[Next Week! Last chance to apply!] Barcelona - Itnig - 22nd April. CTO Networking + Tapas + Short talks from CTOs. Apply here.

Product engineers are not new. We are just noticing them more now.

People like Gergely Orosz and Marty Cagan have been talking about this for years. The difference is that AI is making it painfully obvious.

A strong product engineer builds better. They can own a project end to end, make smarter tradeoffs, catch edge cases early, understand business priorities, and avoid a lot of pointless rework.

There used to be an excuse: coding was hard, engineering time was expensive, so developers should just code and execute neatly packaged requirements. That was already weak logic before. With AI, it looks even weaker.

A few opinions on this trend:

It's much easier to learn enough product to be useful than enough engineering to be useful. Basic product judgment can already make you practical. Basic engineering knowledge usually cannot. AI can handle a lot of the shallow implementation work. The engineering knowledge that really matters now is the deep kind: understanding systems, reading code critically, spotting risks, and making sound technical decisions.

Rigorous code reading and review matter more.

Communication and good team dynamics matter more.

Architectural judgment matters more.

Learning to code well may actually get harder for juniors or product people, especially when they are pushed to ship fast with AI. They often do not understand what they are doing.

People who are genuinely strong at both product and engineering are rare, but probably more valuable than ever. I believe the ideal product and engineering team is made up mostly of product engineers with full ownership, supported by roles such as DevOps, DevEx, SREs, and platform engineers.

Reads on this trend:

Product engineers at Posthog

The Product-Minded Software Engineer by Gergely Orosz

Empowered Product Teams by Marty Cagan

GitHub is launching native Stacked PRs, and I'm genuinely excited about it. I've been a long-time advocate of stacked PRs. They're a great way to avoid huge pull requests with a LGTM comment or that take forever to review.

Anthropic has launched Routines for Claude Code, which let users automate software workflows on a schedule or via API calls. I think this could end up killing some startups, especially those focused on things like generating PRs when bugs are detected, investigating likely root causes of failures, or even doing AI-powered code reviews.

Like Anthropic, OpenAI Will Share Latest Technology Only With Trusted Companies. OpenAI has launched GPT-5.4-Cyber, a cyber-focused model, with initial access limited to selected security partners before a broader rollout.

Thoughts on slowing the fuck down. Teams are overusing coding agents and ending up with fragile, overly complex software. The takeaway is simple: use AI for scoped help, but keep humans in control of architecture, review, and key decisions.

Inside GitHub's Fake Star Economy. GitHub stars are increasingly being manipulated and that fake popularity can influence investor perception and funding.

A few days ago, I shared a survey asking engineering leaders in the community about their biggest pain points. The most voted topic was: "Programming with AI: what actually works."

That's why I wanted to share this: community member and friend Ezequiel Cura - Founder at Desplega Labs and former VP of Tech at Capchase - is offering free Agentic Programming Workshops for companies.

Worth checking out here.

Founding CTO [EUR 105-125K + 1-3% Equity] - Naiian