The Punk PM #60

Hey there, punk!

This week I watched Matt LeMay's talk from the recent Productized conference on why communication is slowing product teams down, and what to do about it. One question stuck with me: why are we more willing to write documentation for an LLM than for the person sitting next to us?

It got me thinking about writing, and whether it still earns its place in an AI-powered world. As a former journalist and a lifelong writer, my answer is that getting words down on a page matters more than ever. And it's not something we should be so quick to hand over.

Let's dig into it.

Quote of the Week 🙊

Insight 🦉

Somewhere in my bookmarks there's a Medium post Noah Weiss wrote when he ran product at Slack—a fat list of resources for new PMs. The section that stuck was about writing. It linked to Werner Vogel's original Working Backwards approach, to Steven Sinofsky's insistence that writing is thinking, to the idea that bad managers talk, but good managers write: themes that shaped how I worked for a decade.

Now half of us have quietly stopped writing. Why waste time putting our fingers to the keyboard? The machine drafts the PRD, the machine summarises the deck, the machine reads the forty-page report so you don't have to. I've done it. You've done it. It feels like leverage.

It isn't. It's abdication dressed as efficiency.

Here's the confusion at the heart of it. We've decided writing is an output—a thing to be produced, and therefore a thing that can be automated. But writing was never the output. Steven Sinofsky was right. When an interviewer described Richard Feynman's notebooks as a record of his work, Feynman corrected him. The notes weren't a record of the thinking, they were the thinking. The work didn't exist somewhere in his head waiting to be captured. It didn’t happen until his pen hit the paper.

That's the part the productivity pitch skips. When you hand the writing to an LLM, you're not saving the boring bit and keeping the good bit. You're outsourcing the good bit. The struggle on the page—where you find out your position is much weaker than you thought, where the argument you were sure of collapses under its own weight. That's not friction to be removed. That's the whole point.

Product managers are meant to be opinionated. It's close to the only thing that doesn't commoditise, and an opinion worth holding doesn't arrive by prompt. It gets forged. It's the same thing I keep coming back to with the producer analogy I've been circling round for the last few weeks. The value isn't in working the desk, it's in knowing what the thing should be, and that conviction has to be earned.

There's a harder edge to this too, and Andy Grove got there decades ago. Data, he warned, only ever describes the past. It can inform a decision but it can't make one, because the future you're actually betting on is one the numbers have never seen. LLMs are built entirely from what has already been written. So if you let one do it for you, you get a fluent, confident account of a world that already exists: averaged and backward-facing. Precisely the wrong instrument for a call about what comes next.

None of this is nostalgia for the artefact. I've argued for years that the PRD is dead, that prototypes are the new spec, that shipped software is the new prototype. I still believe it. But killing the documentation was never permission to kill the thinking. The thinking still has to happen somewhere, in some form, with your own hands on it.

AI can be the R2-D2 in the back of your X-wing. Genuinely useful, occasionally the difference between making it home and not. But you still line up the shot. The surest way to work out what you actually think about something is still the old-fashioned way. Open a blank page. Start writing.

Action 🚀

This week, take the next document you'd normally hand to an LLM and write the first draft yourself. No prompt, just you and a blank page.

It'll be slower, but that's the point. Notice where you get stuck, because that's not the writing failing you. That's you finding out you haven’t got it figured out yet.

Let the machine tidy it up if you like. But the thinking's already done. And it's yours.

Inspiration 💡

The Age of Reading Is Over – Rose Horowitch makes the backdrop concrete. We read more fragments than ever and fewer long things, and the muscle for sustained thought wastes away with it. If reading is a workout for attention, writing is the same workout for judgement. Hand both to a machine and you don't just lose the output—you lose the ability to produce it. Read more

I'll Know It When I Build It – Christina Wodtke is the best argument against everything I've just written. She builds an app with Claude Code by feel— no spec, no plan—and only works out what she wants by watching it take shape. The building was the thinking. The app became, in her words, a mirror for her own cognition. If writing is thinking, then so is this, and it puts the machine right in the loop I've been trying to keep it out of. Maybe writing isn't the only way to think something through. Read more

Stop Thinking of AI as a Coworker. It's an Exoskeleton – Ben Gregory digs deeper into where I think we land with LLMs. AI doesn't replace the person, it amplifies whatever they already bring. Same point from the other end: the exoskeleton is useless without someone inside it deciding where to walk. Bring no thinking of your own and you've strapped yourself into an expensive machine that goes nowhere. Read more

Signing Off ✍️

If this resonates with you, hit reply and let me know. And if you think a friend or colleague would enjoy The Punk PM, feel free to share it with them!

Play it your way,

Toby