AI Coding Weekly

Zechner says deep knowledge still matters in 2026

A one line quote about agents writing the code turned into the week's loudest argument about craft.

Peter Steinberger posted a single quoted line and let it sit there. "Don't worry, you don't need to learn how to write this. Your agents will. And we made them really good at it." That was the whole post, and it was enough to kick off the week's loudest argument about whether craft still earns its keep.

Mario Zechner took the other side, hard. He pointed at the people who got Linux running on Apple Silicon, reverse engineering the GPU instruction set among a gazillion other things, all pre-AI, and set them against "the guy who thinks having your agent rewrite shitty rust he never looked at in x86_64 is excellence". His conclusion was that true knowledge and skill are still relevant in 2026 despite what, in his words, the clowns want to make you think.

Mario Zechner
@badlogicgames
X
the clowns will never ever know the beauty and power of true knowledge and skill, which is still relevant in 2026
Nov 9, 2026 · View on X

The skill he says you cannot drop

Zechner's constructive version of the argument is narrower and more useful than the insult. In a separate post he named the one thing he thinks you have to keep. Being able to look at what your agent produced and say no, not like this. Stop doing that, he says, and you are finished, whatever the devfluencers tell you.

Mario Zechner
@badlogicgames
X
the most import skill in 2026 is still being able to tell your agents "wtf is this crap, no, not like this
Nov 9, 2026 · View on X

Notice what is not in dispute. Nobody in this exchange is arguing against running agents. Zechner assumes you have agents and that they produce output you have to judge. The disagreement is about what you must still be able to judge, and whether that judgment survives if you never learned to write the thing yourself.

A third way to cut it

Running alongside the fight, Matt Pocock pulled a line from @poteto's talk at AIE Code that he called the most original thing in it. "nuance tends to lead to slop." The prescription is to design APIs and automated checks that take nuance out of the agent's hands. Ban useEffect, ban 99% of explanatory comments because they just lead to workarounds, and hide complexity behind interfaces the agent cannot get wrong.

Matt Pocock
@mattpocockuk
X
To increase quality, design API's and automated checks that remove the nuance from agents' decisions.
Nov 16, 2026 · View on X

For working engineers, that is the same split in different clothes. Zechner puts the quality gate in the reviewer's head. Poteto's framing puts it in the codebase, where it runs without you. Neither works if you cannot tell good output from bad in the first place.

Have the good opinion, not the loud one

One email a day. Under three minutes. What working engineers found out about coding with AI.

Get the next one by email

Coding with AI, read daily so you do not have to. The experiments, the receipts and the arguments from engineers shipping real software. Not a changelog.