Mitchell Bryson

Today in AI — 23 March 2026

Today's top AI news — curated links and commentary on the stories that matter for product builders.

·3 min read

The economic assumptions underpinning AI are being rewritten. Nvidia is paying engineers in inference tokens, open-source models match frontier performance at a fraction of the cost, and enterprise spending has tilted sharply in Anthropic's favour. Together, these shifts point to an AI economy where compute, not headcount, is becoming the primary unit of value.

Compute as compensation

When token spend per head approaches salary, the maths start pointing toward fewer heads.

OpenAI: hire fast, IPO faster

Twelve new hires per day to hit 8,000 by year-end, a ChatGPT productivity pivot, and a Q4 IPO at up to $1 trillion. Headcount now, margins later.

Models commoditise

OpenClaw, MiMo-V2-Pro, and MiroThinker 72B each took a different route to near-frontier performance at a fraction of the cost. The proprietary moat is eroding quarter by quarter.

Enterprise spending reshuffles

Anthropic now captures 73% of first-time enterprise AI buyers; Salesforce quietly added 6,000 Agentforce customers in a single quarter. Returns are real, just not always where people expected.

Agents go multi-channel

Agent deployment matters as much as model training now. The infrastructure layer is filling in from edge inference to messaging integrations.

Content ownership under siege

AI-rewritten headlines, copyleft erosion, and copyright lawsuits: three fronts in the same fight over who controls content when models can reproduce and reframe it at will.

If you're building on frontier models, the pricing and competitive assumptions you made six months ago may already be wrong.

Stay up to date

Get notified when I publish something new, and unsubscribe at any time.

More news

Compute is the new equity

In a single week, Nvidia proposed paying engineers half their salary in AI inference tokens, OpenAI announced plans to nearly double its headcount to 8,000, and data showed Anthropic capturing 73% of new enterprise AI spending with a fraction of the workforce. The AI industry is splitting into three competing theories of value: raw compute, human talent, and enterprise relationships, and Jensen Huang's token paycheck is the clearest signal yet that the boundaries between them are collapsing. When your raise comes in tokens and your productivity is measured in inference, the employment contract is being rewritten in real time.

Today in AI — 21 March 2026

Today's top AI news — curated links and commentary on the stories that matter for product builders.

The superapp moment has arrived

OpenAI's Fidji Simo told employees the company was 'spreading efforts across too many apps.' Google is sunsetting Firebase Studio to fold everything into AI Studio. Meanwhile, Carta data shows the same consolidation logic playing out in capital markets: 10% of AI startups now capture half of all venture funding. After two years of tool proliferation, the platforms have decided the winner isn't the best model — it's the best starting screen.