Lex
notesbylex.com
Senior MLE at @Canva. Full-stack developer. Talks about Software dev, LLMs, Agentic Reasoning, Obsidian, note-taking, dog photos.
👤 Joined on 29 November 2023
91
Followers
11
Following
83
Total Posts
📜 Kullanıcı Geçmişi
- 29 Kas 2023, 23:36
🎉 Bluesky'e katıldı - İlk kullanıcı adı:
@lext.bsky.social - 17 Kas 2024, 06:08
✏️ Kullanıcı adı değişti:
@lext.bsky.social→@notesbylex.com
🚫 Blokladığı Kullanıcılar
❤️ Beğendiği Gönderiler
Really enjoyed this! It got me thinking about how much of building useful AI products is really about the harness, not just the LLM. Curious what you'd focus on next to make this work reliably across different tasks: evals, context management, or failure recovery?
Show Post
We may be growing tired of new LLM benchmarks, but this post by @notesbylex.com re: a new "Degradation Benchmark" is both interesting/scary AND makes me feel less paranoid when I add "do not edit any files unless absolutely necessary; any files edited MUST be added to CHANGELOG.md" to every prompt.
Show Post