4 min read

[Weekly Retro] Too many layers. Too little judgment.

#279 - May.2026

Weekly Retro is a short e-mail with a wrap-up of ideas from the week, interesting links I found, and food for thought before you head off for the weekend.

Hi there!

💡 Here's a quick idea for the weekend:

More PMs will not fix the bottleneck.

Many people are concluding that, as engineers can deliver more in parallel, the solution is to add more PMs. 

This observation makes sense with I'm seeing today: More delivery capacity = more options to choose from = more decisions that a PM needs to take. 

But I don't think the solution is just increasing the PM:Engineers ratio.

It goes beyond: rethinking how PMs make decisions (high judgement over noise) and how they cut the coordination layers between customer needs and what gets built.

Every coordination step that isn't needed adds complexity, slows the team, and loses clarity.

This is the one thing PM can develop now: the taste and AI-steering skill to choose over an increasing amount of options, keeping the signal-to-outcome product cycle short.

Instead of delegating the hard judgment calls to AI, delegate the coordination overhead.

Let’s go build something great!

¡Saludos!

🎙️ Worth watching

This clip that started this conversation. Ng argues we should rethink the PM:Engineer ratio. The prescription is debatable. Good food for thought.


📖 Good tips for creative work

Claude for Creative Work
Anthropic is an AI safety and research company that’s working to build reliable, interpretable, and steerable AI systems.

Interesting framing:

“Claude can't replace taste or imagination, but it can open up new ways of working.”

9 new connectors linking Claude to Adobe Creative Cloud, Blender, Ableton, and others. Tools handle the repetition. Humans owns the judgment.


🤔 Article I've been thinking about

Who Owns the Code Claude Wrote?
AI-generated code copyright explained for builders.

Copyright only protects human-authored work. How about AI-generated code?


📖 Article I've been thinking about

Report: Meta will train AI agents by tracking employees’ mouse, keyboard use
Move highlights the difficulty of finding high-quality interactive training data.

This is a bit odd (and scary). Meta is reportedly tracking employee computer activity to generate training data for AI agents.

🖋️ Quote of the week

“Good design is as little design as possible.” — Dieter Rams
César Rodríguez
César Rodríguez
LinkedIn / Instagram / X