@techtalkjp: You can probably build a software factory in about a week. Once you're confident it's reliable, take it to production. Output goes up 10x. But at the same time, token costs and CI costs also go 10x. If revenue doesn't change as a result, then you've just made costs 10x higher, meaning productivity in the true sense drops to 1/10...
Summary
A reflection on the economics of the AI coding era: building a software factory in one week and taking it to production can boost output 10x, but token costs and CI costs also rise 10x — so if revenue doesn't increase, real productivity effectively drops to 1/10.
View Cached Full Text
Cached at: 10/18/26, 05:07 AM
A software factory can probably be built in about a week. Once we determine it’s reliable, we go to full production. Output increases 10x.
At the same time, token costs and CI costs also go up 10x.
If revenue doesn’t change, we’ve just increased costs by 10x — meaning our true productivity has dropped to 1/10.
What do we do?
Similar Articles
@FinanceYF5: Chamath said he asked the CTO about the company's AI token spending today and got a shocking answer: token costs double every 45 days, but downstream productivity improves by at most 5%. His exact words were that costs double, benefits stay roughly flat, and the team found that to iterate to the next generation of capabilities, the amount of tokens required grows exponentially, because...
Chamath revealed that his company found AI token costs double every 45 days while downstream productivity improves by at most 5%. He believes AI development is approaching a bottleneck and suggests companies reconsider their strategies or even consider exiting.
@0xCheshire: Chamath just revealed a deeply unsettling truth for the AI industry. He asked his own CTO to review the company's spending, and the result was staggering: "Our token costs are doubling every 45 days," yet the downstream productivity gains are at most about 5%. Costs are skyrocketing exponentially, while returns remain basically flat...
Chamath reveals the harsh reality of AI costs vs. returns: token costs double every 45 days, but downstream productivity gains are at most 5%. Large model capability improvement has hit an asymptote, and within the next 3-4 years, every company will face an ultimate reckoning between cost and benefit.
@seclink: https://x.com/seclink/status/2057086514975404108
Anthropic engineer shared development experience of long-duration AI agents, including multi-role division and independent evaluators, enabling AI to automatically generate complete applications within 3-5 hours, with a 12x improvement in continuous operation capability.
Where is AI boost in software?
作者质疑AI对软件行业的实际影响：如果软件工程真的被AI解决，理应看到软件降价、厂商节省成本或功能增加，但目前这些现象都没有明显出现，原因是什么？
@rohanpaul_ai: Productivity gains from AI are not linear, they accelerate with depth of use. Workers who use advanced ChatGPT capabili…
OpenAI的研究表明，AI的生产力提升并非线性，而是随着使用深度加速增长。重度用户（使用GPT-5 Thinking、Deep Research等高级功能的用户）每周节省的时间远超轻度用户，并且使用的AI积分是后者的8倍。