HackerReposA FIELD GUIDE TO OPEN SOURCE
ProSponsor the shelf♧ 0
← Back to top all-time
gepa-ai
gepa
Jupyter Notebook2025–2026
gepa-ai / gepa

gepa

Optimize prompts, code, and more with AI-powered Reflective Optimization

4Distinct people
7Mentions
0Explicit recommenders
7.4KGitHub stars
REPOSITORY AT A GLANCE

What to know before exploring

What it is

Optimize prompts, code, and more with AI-powered Reflective Optimization

GitHub

Where it is used

No documented product use has been verified yet.

What people are saying

No recommendation phrase was found in the indexed comments.

No classified warning is attached to this page. Read the discussions before treating silence as approval.

Also in these discussions

No other repository is repeatedly linked in the same discussions.

This week on Hacker News

Nov 9–Nov 16 UTC · Direct repository links and separately labeled related project stories. Indexed through 11/16/2026, 7:57:46 PM UTC.

No direct repository links were found in the indexed HN discussions over the past seven days. Older discussions are available below.

Mentions over time

HN mentions by week7 dated mentions

Observed mentions: 2025-09-10 05:21:17 UTC to 2026-03-23 20:17:21 UTC. Weeks start Monday, UTC.

01232025-09-08 00:00:00 UTC to 2025-09-15 00:00:00 UTC (exclusive): 1 mentions; 1 submissions, 0 commentsSep 82025-09-15 00:00:00 UTC to 2025-09-22 00:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2025-09-22 00:00:00 UTC to 2025-09-29 00:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2025-09-29 00:00:00 UTC to 2025-10-06 00:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2025-10-06 00:00:00 UTC to 2025-10-13 00:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2025-10-13 00:00:00 UTC to 2025-10-20 00:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2025-10-20 00:00:00 UTC to 2025-10-27 00:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2025-10-27 00:00:00 UTC to 2025-11-03 00:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2025-11-03 00:00:00 UTC to 2025-11-10 00:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2025-11-10 00:00:00 UTC to 2025-11-17 00:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2025-11-17 00:00:00 UTC to 2025-11-24 00:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2025-11-24 00:00:00 UTC to 2025-12-01 00:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2025-12-01 00:00:00 UTC to 2025-12-08 00:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2025-12-08 00:00:00 UTC to 2025-12-15 00:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2025-12-15 00:00:00 UTC to 2025-12-22 00:00:00 UTC (exclusive): 2 mentions; 0 submissions, 2 commentsDec 152025-12-22 00:00:00 UTC to 2025-12-29 00:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2025-12-29 00:00:00 UTC to 2026-01-05 00:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2026-01-05 00:00:00 UTC to 2026-01-12 00:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2026-01-12 00:00:00 UTC to 2026-01-19 00:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2026-01-19 00:00:00 UTC to 2026-01-26 00:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2026-01-26 00:00:00 UTC to 2026-02-02 00:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2026-02-02 00:00:00 UTC to 2026-02-09 00:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2026-02-09 00:00:00 UTC to 2026-02-16 00:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2026-02-16 00:00:00 UTC to 2026-02-23 00:00:00 UTC (exclusive): 3 mentions; 2 submissions, 1 comments2026-02-23 00:00:00 UTC to 2026-03-02 00:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2026-03-02 00:00:00 UTC to 2026-03-09 00:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2026-03-09 00:00:00 UTC to 2026-03-16 00:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2026-03-16 00:00:00 UTC to 2026-03-23 00:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2026-03-23 00:00:00 UTC to 2026-03-30 00:00:00 UTC (exclusive): 1 mentions; 0 submissions, 1 commentsMar 23

Counts come from the repository’s dated source evidence. Index cutoff: 2026-11-02 08:16:59 UTC.

