Writing
Notes on engineering leadership, platforms, and applied AI. Direct, and built from what actually shipped.
Timeline
The worst part of managing engineers is the nagging. So I built a bot to do it - privately - and stopped being the villain of my own team.
The standard story says you go up the ladder and lose the keys to the codebase. Agents broke that trade. The bottleneck moved off typing and onto judgment, which is the one thing a decade of management actually sharpens.
A meeting transcriber is one API call and a two-quarter product. The transcription is a commodity. The privacy, the identity, the ingestion, and the authz are the actual work, and that is exactly the work agents now let one person do.
Everyone wants the AI ROI number. Finance wants cost per team. You can get it. You cannot get it clean. Attribution is the entire game, and the moment a fuzzy number is allowed to rate a person, the data dies.
The first move when a team isn't delivering is almost never to hire. It's to find out why (ownership gaps, the wrong people, too much noise) and fix that.
Everyone has a loud opinion on whether AI coding tools help. Almost nobody has the data. So I built a tiny thing to get it, used it, and killed it.
Hundreds of "review this for security" sessions in my logs looked like noise. They were a review gate I hadn't bothered to trust. So I wired them as one.
Running Claude, Codex, and Gemini in parallel beats standardizing on a single model. Three different blind spots rarely line up on the same miss.
I ran a hiring overhaul that worked everywhere except the one place I cared about. You cannot fix a salary band by interviewing more charmingly.
The fastest way to lose your best engineer is to promote him. A case study in doing exactly that.
"We need to talk" is the scariest four words a manager hears. How to mine the honest feedback, and why the best exit interview is the one you never run.
Two phases both get called planning, and merging them is how good teams build the wrong product beautifully. The stakeholder you forget is the one who fills your queue.
Plan the backend now, or fight the dumpster fire at 3am. When the conception phase earns its keep, and when to skip it.
You will not out-argue the PM who wants you to "make it pop." How to work with the difficult one, minus the sabotage fantasies.
I said yes to everything until it nearly broke me. Then I learned to decline on purpose.
You are experienced and you are also the new person. How to spend the first months without faceplanting.
Day one in a new company is a reset button. The logistics, the watching, and the classic rookie faceplants.
Made, mostly, with one asterisk that matters more than the rule. What you can train into a tech lead and what you cannot.
"Trust your gut" is the best advice you can give a senior and the worst you can give a junior. A gut is compressed experience, and the only skill that matters is knowing which of yours is still calibrated.
My youngest tech lead asked why nobody warned him the role sucks. How to prepare a new lead so the job does not break them.
A clean no that leaves someone stranded is just abandonment with manners. Saying no to the task while staying on the hook for the person.
Your calendar stops being yours the day you become a lead. One dumb 2x2 grid that helped me get it back.
A Pulse Survey nearly gave me a heart attack, then taught me the real lesson: headquarters and I were grading the same team with different rulers.
Hand an agent a big feature and you get a big diff you cannot review and should not trust. The fix is not a better prompt. It is a smaller unit. Spec-first, one independently-deployable slice at a time, each in its own worktree.
Engineers swear we are not creative, then refactor the same job five ways before lunch. The two modes of technical creativity.
Not years, not salary, not a tour of eighteen technologies at Hello World depth. What "senior" actually means.
SMART is the most-quoted, least-applied acronym in our industry. The two letters everyone quietly cheats on.
Most objectives are written once and read once, doing nothing in between. The manager's job is to make them a stairway, not a wish with a deadline.
Every team is a small country with unwritten laws. The Team Canvas drags them onto a wall, and only works if people tell the truth.
The hardest part of a good 1:1 is keeping your own mouth shut. The craft of listening, preparing, and ending the meeting on something real.
The first meeting that gets cancelled when the week catches fire is the one that matters most. Why the 1:1 is the cheapest insurance you will ever buy.
