conv.

All stories
AIQuiet 11d · day 11

GPT-5.6 Sol completes Factorio peaceful run in 13.5 hours

Researcher Ryan Madden demonstrates OpenAI's latest model autonomously playing and beating the factory-building game.

What to know

  • GPT-5.6 Sol completed a full Factorio peaceful-mode playthrough in 13.5 hours, showing improvement over GPT-5.5 but revealing persistent weaknesses in late-game scaling and transport belt logic.
  • The model relies on manual resource ferrying and chest-based inventory rather than optimized belt routing—the core mechanic human players prioritize—suggesting incomplete understanding of Factorio's systems.
  • Commenters debate whether the achievement is directly comparable to human play, noting that peaceful mode removes time pressure and LLMs have access to all published guides and videos.

The dispute Whether Sol's ability to complete Factorio demonstrates genuine problem-solving capability or merely skilled synthesis of training data, and what hidden constraints (peaceful mode, information access) mean for fair comparison with human performance. · positions read across 15 posts and comments

most voices

The demonstration reveals genuine AI reasoning but exposes core gaps in mechanical understanding.

  • “It doesn't seem to understand the core mechanic of Factorio, which is putting outputs onto belts and routing them to other inputs.”

    fasterik · Hacker News ↗
many voices

The comparison is unfair because peaceful mode and training data access eliminate constraints humans face.

  • “Putting into "peaceful" mode makes response time a non-factor... Most humans, especially on first play, learn game systems and interactions from scratch, then develop novel strategies and plans and "hypothesis test" their way into beating…”

    program_whiz · Hacker News ↗
some voices

The work is valuable for understanding AI behavior and has real-world applications beyond gaming.

  • “I'm optimistic about autonomous Factorio play as a medium for building intuition about the nature of model capabilities.”

    Ryan Madden · ryanmadden.net ↗

Ryan Madden Researcher / bloggerGPT-5.6 Sol AI model under test

GPT-5.6 Sol completes Factorio peaceful run in 13.5 hours
ryanmadden.net

How it unfolded 3 developments, newest first · click a bar or a number to jump articlespostscomments

Peak 12 pieces in 3h at Oct 31, 1 PM; 17 pieces over 11 days (1 article · 1 post · 15 comments) Oct 31, 1 PM — 12 pieces · 1 article · 1 post · 10 comments — Hacker News 11, Newswires 1Oct 31, 4 PM — 3 pieces · 3 comments — Hacker News 3Oct 31, 7 PM — 1 piece · 1 comment — Hacker News 1Oct 31, 10 PM — quietNov 1, 1 AM — 1 piece · 1 comment — Hacker News 1Nov 1, 3 AM — quietNov 1, 6 AM — quietNov 1, 9 AM — quietNov 1, 12 PM — quietNov 1, 3 PM — quietNov 1, 6 PM — quietNov 1, 9 PM — quietNov 2, 12 AM — quietNov 2, 3 AM — quietNov 2, 6 AM — quietNov 2, 9 AM — quietNov 2, 12 PM — quietNov 2, 3 PM — quietNov 2, 6 PM — quietNov 2, 9 PM — quietNov 3, 12 AM — quietNov 3, 3 AM — quietNov 3, 6 AM — quietNov 3, 9 AM — quietNov 3, 12 PM — quietNov 3, 3 PM — quietNov 3, 6 PM — quietNov 3, 9 PM — quietNov 4, 12 AM — quietNov 4, 3 AM — quietNov 4, 6 AM — quietNov 4, 9 AM — quietNov 4, 12 PM — quietNov 4, 3 PM — quietNov 4, 6 PM — quietNov 4, 9 PM — quietNov 5, 12 AM — quietNov 5, 3 AM — quietNov 5, 6 AM — quietNov 5, 9 AM — quietNov 5, 12 PM — quietNov 5, 3 PM — quietNov 5, 6 PM — quietNov 5, 9 PM — quietNov 6, 12 AM — quietNov 6, 3 AM — quietNov 6, 6 AM — quietNov 6, 9 AM — quietNov 6, 12 PM — quietNov 6, 3 PM — quietNov 6, 6 PM — quietNov 6, 9 PM — quietNov 7, 12 AM — quietNov 7, 3 AM — quietNov 7, 6 AM — quietNov 7, 9 AM — quietNov 7, 12 PM — quietNov 7, 3 PM — quietNov 7, 6 PM — quietNov 7, 9 PM — quietNov 8, 12 AM — quietNov 8, 3 AM — quietNov 8, 6 AM — quietNov 8, 9 AM — quietNov 8, 12 PM — quietNov 8, 3 PM — quietNov 8, 6 PM — quietNov 8, 9 PM — quietNov 9, 12 AM — quietNov 9, 3 AM — quietNov 9, 6 AM — quietNov 9, 9 AM — quietNov 9, 12 PM — quietNov 9, 3 PM — quietNov 9, 6 PM — quietNov 9, 9 PM — quietYesterday, 12 AM — quietYesterday, 3 AM — quietYesterday, 6 AM — quietYesterday, 9 AM — quietYesterday, 12 PM — quietYesterday, 3 PM — quietYesterday, 6 PM — quietYesterday, 9 PM — quietToday, 12 AM — quietToday, 3 AM — quietToday, 6 AM — quietToday, 9 AM — quietToday, 12 PM — quietToday, 3 PM — quietToday, 6 PM — quiet 1–3
Nov 1Nov 2Nov 3Nov 4Nov 5Nov 6Nov 7Nov 8Nov 9yesterdaynow · 9:51 PM ET
  1. 3

