ECONOMY - AI Dynamics

Kimi K3 Approaches Sol at Slightly Lower Cost per Task

By

–

16 July 2026 21h28

Here is the chart where you can see how Kimi K3 stays very close to GPT 5.6 Sol at a slightly lower cost per task. [Traduit de FR vers EN]

→ View original post on X — @dotcsv

16 July 2026

A Cost-Performance Comparison With GPT-5.6 Sol Max

By

@dotcsv

–

16 July 2026 21h28

That, of course, also affects its price, which on this occasion could not be as competitive against private models as in previous releases. In cost-performance terms, it’s very close to GPT 5.6 Sol (max). [Traduit de ES vers EN]

→ View original post on X — @dotcsv

16 July 2026

Use a Larger Model for Planning and Smaller Models for Execution

By

@theahmadosman

–

16 July 2026 18h21

I do and my weekend plans are already made to implement it for my hardware hahaha Answering your question: most folks will not need this model except for high-level planning and then smaller models will execute just fine: paid api for those needs

→ View original post on X — @theahmadosman

16 July 2026

Electricity Is a Fundamental Input to AI Token Generation

By

@ninadschick

–

16 July 2026 15h42

At its core, the ‘Intelligence Explosion’ is a conversion of energy to tokens (IE non-biological Intelligence). This the alchemy of our age, and it is why electricity is the fundamental input @elonmusk understands this, and it is why he is launching orbital compute to

→ View original post on X — @ninadschick

16 July 2026

AI Chip Suppliers’ Growth Raises Questions About Sustaining Spending

By

@tiernanraytech

–

16 July 2026 15h38

The better things get in the outlook of AI suppliers such as TSM, ASML, and Micron, the more it causes people anxiety about where all this spending is heading, and whether it can be sustained. Taiwan Semi sells off on higher spending despite stellar outlook The forecast from

→ View original post on X — @tiernanraytech

16 July 2026

AI21’s Hybrid Coding Pipeline Reduces Frontier Model Spending

By

@ai21labs

–

16 July 2026 13h16

3/3 Frontier is just 25% of the budget, so the pipeline runs ~⅓ the cost of a frontier agent end-to-end while beating it on quality. Well ahead of other hybrids too: closest comparison hit 66.1% (60-task subset vs. our full set). Full write-up: https://
ai21.com/blog/better-an
d-cheaper-together-open-models-explore-frontier-models-patch/?utm_source=org-twitter
…

→ View original post on X — @ai21labs

16 July 2026

TSMC Reports Higher Revenue and Profit Amid the AI Boom

By

@kimmonismus

–

16 July 2026 12h06

TSMC Q2 2026 Results (As always, the biggest winners of the AI boom are the "shovel sellers", TSMC and NVIDIA.) -Revenue: NT$1.27T (~US$40B) (+36% YoY)
-Net profit: NT$706.6B (~US$22B) (+77% YoY, record high)
-Beat analyst consensus of NT$632.6B by 11.7%
-Gross margin: 67.7%

→ View original post on X — @kimmonismus

16 July 2026