View mention counts and date ranges
UTC ranges include their start and exclude their end. Select a period to filter the source evidence.
PeriodStart (UTC)End (exclusive, UTC)MentionsSubmissionsComments
Sep 82025-09-08 00:00:00 UTC2025-09-15 00:00:00 UTC110
Sep 152025-09-15 00:00:00 UTC2025-09-22 00:00:00 UTC000
Sep 222025-09-22 00:00:00 UTC2025-09-29 00:00:00 UTC000
Sep 292025-09-29 00:00:00 UTC2025-10-06 00:00:00 UTC000
Oct 62025-10-06 00:00:00 UTC2025-10-13 00:00:00 UTC000
Oct 132025-10-13 00:00:00 UTC2025-10-20 00:00:00 UTC000
Oct 202025-10-20 00:00:00 UTC2025-10-27 00:00:00 UTC000
Oct 272025-10-27 00:00:00 UTC2025-11-03 00:00:00 UTC000
Nov 32025-11-03 00:00:00 UTC2025-11-10 00:00:00 UTC000
Nov 102025-11-10 00:00:00 UTC2025-11-17 00:00:00 UTC000
Nov 172025-11-17 00:00:00 UTC2025-11-24 00:00:00 UTC000
Nov 242025-11-24 00:00:00 UTC2025-12-01 00:00:00 UTC000
Dec 12025-12-01 00:00:00 UTC2025-12-08 00:00:00 UTC000
Dec 82025-12-08 00:00:00 UTC2025-12-15 00:00:00 UTC000
Dec 152025-12-15 00:00:00 UTC2025-12-22 00:00:00 UTC202
Dec 222025-12-22 00:00:00 UTC2025-12-29 00:00:00 UTC000
Dec 292025-12-29 00:00:00 UTC2026-01-05 00:00:00 UTC000
Jan 52026-01-05 00:00:00 UTC2026-01-12 00:00:00 UTC000
Jan 122026-01-12 00:00:00 UTC2026-01-19 00:00:00 UTC000
Jan 192026-01-19 00:00:00 UTC2026-01-26 00:00:00 UTC000
Jan 262026-01-26 00:00:00 UTC2026-02-02 00:00:00 UTC000
Feb 22026-02-02 00:00:00 UTC2026-02-09 00:00:00 UTC000
Feb 92026-02-09 00:00:00 UTC2026-02-16 00:00:00 UTC000
Feb 162026-02-16 00:00:00 UTC2026-02-23 00:00:00 UTC321
Feb 232026-02-23 00:00:00 UTC2026-03-02 00:00:00 UTC000
Mar 22026-03-02 00:00:00 UTC2026-03-09 00:00:00 UTC000
Mar 92026-03-09 00:00:00 UTC2026-03-16 00:00:00 UTC000
Mar 162026-03-16 00:00:00 UTC2026-03-23 00:00:00 UTC000
Mar 232026-03-23 00:00:00 UTC2026-03-30 00:00:00 UTC101

