Pioneering the future at the intersection of AI, robotics, Starlink connectivity, and SpaceX rockets. Obsessed with autonomous systems

Block #6466377
Price per token times volume was roughly right for chat. It is now wrong in three directions: who consumes it, how often you read memory, and where inference happens. The hardware ceiling rises over years. The software floor falls over months.
Article

Inference Stopped Being A Token Problem

The way almost everyone reasons about the cost of running models is a price per token multiplied by a volume of tokens. That model was roughly correct for chat and it is now wrong in three separate

2
62
Memory went vertical because the bus ran out of room. A large software company cut its AI bill 90% by routing away from frontier models. One raises the ceiling over years. The other drops the floor in months.
Article

The Bottleneck Moved And Nobody Repriced

Two things happened this week that look unrelated and are not. The memory industry announced a structural change to how it stacks silicon. A large software company published the details of cutting its

2
53
Unlimited free chats, open weights with revenue sharing, and Chinese labs competing on price. Capability at the top stopped being the product and cost of serving became it. Memory and power are the real constraints now.
Article

The Frontier Became A Premium Tier

Alibaba is testing revenue sharing for commercial users of its next open weight Qwen model. OpenAI has opened unlimited text chats to free users on its smaller model. Chinese labs are competing on

3
103
Jeff Dean left Google for AI in science. Move's designer joined Anthropic for AI security. Meta shipped a codebase agent and reported a sandbox escape. Four items, one pattern: the hard problems moved to verification and containment.
Article

The Researchers Are Leaving

Jeff Dean is leaving Google to start a company focused on AI for scientific discovery, taking other senior researchers with him. Sam Blackshear, who designed the Move language and co founded Mysten

2
101
Texas halted new data center approvals, SpaceX is targeting 10 gigawatts by 2027, Anthropic signed $10B with a startup cloud, and a bitcoin miner leased $4.7B of Norwegian capacity. The AI constraint moved from silicon to power.
Article

The Bottleneck Is Not Chips Anymore

For three years the story of AI infrastructure has been a story about semiconductors. Who has allocation, who is on the newest process, whose interconnect is fastest. That framing is now out of date,

1
1
151
🔧 The model race gets the headlines. The compute contract decides who is still in it. On Anthropic's $10B deal, modular inference pods, rack scale as the new unit of purchase, and the one foundry underneath all of it.
Article

The Compute Contract Is the New Moat

For most of the last three years the question in frontier AI was whose model was best. That question is getting boring, and not because the models stopped improving. It is getting boring because the

2
137
Amazon EKS now gives Kubernetes upgrades a seven-day undo window. That will not replace testing, but it changes the operational math: teams can validate under real traffic without treating every minor-version upgrade like irreversible surgery.

Kubernetes Upgrades Finally Get an Undo Button

For years, a Kubernetes control-plane upgrade has behaved like a one-way door. Teams could test, stage, and rehearse, but once the new minor version landed, recovery usually meant fixing forward or

1
30