A company's numbers do not say what it was built to do. Revenue, margins and growth describe what it is doing now. They are silent on what it is for, and that silence is where a good many investments go wrong.

The author's great-grandmother lost her hearing near the end of her life, but her mind remained fully functional, making it impossible for her to participate in or follow the conversations going on around her. This essay begins with that defeat and asks why the world's largest corporations are rushing to develop humanoid robots, and whether logistics, a task a basic overhead rail can perform, is the real solution. It asks what it would take for someone to continue in a new body, and who gets to decide, drawing on Atlas, Optimus, Asimov, Gibson, and the Ship of Theseus. Intro by Mackenzie Swain, reviewer.

A trade essay says America's machine-shop capacity is stranded by the quote. Companies have answered it in three ways, and some of the capacity sits in garages and spare rooms, owned by people who want self-sufficiency, not customers.

Machines have turned human language into numbers by three routes: hand-written rules, taxonomies, and learned statistics. Each moved where the structure lives, and the last returns the specification to plain English, so a person still has to check the answer.

Want to know if an institution is any good? Ignore what it does. Watch where its people go next. After a decade inside In-Q-Tel, the CIA-backed venture firm, Troy M. Pearsall traces his old colleagues to NASA, Lockheed Martin, and venture firms. He argues that when a cohort reassembles somewhere new, it's a stronger signal than any single résumé. Intro by Mackenzie Swain, reviewer.

Secondhand servers, drives and memory are the bottom rung of a chain priced at the top. AI buyers moved the top, and the price is coming down the chain.

The software around an AI model does much of the work, but the most valuable part is the read: judging a machine's output, a room or a person. That part stays with us.

A company's name is the first claim it makes, before any pitch. How founders pick names, what a name tells the world about what they are building, and why the explanation is worth asking for.

A mold turns a large fixed cost into a tiny cost per part and wins if the design holds and the parts sell. A printer has a flat cost per part and wins if the design keeps changing. The line between them moves when someone builds a bigger farm.

A controlled layer lasts only as long as leaving it stays costly. A look at how Unix, the Mac's processors, Alpha, Linux and Windows on Arm each loosened, and what a bridge or an open alternative does to the cost of leaving.

The full walk behind "Nothing Is Alone": the evidence, the people, what synthesis costs, and where the author is likely to be wrong.

Why a cold place with heavy floors and power, like a shipyard town in Maine, could hold an AI data center: the physics of the rack, who has to retrofit, the capital and staffing arguments, Maine's moratorium fight, and Newfoundland and Iceland. For investors.