    Discussion identifies limitations of the demonstration

    Commenters point out that peaceful mode removes time-pressure constraints and that LLMs have access to all published Factorio guides and videos—advantages human players rarely have on first playthrough—making direct capability comparison difficult.

    “Putting into "peaceful" mode makes response time a non-factor... Most humans, especially on first play, learn game systems and interactions from scratch, then develop novel strategies and plans and "hypothesis test" their way into beating the game.”
    — program_whiz
    1. first by HN Frontpage, 11d ago

    • One limitation of these type of demos (and similar with Minecraft, etc) is that: 1. Putting into "peaceful" mode makes response time a non-factor 2. These models have read and have access to every guide and video ever published on the subject. Most humans, especially on first play, learn game systems and interactions from scratch, then develop…

      program_whizHacker News11d agoview on Hacker News ↗
    2 more of the top 3 · 9 posts in this stretch
    • Honestly, I'm more interested in the harness. Having an LLM gardener that could expand outposts to connect to the main network, or just pop in to observe if there's any major problem to report could be a cool project for long-running multiplayer bases.I see it at https://github.com/ryanmadden/autorio and…

      ArcHoundHacker News11d agoview on Hacker News ↗
    • > The rocket launch was only achieved through significant ‘manual’ effort on Sol’s part as it papered over its late-game failures to connect certain production systems by repeatedly ferrying supplies between areas of the factory.Looking through the video, around 12m one can see the Sol trace report the cancellation of a conflicting 'courier'…

      MajromaxHacker News11d agoview on Hacker News ↗
    all of them →
  2. 2

    Commenters highlight Sol's unconventional factory design

    Hacker News discussion emerged with multiple commenters noting that Sol's approach—relying on chests and manual insertion rather than belt routing—differs markedly from typical human strategies and raises questions about whether the model truly grasps Factorio's core mechanics.

    “It doesn't seem to understand the core mechanic of Factorio, which is putting outputs onto belts and routing them to other inputs. It does this in a few places, but for the most part everything is put into chests and moved around by the player.”
    — fasterik · source
    • It is weird and almost alien how AI designs factories. When Claude announced Fable 5 they also did a Factorio demo that was the same kind of designs and a lot of hand feeding and direct insertion. In some sense it proves to me that llms are not just copying and are actually creating a bespoke solution. But I honestly expected more from frontier…

      kanemcgrathHacker News11d agoview on Hacker News ↗
    2 more of the top 3 · 6 posts in this stretch
    • > The rocket launch was only achieved through significant ‘manual’ effort on Sol’s part as it papered over its late-game failures to connect certain production systems by repeatedly ferrying supplies between areas of the factory.Robots: they're just like us!

      mwtHacker News11d agoview on Hacker News ↗
    • It doesn't seem to understand the core mechanic of Factorio, which is putting outputs onto belts and routing them to other inputs. It does this in a few places, but for the most part everything is put into chests and moved around by the player.

      fasterikHacker News11d agoview on Hacker News ↗
    all of them →
  3. 1

    Sol shows improvement over GPT-5.5 but faces scaling limits

    Madden noted that Sol surpassed GPT-5.5's performance (which failed to use transport belts) but encountered scaling issues in the blue science stage that slowed progress significantly. The final rocket launch required extensive manual effort by Sol, repeatedly ferrying supplies between disconnected factory areas rather than optimizing production systems.

    “The rocket launch was only achieved through significant 'manual' effort on Sol's part as it papered over its late-game failures to connect certain production systems by repeatedly ferrying supplies between areas of the factory.”
    — Ryan Madden
  4. background

    Madden publishes GPT-5.6 Sol's Factorio completion — Ryan Madden released results showing GPT-5.6 Sol (running in Codex with low thinking level) successfully completing a peaceful Factorio run in 13.5 hours of game time using a Python game manager that enforces human-like constraints such as player reach and fog of war.

What people are saying 9 voices from 1 site · best of 15 · verbatim