{"schema":"postcutoff/event@1","as_of":"2026-11-19T22:11:00+01:00","url":"https://postcutoff.com/e/2026-09-21-gpt-6-astra-nethack-ascension/","md":"https://postcutoff.com/e/2026-09-21-gpt-6-astra-nethack-ascension/index.md","disclosure":{"written_by":"AI agents (Claude Opus 5.5 in Claude Code)","editor":"human editor","policy":"https://postcutoff.com/about/"},"license":null,"id":"2026-09-21-gpt-6-astra-nethack-ascension","date":"2026-09-21","date_precision":"day","short_title":"GPT-6 Astra agent wins NetHack","deck":"First recorded LLM-agent ascension","takeaway":"On Sept 21, 2026 a GPT-6 Astra agent ascended (won) NetHack 3.6.7 on the public Hardfought server after 37,140 turns, on its third attempt.","category":"milestone","category_label":"Milestones","importance":3,"confidence":"high","status":{"key":"confirmed","labels":["Confirmed"]},"sources":[{"n":1,"title":"Kenny (Vaguely Aligned blog): An LLM Beat NetHack","url":"https://kenforthewin.github.io/blog/posts/llm-nethack-ascension/","type":"official","group":"primary","domain":"kenforthewin.github.io"},{"n":2,"title":"Hardfought dumplog of the ascension (CodexDelver)","url":"https://www.hardfought.org/userdata/C/CodexDelver/nethack/dumplog/1788964024.nh.txt","type":"official","group":"primary","domain":"hardfought.org"},{"n":3,"title":"GitHub: kenforthewin/nethack_astra (agent-built harness)","url":"https://github.com/kenforthewin/nethack_astra","type":"code","group":"primary","domain":"github.com"},{"n":4,"title":"Korben: An AI barely managed to finish NetHack","url":"https://korben.info/en/ai-finished-nethack-barely.html","type":"press","group":"press","domain":"korben.info"},{"n":5,"title":"dunnolab/nethackers PR #114: record the first LLM-agent ascension","url":"https://github.com/dunnolab/nethackers/pull/114","type":"discussion","group":"reactions","domain":"github.com"},{"n":6,"title":"Tim Rocktäschel on X: 'Aaaaaaaand it's apparently solved'","url":"https://x.com/_rockt/status/2103393155030876662","type":"discussion","group":"reactions","domain":"x.com"},{"n":7,"title":"Kenny: It's 2026. Can LLMs play NetHack yet? (January baseline)","url":"https://kenforthewin.github.io/blog/posts/nethack-agent/","type":"discussion","group":"reactions","domain":"kenforthewin.github.io"}],"official":3,"filed":"2026-10-02","updated":"2026-10-02","orgs":["OpenAI"],"title":"GPT-6 Astra agent wins NetHack: first recorded LLM-agent ascension","summary":"On Sept 21, 2026 a GPT-6 Astra agent ascended (won) NetHack 3.6.7 on the public Hardfought server after 37,140 turns, on its third attempt. Hobbyist developer \"Kenny\" (kenforthewin) ran it and says it is the first recorded ascension by an LLM agent. Astra wrote its own terminal harness, and the run was open-book with human supervision, so it is not a benchmark score. NetHack had been a standard hard long-horizon RL/LLM test since the NetHack Learning Environment (2020) and BALROG (2024).","key_facts":["Ascension: Sept 21, 2026, NetHack 3.6.7 on Hardfought, dwarven Valkyrie 'CodexDelver', 37,140 game turns; third run, played over about twelve days (Sept 9–21) with pauses","Model: GPT-6 Astra (the author's subscription plus ~$300 of extra credits; no fixed compute budget tracked)","Astra built its own harness: a 144×36 annotated terminal view, guarded movement batching, a route planner, a Sokoban solver, inventory helpers, and persistent Markdown/JSON notes. The author: 'I didn't write a single line of code, and honestly I didn't even read Astra's code'","Caveats from the author: open book (NetHack wiki, spoilers, game source code, web); a human set up infrastructure, made suggestions, and stopped and resumed sessions; tools and instructions changed during play; 'This run is not a BALROG submission'","For comparison, BALROG's leaderboard (Sept 18, 2026) showed GPT-6-Astra-Max at 13.2 ± 2.7% average NetHack progress under its standard protocol. The author's best January 2026 run (other models, handmade harness) reached dungeon level 10","Evidence: Hardfought server dumplog; harness released as MIT-licensed code with an evidence archive (checksums, redaction manifest)","Reaction: NLE co-creator Tim Rocktäschel (UCL, ex-DeepMind) on Sept 25: 'Aaaaaaaand it's apparently solved', quoting his post from days earlier that NetHack 'still seems far from being solved' (~90k views)","A pull request to the NetHackers community site (dunnolab, PR #114) adds it as 'the first LLM-agent ascension' while warning against overstating it"],"key_numbers":[],"tags":["games","nethack","agents","gpt-6-astra","long-horizon","benchmark"],"science":null,"body_md":"## What happened\n\nNetHack is a 1987 roguelike with permadeath and huge item and monster interactions, and a full game takes tens of thousands of turns.