What I am seeing in current builds, frontier AI, and software calls. Written so an operator can decide what to do next.
How we apply this
FeaturedAug 28, 20266 min read
Newsletter
Practical AI, shipping, and governance notes. No spam. Unsubscribe anytime.
Browse by focus
The three piles I write in most. Counts are matching notes in the archive.
AI Policy
The U.S. Treasury Department launched an AI Innovation Series with four roundtables to push AI adoption in finance. The real headline? They're now treating failure to adopt AI as a threat to financial stability.
AI Industry
OpenAI plans to nearly double its headcount to 8,000 by end of 2026, fresh off an $840 billion valuation. Meanwhile, Sam Altman hit the internal panic button over Google Gemini 3. Here's what this hiring spree actually tells us.
AI Industry
Microsoft's in-house image model just landed at #3 on Arena.ai, right behind Google and OpenAI. The photorealism is real. The restrictions are brutal. And the strategic implications are bigger than the model itself.
AI Industry
Hunter Alpha appeared on OpenRouter with no name attached, topped the leaderboard, and had everyone convinced it was DeepSeek V4. Turns out it was Xiaomi's MiMo-V2-Pro, and that changes a few things.
AI Industry
Meta is reportedly planning to cut 20% of its workforce — roughly 16,000 people — while pouring $600 billion into AI data centers. The math is brutal but simple: companies are betting that AI replaces headcount, not supplements it.
AI
China's 15th Five-Year Plan is 141 pages of ambition: humanoid robots, brain-computer interfaces, flying cars, quantum tech, 6G. The term 'artificial intelligence' appears 50+ times. With 2 million robots already in Chinese factories and 150+ humanoid robot companies, this isn't a wish list. It's a procurement order.
AI
Beijing coined a new term — 'intelligent economy' — and put a $1.4 trillion price tag on its AI industry by 2030. The 15th Five-Year Plan isn't vague aspiration. It's a national blueprint with specific targets, including 90% AI agent penetration. Here's why that matters for everyone else.
AI
Google embedded Gemini directly into Chrome this week, rolling out globally. Summarization, shopping, cross-app context. Plus Meta's AI glasses have a privacy problem nobody wants to talk about.
AI
Perplexity just announced Personal Computer — an always-on AI agent running locally on a Mac mini, 24/7. As someone who runs an always-on agent daily, here's what this actually means for the market.
AI
Yann LeCun left Meta, launched AMI Labs with 12 people, and just raised over $1 billion at a $3.5B valuation. He thinks transformers are a dead end. Is he right, or is this the bubble talking?
AI Ethics
Anthropic got blacklisted. OpenAI took the Pentagon deal. 900 employees signed a petition. A robotics chief resigned. Dario and Sam are trading public insults. This stopped being a policy debate and became something much uglier.
OpenAI
Caitlin Kalinowski resigned as OpenAI's head of robotics after the company rushed a deal to deploy AI on the Pentagon's classified networks. Sam Altman says the optics don't look good. No kidding.
OpenAI
OpenAI just shipped GPT-5.4 with native computer use baked in. Two days after GPT-5.3 Instant. The model can autonomously operate across apps on your machine, includes financial plugins for Excel and Google Sheets, and is already live on OpenRouter. OpenAI is shipping at a pace that should make everyone pay attention.
AI News
Meta is paying News Corp $100M to train on journalism. Perplexity's browser just got caught with a nasty security flaw. Agentic AI is moving on-prem. And a new study says consumers aren't experimenting with AI anymore, they just use it. Four stories from one day that tell you where we actually are.
AI Infrastructure
CoreWeave just locked in a multi-year deal with Perplexity AI for inference workloads. Stock popped 8%. Meanwhile, Perplexity's own Comet browser has a nasty prompt injection bug, and Broadcom is projecting $1B+ in AI chip revenue by 2027. The infrastructure layer is where the real bets are being placed.
AI Policy
The Pentagon just classified Anthropic as a supply-chain risk to national security, forcing defense contractors to purge Claude from their systems. OpenAI gets the green light. Meanwhile, hundreds march through London screaming "Stop the slop." Welcome to March 2026.
