The challenge of the "social productivity paradox" in AI and our ideas about how to address it with an "attention reinvestment cycle" are now available in Science and Public Policy.
ECONOMY
-

Opus Uses Multiple Agents and Excessive Tokens for an Infographic
By
–

Opus is treating my cheese infographic with the seriousness of a new Erdos problem. We are a million tokens and 5 agents in.
-
GPT-5.6 Update Aims to Improve Token Efficiency by 18%
By
–
Another reset and a fix that should make GPT-5.6 18% more token efficient. Both are very welcome. The effort OpenAI is currently putting into Codex and its model is impressive. (I’d like to know how much money the repeated resets cost OpenAI. But surely millions.)
-
OpenAI Reports Improvements That Extend Typical Sol Usage by 18%
By
–
Hola! quick usage update on usage patterns for Sol: We’ve shipped several improvements that should make typical Sol usage last ~18% longer, with significantly larger gains for some power users. The five-hour limit will also return tomorrow. The reason: Sol works longer, calls
-

Estimate Explores Revenue Needed to Justify Big Tech AI Investment
By
–
“I assume that big tech firms need to cover depreciation and make a small profit on each unit of AI capex. This gets me to $2.5trn of "AI revenues" every year in perpetuity… so far we are not anywhere near that…”
-

AI Spending Concerns Weigh on Big Tech Cash Flow and Stocks
By
–
“Big tech has turned from a cash machine to something that, on net, consumes cash. Recent declines in AI stock prices reflect these concerns.”
-

Databricks Discusses Shifting Enterprise AI from Tokens to ROI
By
–
Arsalan Tavakoli, Databricks Co-Founder and SVP of Field Engineering, joined @Bloomberg The Close to discuss how the AI conversation is shifting from token maxxing to value maxxing as enterprises increase focus on ROI, governance, and choosing the right model for each use case.
-
Local Kimi K3 and Fable Show Different Token and Game Results
By
–
Tokens and cost stats from the test > Kimi K3, local on 8x B300: ~237K tokens, $0
> Claude Fable 5: 92K tokens, $4.50 Kimi's version of the snake came out as the most detailed scene of the six, with soft shadows, floating dust, and a bot that checks it can still reach its -

Benchmark Indicates Opus 5 May Use Many Tokens
By
–

I can confirm with our benchmark as well. Opus 5 seems to eat tokens like crazy!
-
Central Europe’s Opportunity in Artificial Intelligence
By
–
Capturing Central Europe’s AI opportunity
#AI #Economy #AIInnovation #MachineLearning #Policy @timnitgebru @oriolvinyalsml @ceobillionaire @soumithchintala @waitin4agi_ @sallyeaves @bernardmarr