The Rewrite That Actually Worked | Exec Engineering #201
Also: the best engineer your hiring keeps turning down, why a good AI mandate is just cover for the team while they learn, and the quiet estimate-padding that eats nearly half your R&D capacity.
Heyđ
Thanks for reading Exec Engineering, a weekly digest for the busy tech executive.
I hope this edition brings you value.
The Digest
The Rewrite That Actually Worked (Roman / High Impact Engineering)
System rewrites are the project every leader watches die at 80% done. Roman, a startup CTO, shipped his in three months with an AI coding agent. For once, the estimate held. It held because of the setup around that agent. The agent is only a multiplier, he says, but âyour environment is the exponent.â Instead of static Figma mockups, they built working prototypes the team could click through. That surfaced the real edge cases before production code. Point a raw agent at your mess and you get a faster mess.
The Engineer in the Half-Space (Yusuf Aytas)
Some engineers do their best work as the absence of a problem. Think of the outage that never fired, or the migration that did not break on old customer data. Yusuf Aytas calls this person the gap reader, the one who spots the missing owner and closes the gap before it gets costly. The catch is your hiring keeps passing on them. Your interviews score algorithms and system design, and this is neither. It looks like someone asking plain, practical questions. Worth reading to spot the people your process misses.
In defense of AI mandates (Charity Majors)
Most AI mandates deserve the hate they get. Charity Majors makes the case for the rare good one. A real mandate is the leader giving people cover. It admits some deadlines will slip while the team learns to really build with AI. It promises the team will not carry that slip alone. Skip it and you have told people to pick up AI on their own time. So pick a lane. If you are going to force it, reality had better prove you right in a hurry.
How to think about span of control (Jade Rubick)
There are two ways to shape an org. Leaders usually go flat to move faster, with fewer layers and quicker calls. But flat is not free. It clears the tangle of extra layers only to drop a heavier load onto your front-line managers, who each run far more people. It pays off only when you have set up the right conditions first, and Rubick is sharp on what those are. Flatten without them and you just move the pain downhill.
AI productivity is burning out your best engineers (Sreenivasa Reddy / LeadDev)
Your fastest team might be quietly losing its middle. Juniors and seniors get a real lift from AI. But someone still has to catch the confident nonsense the model hands back, and it is almost always your mid-level engineers. One of them told Sreenivasa Reddy why she quit. âI cannot handle this pressure of making things right without any credit for it.â Reddy lost three of the teamâs best inside two months, and never saw it coming.
Stop Lying to Your Cofounder (Aviv Ben-Yosef)
Padded estimates feel harmless, so Aviv Ben-Yosef ran the math. The engineer pads a task by 20% to feel safe, then the manager pads it again, then the VP piles on more. Stack those hidden cushions and you spend almost half your R&D capacity protecting the team from itself. And trust does not fade slowly. One caught white lie colors everything you ever said. So hand over your real estimates with the risks spelled out, and bring your own bad news early, before anyone has to dig for it.
If You Canât Be Replaced You Canât Be Promoted (Dave Anderson / Scarlet Ink)
The fastest way to get stuck is to be the only one who can do your job. Dave Anderson made hiring his own backup his first move, and got promoted right on time. Stay the one nobody can replace, and you stay too valuable to move.
The Four Core Areas of Responsibility for an Engineering Manager (James Samuel / ESLâs Newsletter)
You have seen the manager job split into people, product, delivery and technical work, so that is old news. The real point is simpler. Your job is to make the team greater than the sum of its parts. So where you spend your time shifts with whatever the team needs most.
Dialog
I asked Manuel Drews, Director of Engineering at Native Instruments, why AI has made so many teams faster at building without shipping value any faster.
His take is that doubling your build speed mostly drags the real bottleneck into view. If the deciding and the customer-adoption around it stay just as slow, you have only shifted the jam somewhere new.
We also got into why a struggling team is usually badly shaped rather than short on talent, and what he does to fix one. A good read if your dashboards look faster this year but the outcomes have not moved.
More reads
The Cost YAGNI Was Never About Kent Beck â Still Burning (Kent Beck)
Good Judgment Beats Good Prompts Art of TPM (Aadil Maan)
The Two Words I Never Got to Hear (Deb Liu)
FIREing Our Way to Financial Literacy in|retrospect (Allen Cheung)
About Exec Engineering
Iâm Yassine đ I spend a big chunk of my time digging into engineering leadership, talent, and craft, plus how AI impacts all three. I share the most valuable gems I come across in this newsletter every week, all curated by hand.