Act IBefore the code
What to build and where to land it
A ticket without a problem behind it is future debt, no matter how well it's implemented.
- 01 · Product criterion
- 02 · Technical diagnosis
Guide for Product EngineersProduct,
Product,
not code.
The craft of the Product Engineer when AI writes the code by default.
Code is a means, not the end.
Free. Email for the PDF + weekly newsletter.
37 pages
3 acts · 6 dimensions
4 printables
Opening
Generating code has become cheap. Making product has not.
You probably already feel it. AI writes the code for you. Some afternoons you realize you've spent three hours accepting suggestions, another three iterating on a plan, and when you close your laptop you couldn't reconstruct what you did. Apparent speed went up, but something is creaking underneath.
If your value depended on how much code you wrote, AI has the upper hand on you. The difference is what I call craft: six concrete practices, almost always invisible in your Jira, that separate the engineer who ships product from the one who ships tickets.
A note before going on. This is not a document about AI. It's a document about what remains yours when AI sits inside the editor.
What's inside
Your real work cycle, in three acts.
Each act develops two dimensions of the craft. Every dimension follows the same pattern: symptom of atrophy, mental shift, three practices for Monday, and the table of what it costs not to do them.
Act IIIn the code
Resist complexity, direct the AI
The craft inside the code is not producing, it's resisting. AI pushes in the opposite direction, not by mistake, but by training.
- 03 · Sustainable code
- 04 · Collaboration with AI
Act IIIAfter the code
From the merge to the metric, and to the narrative
The merge is not the delivery. The delivery is the metric moving, and someone who knows how to tell the story.
- 05 · Outcome ownership
- 06 · Repositioning
The six dimensions of the craft
What doesn't show up in the repository.
When code becomes cheap, what separates the engineer the tool replaces from the engineer the tool amplifies are these six practices. The rest is done by the tool.
01Product criterion
Your unit of work is not the ticket, it's the problem the ticket represents. The three whys, the five-line briefing, and the conversation about alternatives before implementing.
02Technical diagnosis
Before touching a module, understand it. Impact mapping in fifteen minutes, the rule of the file next door, and the curious git blame, so your PR doesn't break what nobody anticipated.
03Sustainable code
Code holds up by simplicity, not by cleverness. A complexity budget, a second reading of the PR, and the smallest-file rule against the AI reflex of adding layers.
04Collaboration with AI
You direct, even when you don't type. AI writes, tests, and operates more and more, but it doesn't take on the consequence of choosing badly: that's on you. Ask it for alternatives instead of solutions, use it to interrogate the system rather than just modify it, and always sign off yourself.
05Outcome ownership
Your work doesn't end at the merge. Close the loop all the way to the metric, show up when your thing is on fire, and leave explicit handoffs for whoever comes next, including your future self three sprints from now.
06Repositioning
If your manager doesn't see the craft, the craft doesn't exist for your career. A decision journal, a 1:1 structured around outcomes, and three honest sentences for when the opportunity shows up.
The thesis
“Code was never the product,
it was the means.”
What you're paid for as a Product Engineer (deciding what to build, writing what holds up, and shipping beyond the merge) is still human work. AI accelerates the means, not the end. This document is about not confusing them.
The annex · 4 printables
Sheets to use, not to file.
Four templates with explicit cadence (Monday, during the week, Friday, every 1-2 weeks) so the craft enters your routine without rewriting your calendar.
Canvas typist ↔ owner
Every 1-2 weeksA scale per dimension to see, on a single sheet, where you're operating as a typist and where as an owner. Concrete evidence plus a reflective close.
1:1 template with your manager
Monday (or before the 1:1)Four sections to stop reporting and start narrating: what I decided, what I learned, what I'm watching, where I need cover.
Weekly craft checklist
FridaySix questions (one per dimension) to look at the week before closing it. Tick boxes, no scoring, no gamification.
Decision journal
During the weekTwo entries per sheet with Decision / Context / Alternatives / Outcome. Twenty entries in a quarter tell a story no list of PRs ever will.
About the author
Who writes this.
Emilio Carrión
Staff Engineer · Mercadona Tech · PhD candidate at UPV · Software production methods
I'm a Staff Engineer at Mercadona Tech: I build systems at real scale in physical retail. I combine this with a PhD at UPV on software production methods. Outside of work I write for thousands of senior engineers, give talks, and selectively collaborate with technical leaders and Product Engineers who are redefining their craft now that AI writes the code.
Take it and start on Monday.
It's free. I'll send it to your email and add you to my newsletter, where I write every week about the same thing.