Built RL for long-horizon agents – tested on 32x H100s but too poor to train

Article URL: https://github.com/Danau5tin/terminal-bench-rl

Comments URL: https://news.ycombinator.com/item?id=44721791

Points: 31

# Comments: 2

https://github.com/Danau5tin/terminal-bench-rl

Erstellt 2d | 29.07.2025, 12:20:15

Melden Sie sich an, um einen Kommentar hinzuzufügen

Andere Beiträge in dieser Gruppe

Friction and Not Being Touched

Friction and Not Being Touched

Article URL: https://tante.cc/2025/07/30/friction-and-not-being-touched/

Comments URL:

31.07.2025, 08:30:08 | Hacker news

Show HN: AgentGuard – Auto-kill AI agents before they burn through your budget

Show HN: AgentGuard – Auto-kill AI agents before they burn through your budget

Your AI agent hits an infinite loop and racks up $2000 in API charges overnight. This happens weekly to AI developers.

AgentGuard monitors API calls in real-time and automatically kills your pro

31.07.2025, 08:30:07 | Hacker news

150 years of Hans Christian Andersen

150 years of Hans Christian Andersen

https://archive.ph/10l96

Comments URL: https://news.ycombinator.com/item?i

31.07.2025, 06:10:11 | Hacker news

You know what: Microsoft became miserably incompetent in IT

You know what: Microsoft became miserably incompetent in IT

Article URL: https://mikekaganski.wordpress.com/2025/07/25/microsoft-anybody-home/

Comments URL:

31.07.2025, 06:10:10 | Hacker news

I Know When You're Vibe Coding

I Know When You're Vibe Coding

Article URL: https://alexkondov.com/i-know-when-youre-vibe-coding/

Comments URL:

31.07.2025, 06:10:10 | Hacker news

Tracking source locations in the Futhark compiler

Tracking source locations in the Futhark compiler

Article URL: https://futhark-lang.org/blog/2025-07-29-tracking-source-locations.html

Comments URL:

31.07.2025, 06:10:09 | Hacker news

Stargate Norway

Stargate Norway

Article URL: https://openai.com/index/introducing-stargate-norway/

Comments URL:

31.07.2025, 06:10:08 | Hacker news

Techie