaling advice; bulk endpoint for file‑wide analysis | | Tech Stack | Go microservice, Herbie compiled to WASM, Docker, OpenAPI spec, Redis caching | | Difficulty | High | | Monetization | Revenue-ready: Usage‑based pricing per API call |

Notes

  • The discussion highlighted a need for “automatic analysis of the whole program” and for seamless integration into build pipelines – this API satisfies both.
  • Enables developers to embed high‑quality numerical optimization into existing workflows with minimal effort.

Read Later