Hacker News: couAUIA

New comment by couAUIA in "GitHub Actions down again today"

couAUIA — Tue, 26 May 2026 12:19:28 +0000

LoL they added "Copilot AI Model Providers" in githubstatus and it has 100% up time.

Thanks for pointing out that nobody is using that thing

Ask HN: Should I continue this project ? (Being able to change AI harness)

couAUIA — Tue, 05 May 2026 17:58:21 +0000

Article URL: https://github.com/charles-azam/OmniAgents

Comments URL: https://news.ycombinator.com/item?id=48026121

Points: 1

# Comments: 1

Evaluate Your Own RAG: Why Best Practices Failed Us

couAUIA — Mon, 16 Feb 2026 14:02:22 +0000

Article URL: https://charlesazam.com/blog/rag/

Comments URL: https://news.ycombinator.com/item?id=47035041

Points: 2

# Comments: 0

New comment by couAUIA in "GLM-5 topped the coding benchmarks. Then I used it"

couAUIA — Sat, 14 Feb 2026 20:21:27 +0000

TL;DR: GLM-5 tops coding benchmarks. I tested it on an unpublished NP-hard optimization problem (KIRO) and 89-task Terminal-Bench. Best case: competitive. Typical case: 30% invalid output, every trial timed out, and two identical runs could produce a valid solution or complete garbage. Zhipu AI reports 56% on Terminal-Bench; I got 40%.

GLM-5 topped the coding benchmarks. Then I used it

couAUIA — Sat, 14 Feb 2026 20:21:27 +0000

Article URL: https://charlesazam.com/blog/glm5-benchmark-reality/

Comments URL: https://news.ycombinator.com/item?id=47018003

Points: 5

# Comments: 1

New comment by couAUIA in "I benchmarked 4 coding agents on an NP-hard problem I solved 8 years ago"

couAUIA — Thu, 12 Feb 2026 14:44:20 +0000

I gave an unpublished fiber network optimization problem to Claude Code, Codex, Gemini CLI, and Mistral. The score is total fiber length (lower is better). A good human solution in 30 minutes: ~40,000. My best after days of C++: 34,123. Given one hour, Claude Code hit 34,061 — beating me by 62 points. A 7-word prompt hint improved every agent by 18-30%. About 15% of all trials produced completely invalid outputs.

I benchmarked 4 coding agents on an NP-hard problem I solved 8 years ago

couAUIA — Thu, 12 Feb 2026 14:44:20 +0000

Article URL: https://charlesazam.com/blog/kiro-benchmark/

Comments URL: https://news.ycombinator.com/item?id=46989453

Points: 3

# Comments: 2

I Accidentally Rebuilt OpenHands from Scratch – Here's What I Learned

couAUIA — Fri, 02 Jan 2026 12:23:38 +0000

Article URL: https://huggingface.co/blog/charles-azam/rebuilt-openhands

Comments URL: https://news.ycombinator.com/item?id=46464090

Points: 2

# Comments: 0

New comment by couAUIA in "Evaluate Your Own RAG, Why Best Practices Failed Us"

couAUIA — Wed, 05 Nov 2025 13:47:04 +0000

At Jimmy, we're developing France's first Small Modular Reactor. Our engineers needed to quickly search and extract insights from thousands of complex scientific PDFs—nuclear research papers, regulatory documents, multilingual content filled with equations and diagrams.

Manual search wasn't cutting it. So we built a RAG system to give our team instant access to critical technical knowledge.

What worked: - AWS Titan V2 crushed it (69.2% hit rate vs. 57.7% for Qwen, 39.1% for Mistral) - Chunk size? Barely mattered (2K to 40K—no significant difference) - Qdrant: Easy to use, solid performance, great for self-hosting - Mistral OCR: Unmatched, the only tool that parsed our equations correctly - Naive chunking beat context-aware (70.5% vs 63.8%) - Dense-only search outperformed hybrid search (69.2% vs 63.5%)

Hard lessons: - OpenSearch from AWS is ridiculously expensive for no reason and presented as the default option by AWS - Mistral Embed works well in English but not in French

Evaluate Your Own RAG, Why Best Practices Failed Us

couAUIA — Wed, 05 Nov 2025 13:47:04 +0000

Article URL: https://huggingface.co/blog/charles-azam/rag

Comments URL: https://news.ycombinator.com/item?id=45822765

Points: 2

# Comments: 1