Oh very interesting. I can confirm the overall findings with Qwen3.6 (my blogpost from a few weeks back). I.e., Claude Code uses 2x as many tokens as Qwen Code and Codex CLI
ECONOMY
-
AI Demand Outpaces Available Infrastructure, Energy, and Talent
By
–
-
Continuous Agent Loops Expose Costs Missing from AI Pilot Designs
By
–
The real story is not that AI pilots are expensive. The real story is that most enterprise architectures were designed for demonstration — not for the continuous loops that production agents actually run. A chatbot request costs cents. An agentic task consuming context, memory,
-

Pruna and Ideogram Optimize Image Generation for Typography and Layout
By
–
P-Image-Ideogram from @PrunaAI is here. Built in partner with @ideogram_ai for high-quality stylistic generation with strong typography + layout capabilities. Now, optimized for as low as $0.003/image.
-
Open-Weight Models Could Give Customers More Control Over AI Economics
By
–
Which is another reason. Why I hope that Open-Source (open-weights) win because over time imho it is better for both the customer and overall economy “the frontier lab playbook is simple: 1. sell their models to the world’s most valuable businesses
2. help wire them into those -
Report Links AI Market Losses to a Large Hedge Fund
By
–
BREAKING: Leopold Aschenbrenner’s $24B hedge fund is facing steep AI losses:
-

OpenAI CFO Reportedly Says July Recurring Revenue Surpassed Q2
By
–

OpenAI is on such a run right now: CFO Sarah Friar told employees that OpenAI's annualized recurring revenue in July exceeded all of the second quarter. "And Q2 was no slouch," she said at Wednesday's internal meeting, according to a partial transcript reviewed by CNBC. Friar
-
Open-Source Workflow Reduces Fable 5 Tokens and Errors
By
–
I cut Fable 5 token usage 2.5x with just one change!
— Akshay 🚀 (@akshay_pachaar) 30 juillet 2026
(100% open-source solution)
– Before: 5.5 M tokens · 7 errors · $8.94
– After: 2.3 M tokens · 0 errors · $4.17
The final build was the same for both, but the path the agent took wildly differed.
In both runs, the agent… https://t.co/5Et5Nrsuai pic.twitter.com/k65jI4iXitI cut Fable 5 token usage 2.5x with just one change! (100% open-source solution) – Before: 5.5 M tokens · 7 errors · $8.94
– After: 2.3 M tokens · 0 errors · $4.17 The final build was the same for both, but the path the agent took wildly differed. In both runs, the agent -

Slow Grid Connections Could Constrain Rapid AI Compute Growth
By
–
AI moves at software speed. Power infrastructure moves at engineering speed. Models can scale in weeks. Grid connections and substations may require years. That timeline gap could become AI’s next major bottleneck. https://
go.kenovy.com/QICt Follow @ingliguori #AI -
Pinning Models Through Long Sessions Avoids Unnecessary Routing Costs
By
–
Yess, session unit vs request unit is the whole game
pin for the long job, re-route only when the task actually resets