What Hacker News says

3 sources
We built optimize_anything, an API that optimizes any artifact representable as text — code, prompts, agent architectures, configs, even SVGs. It extends GEPA (our prompt optimizer, discussed here previously: https://arxiv.org/abs/2507.19457) far beyond prompts. The API is deliberately minimal. You provide what to optimize and how to measure it: import gepa.optimize_anything as oa def evaluate(candidate: str) -> tuple[float, dict]: result = run_my_system(candidate) return result.score, {"error": result.stderr, "runtime": f"{result.time_ms}ms"} result = oa.optimize_anything( seed_candidate="<your artifact>", evaluator=evaluate, ) The evaluator returns a score plus diagnostic feedback (we call it "Actionable Side Information" — stack traces, rendered images, profiler output, whatever helps diagnose failures). An LLM proposer reads this feed
Read the rest
back during a reflection step and proposes targeted fixes, not blind mutations. Candidates are selected via a Pareto frontier across metrics/examples, so a candidate that's best at one thing survives even if its average is mediocre. Two ideas distinguish this from AlphaEvolve/OpenEvolve/ShinkaEvolve-style LLM evolution: (1) diagnostic feedback is a first-class API concept rather than a framework-specific mechanism, and (2) the API unifies three optimization modes — single-task search (solve one hard problem), multi-task search (solve related problems with cross-transfer), and generalization (build artifacts that transfer to unseen inputs). Prior frameworks only express mode 1. We tested across 8 domains. Selected results: Coding agent skills: Learned repo-specific skills push Claude Code to near-perfect task completion and make it 47% faster Cloud scheduling: Discovered algorithms that cut costs 40%, topping the ADRS leaderboard over expert heuristics and other LLM-evolution frameworks Agent architecture: Evolved a 10-line stub into a 300+ line ARC-AGI agent, improving Gemini Flash from 32.5% → 89.5% Circle packing (n=26): Outperforms AlphaEvolve's published solution Blackbox optimization: Generated problem-specific solvers matching or exceeding Optuna across 56 EvalSet problems CUDA kernels: 87% match or beat baseline; multi-task mode outperforms dedicated single-task runs ``` pip install gepa ``` Blog with full results and runnable code for all 8 case studies: https://gepa-ai.github.io/gepa/blog/2026/02/18/introducing-o... GitHub: https://github.com/gepa-ai/gepa
In Optimize_anything: A Universal API for Optimizing Any Text Parameter

Show HN: Optimize_anything: A Universal API for Optimizing Any Text Parameter

We built optimize_anything, an API that optimizes any artifact representable as text — code, prompts, agent architectures, configs, even SVGs. It extends GEPA (our prompt optimizer, discussed here previously: https://arxiv.org/abs/2507.19457) far beyond prompts. The API is deliberately minimal. You provide what to optimize and how to measure it: import gepa.optimize_anything as oa def evaluate(candidate: str) -> tuple[float, dict]: result = run_my_system(candidate) return result.score, {"error": result.stderr, "runtime": f"{result.time_ms}ms"} result = oa.optimize_anything( seed_candidate="<your artifact>", evaluator=evaluate, ) The evaluator returns a score plus diagnostic feedback (we call it "Actionable Side Information" — stack traces, rendered images, profiler output, whatever helps diagnose failures). An LLM proposer
Read the rest
reads this feedback during a reflection step and proposes targeted fixes, not blind mutations. Candidates are selected via a Pareto frontier across metrics/examples, so a candidate that's best at one thing survives even if its average is mediocre. Two ideas distinguish this from AlphaEvolve/OpenEvolve/ShinkaEvolve-style LLM evolution: (1) diagnostic feedback is a first-class API concept rather than a framework-specific mechanism, and (2) the API unifies three optimization modes — single-task search (solve one hard problem), multi-task search (solve related problems with cross-transfer), and generalization (build artifacts that transfer to unseen inputs). Prior frameworks only express mode 1. We tested across 8 domains. Selected results: Coding agent skills: Learned repo-specific skills push Claude Code to near-perfect task completion and make it 47% faster Cloud scheduling: Discovered algorithms that cut costs 40%, topping the ADRS leaderboard over expert heuristics and other LLM-evolution frameworks Agent architecture: Evolved a 10-line stub into a 300+ line ARC-AGI agent, improving Gemini Flash from 32.5% → 89.5% Circle packing (n=26): Outperforms AlphaEvolve's published solution Blackbox optimization: Generated problem-specific solvers matching or exceeding Optuna across 56 EvalSet problems CUDA kernels: 87% match or beat baseline; multi-task mode outperforms dedicated single-task runs ``` pip install gepa ``` Blog with full results and runnable code for all 8 case studies: https://gepa-ai.github.io/gepa/blog/2026/02/18/introducing-o... GitHub: https://github.com/gepa-ai/gepa
8 points · Open discussion

MAKE YOURSELF AT HOME

A few shortcuts

/
Search the collection
?
Open this guide
Esc
Close dialogs and navigation

Your saved repositories and interests stay in this browser.