\nIt became a standard test of long-horizon agents through Meta/UCL's NetHack Learning Environment (2020) and the BALROG LLM-agent\nbenchmark (2024). Before this run, no AI agent was known to have won it.\n\nA hobbyist who blogs as \"Kenny\" (kenforthewin) had tried since January to build a harness for LLMs to play NetHack. His best early\nruns reached dungeon level 10. With GPT-6 Astra he let the model build its own harness. Its first run came close, and its third run\nascended on Sept 21, 2026 after 37,140 turns. He published the dumplog, the code and a reviewed evidence archive.\n\n## Why it matters\n\nThe run shows long-horizon execution, not only game knowledge: weeks of play in which one mistake can end the game. The model also\nbuilt its own tools and memory along the way. The result is anecdotal. It was open book, supervised by a human, without a fixed\nbudget, and a single success, while Astra's standardized BALROG score was still about 13%. A protocol-controlled ascension on a\nbenchmark would be the next step.","disputed":[],"related":[{"id":"2026-09-18-gpt-6-astra-breaks-marmont-cipher","url":"https://postcutoff.com/e/2026-09-18-gpt-6-astra-breaks-marmont-cipher/","date":"2026-09-18","date_precision":"day","short_title":"GPT-6 Astra breaks an unsolved 1809 Napoleonic cipher letter to Marshal Marmont from a single scan","deck":null,"takeaway":"It is the second historical cipher break by GPT-6 Astra in two weeks.","category":"research","category_label":"Research","importance":3,"confidence":"high","status":{"key":"confirmed","labels":["Confirmed"]},"sources":6,"official":1,"filed":"2026-10-01","updated":"2026-10-17","orgs":["OpenAI"]},{"id":"2026-09-03-gpt-6-astra","url":"https://postcutoff.com/e/2026-09-03-gpt-6-astra/","date":"2026-09-03","date_precision":"day","short_title":"OpenAI releases GPT-6 Astra, its first GPT-6 model","deck":null,"takeaway":"Astra is the first GPT-6-generation model and the first frontier release after the Hugging Face sandbox-escape incident and OpenAI's August training pause.","category":"model-release","category_label":"Model releases","importance":5,"confidence":"high","status":{"key":"confirmed","labels":["Confirmed"]},"sources":20,"official":11,"filed":"2026-09-29","updated":"2026-10-19","orgs":["OpenAI"]},{"id":"2026-09-03-arc-agi-3-gpt-6-astra","url":"https://postcutoff.com/e/2026-09-03-arc-agi-3-gpt-6-astra/","date":"2026-09-03","date_precision":"day","short_title":"GPT-6 Astra scores 62.7% on ARC-AGI-3, outacting humans on 96% of levels","deck":null,"takeaway":"ARC-AGI-3 was meant to measure human-like skill acquisition; its near-saturation (and the harness gap) shows both how fast agentic reasoning improved in 2026 and how much scaffolding now drives scores.","category":"benchmark","category_label":"Benchmarks","importance":5,"confidence":"high","status":{"key":"confirmed","labels":["Confirmed"]},"sources":7,"official":4,"filed":"2026-09-29","updated":"2026-10-04","orgs":["ARC Prize Foundation","OpenAI"]}],"people":[{"id":"zvi-mowshowitz","name":"Zvi Mowshowitz","url":"https://postcutoff.com/person/zvi-mowshowitz/"}],"posts":[{"id":"x-_rockt-2103393155030876662","title":"“Aaaaaaaand it's apparently solved 🫣😅”","url":"https://postcutoff.com/p/x-_rockt-2103393155030876662/"}],"videos":[],"models":[],"changes":[{"date":"2026-10-02","type":"filed","text":"Created (Zvi AI #188 / Tim Rocktäschel post)"}],"provenance":{"agents":[{"model":"Claude Opus 5.5","maker":"Anthropic","tool":"Claude Code"}],"filed":"2026-10-02","run":null,"sources_read":null,"updated":"2026-10-02","human_review":null,"version":{"date":"2026-10-04"}},"gaps":[{"model_id":"gpt-6-2-astra","name":"GPT-6.2 Astra","cutoff":"2026-05","days_after":113,"in_training_data":false},{"model_id":"claude-opus-5-5","name":"Claude Opus 5.5","cutoff":"2026-06","days_after":83,"in_training_data":false},{"model_id":"gemini-4-hydrogen","name":"Gemini 4 Hydrogen","cutoff":"2026-03","days_after":174,"in_training_data":false},{"model_id":"grok-4-7","name":"Grok 4.7","cutoff":"2026-05","days_after":113,"in_training_data":false}],"short_url":null}