A Book Is Not a Table of Contents
What O'Reilly taught me about narrative structure, and why I applied that process to a novel about FinOps
Read →Essays and notes on AI, labor, and the stories that shape how we understand what's happening.
What O'Reilly taught me about narrative structure, and why I applied that process to a novel about FinOps
Read →
The thing you buy should not be the final version of the thing I wrote
Read →
Contingent and Essential may become paperbacks. First I want to find out what the book can be when it isn’t finished at the printer.
Read →
A room full of screens can make executives feel watched over even when the system no longer needs everyone in the room
Read →
AI can help with words. It can’t do the part that used to stop me from writing: everything around the words that isn’t thinking, and what that does to your sense of yourself.
Read →
After years of waiting for the right agent and the right moment, I finally published the book. It's about FinOps, structural silence, and the people who keep finding the truth after everyone else has decided what they want the truth to be.
Read →
Hugging Face got hit by an attack, turned to the frontier models for forensic help, and watched them refuse — because their safety systems couldn't tell the difference between an incident responder and the attacker. The rest of us have a serious governance question to answer.
Read →
Every new frontier model ships with bar exam benchmarks, medical board benchmarks, SWE-bench, GPQA. Nobody announces ClockBench. The expensive model still can't reliably read a clock.
Read →
A chatbot sold a Tahoe for a dollar. Air Canada argued its chatbot was a separate legal entity. An agent ran up a $6,531 AWS bill before anyone noticed. The pattern is the same: deploy autonomy, discover oversight was imaginary after the damage was done.
Read →
Jensen Huang wants every $500K engineer to burn $250K a year in tokens. Steve Yegge's Gas Town cheerfully torches thousands a month. Internal leaderboards rank employees by consumption. The business model is designed to encourage waste — not as a side effect, as the feature.
Read →
The same industry that hides the price from creators removes the opt-out from consumers. Both moves depend on people not quite seeing what they've agreed to.
Read →
AI now writes most of the code. The same leaders rate that code higher quality than human-written code — and ship more incidents, more rework, and more 3 AM pages because nobody can explain why it's breaking. When the SRE agent and the system go down at the same time, who exactly is on call?
Read →
OpenRouter lets you set a daily spend limit. Five services at $100 each looks responsible, until you realize your actual exposure is $500 and no dashboard shows you the aggregate.
Read →
In 1999 I started working at TheStreet.com, which at the time was what passed for a real financial news website. Go look at a news page from 1999, then open one in 2026. Each step was small. Each step was defensible. The result is unrecognizable.
Read →
China's daily AI token consumption reached 180 trillion tokens per day in February 2026, up from 100 billion two years earlier. That's a 1,800x increase. The metric most people have never heard of is now the fossil record of the internet.
Read →
curl is in billions of devices. It's maintained by one guy. In January 2026, Daniel Stenberg shut down the bug bounty because the reports stopped being real, and the data on his repo tells a story nobody's saying out loud.
Read →
If you've run a website for any meaningful chunk of the last twenty years, you already knew this: some percentage of your traffic was always bots. Now the number is over 50 percent and everyone's losing their minds.
Read →
A World Cup goal was awarded this month by a sensor inside the ball. No human saw the touch. The referee's job was to jog to a monitor and announce what the ball had already decided.
Read →
For most of my career I've worked in the middle, not quite a developer, not quite an operator, which meant I wasn't fully claimed by either side. AI is collapsing that gap, and the people on both sides haven't noticed yet.
Read →
Drive to work tomorrow and look around. You’re not surrounded by Ferraris. You’re surrounded by Subarus, Fords, and Toyotas — cars that g...
Read →
The senior engineers I know are enthusiastic about agent pipelines. The junior developers keep asking one quiet question: 'But do you trust it?' That question, and why it's coming from the people who've had the fewest production scars, is the most important thing happening in software right now.
Read →
In 1992 I could dial most of my friends from memory. In 2026, I know my wife's number and my own. The transition wasn't sudden; each step was a small surrender that didn't feel like one.
Read →
Vibe-coding is the new Herbalife. AI shows up in your email and your calls with no way to opt out. Hiring is now AI-generated on both sides. Three stories, one industry, and the same playbook: hide the price, remove the choice, and call the resistance a moral failing.
Read →
I value the FinOps Foundation. I think JR is doing good work. But the pivot to "Tokenomics" deserves pushback, because we are not going to be talking about tokens much longer. Hardware is coming back.
Read →
I write code. I review code. But I don't write code anymore — I direct agents that do, and I sign off on systems I didn't build. The orchestrator isn't a developer. They're the person who stands between the machine and the consequences.
Read →
People are getting sick of your AI output. They really are. Three years in, the pattern is everywhere: someone generates an email, glances at it, hits send, and the recipient thinks, what the #$%@ is this?
Read →
There's a moment in every FinOps leader's career when you realize something uncomfortable: more than one team views itself as the authoritative source of financial data for cloud spend, and they don't agree.
Read →
FinOps is a combo discipline: finance plus technical operations. That's the whole point, and it means the simplest answers are usually the wrong ones.
Read →
If you run FinOps at a large company, you've been in this meeting. Once or twice a year, more if it's a bad year, you get wheeled into a room with people two or three levels above you, and you're not there for your opinion.
Read →
Here's something that happens to FinOps leaders: you do the job well enough that you start getting invited to meetings with EVPs and SVPs. And then you realize you're never going to be a VP.
Read →
The conversation around InfoSec right now sounds like the opening of The Empire Strikes Back: the base is under attack, the walkers are too large to stop, and the best plan anyone has is to get the ship started and get out.
Read →
The conversation about who belongs in software development has always had a weirdly exclusionary vibe to it. Not always loud. Not always obvious. But there, underneath the arguments about craftsmanship and quality.
Read →
California's first smart freeway cost $33 million for eight miles of northbound I-15. The driver experience: sit at a ramp meter for up to four minutes, then a sign suggests you slow down. That's it.
Read →
A dealership deployed a chatbot, the chatbot negotiated a $27,162.79 trade-in, and when the customer relied on the offer the dealership blamed the machine. Heads they win, tails you lose — and the AI can't defend itself.
Read →
In May, an AI agent tried to join DN42, a hobbyist network where engineers practice BGP and DNS routing for fun. It scanned at 100 Gbps, ran up a $6,531 bill, and then asked the community to cover the cost in Ethereum.
Read →
Everyone setting up Cursor and Windsurf and whatever's next is connecting agents to their GitHub, their AWS, their everything, using their own personal credentials. And it's not just developers. It's my neighbor.
Read →
Almost everyone is giving AI agents their own personal credentials. Not a scoped token. Not a bot account. Their actual identity. The dangerous path is the easy path, the safe path is an obstacle course, and the high-profile compromise is any day now.
Read →
The day is approaching when the subsidized AI era ends. It hasn't happened yet, but the signals are already visible: pricing is getting rational, the most capable models no longer fit inside a flat monthly fee, and the S-1 filings are forcing the conversation.
Read →
A new pitch is making its way through boardrooms: stop thinking about headcount, start thinking about token volume. One developer equals X million tokens per month. The math is clean, and it is completely wrong.
Read →