GoKawiilDeveloper finds GPT-6 'Astra' burns billions of tokens with no usable code output
A software engineer tested OpenAI's GPT-6 Astra model by setting up an autonomous 'software factory' where the model managed its own workflow, context and subagents to build a Python variant with virtual threads and lexical scoping. After roughly 35 hours and about 4 billion tokens consumed, the effort produced nothing usable and taught the author little about how to actually apply the model to real software engineering work.
GoKawiil's interpretation of the reporting above, not reported fact.
The experiment highlights a growing gap between benchmark-impressive AI models and their practical usefulness for autonomous, long-horizon coding tasks. It also underscores the steep and possibly wasteful token costs of letting frontier models self-direct complex engineering work without human guardrails, raising questions about whether more capable models actually translate into more productive output.
- GPT-6 Astra excelled at computer use and image/topic understanding but failed to deliver usable code in an autonomous coding test.
- The self-directed 'software factory' experiment consumed about 4 billion tokens over 35 hours with zero useful output.
- The result raises doubts about whether increasingly powerful AI models are translating into real productivity gains for software engineering.
Agricultural Involution by Clifford Geertz — The article's whole framing of AI engineering as "involution" comes straight from Geertz's classic study of Javanese agriculture, where ever-more effort per plot yields no gain per person. Reading the original makes the Neijuan metaphor click in a way a tweet thread can't, and it's a sharp lens for thinking about diminishing returns in any intense field.
See Agricultural Involution by Clifford Geertz on Amazon → Affiliate link — we may earn a commission on purchases, at no extra cost to you. Product picked by AI based on this article; it is not a tested recommendation.Source: lucumr.pocoo.org, 2026-09-11
Published there as: “Astra for Coding: Why Are We Doing This Again?”
Read the original report → The summary and analysis above are GoKawiil's own, written from reporting by the source above. Facts and quotes belong to the original publisher.