tonkkatonka

What evidence do we have that code review is good for spotting bugs? Because the evidence actually says the opposite - we think that code review is good for spotting bugs but what it's actually good for is knowledge sharing.

Quoting from this article: https://thenewstack.io/kill-code-review-theater/

Ask engineers why they review code, and most will say they do it to find bugs. In 2013, Alberto Bacchelli and Christian Bird studied code review at Microsoft and classified 570 review comments. 44% of developers ranked finding defects as their top reason for reviewing. Only 14% of the actual comments were about defects. Five years later, Caitlin Sadowski and her colleagues looked at nine million reviewed changes at Google and reached the same conclusion.

"My working hypothesis is that human beings are too slow when dealing with code. If we're going to move at the speed of the agents, we're all going to have to back off a little bit and maintain high-level supervision."

The best engineers don't want to be botsitters

One approach could be the role of an automated verification engineer, a cross between a developer experience engineer and quality assurance.

Similar: https://www.aviator.co/blog/do-we-even-need-a-better-github/

Maybe.

Humans just can't thoroughly review the amounts of code AI can produce. But the answer is not YOLO and just ship blindly whatever AI created.

The answer is to move the human review/judgment upstream. You review specs. You review the steps. You review the verification rules. You review the contract the code must fulfill.

https://www.latent.space/p/reviews-dead

Some engineering organizations are measuring AI output via tokens burned, some compare engineers to executive chefs, and some rely on self-reporting.

The response to my piece about killing the code review was loud. This is my attempt to address the comments.

Human-written code died in 2025. Code reviews will die in 2026.

What's working, what’s still vaporware, who should be paying attention, and what the real path forward looks like for teams that want to be ready when orchestration goes mainstream.

Stanford researcher shares insights from large-scale studies on developer output, why early AI productivity claims were overstated, and what engineering leaders should (and shouldn’t) measure when rolling out AI across the software development lifecycle.

Four industry veterans — Kent Beck, Bryan Finster, Rahib Amin, and Punit Lad — shared their perspectives on how enterprises can adopt AI coding tools wisely.

Systemic engineering bottlenecks won’t disappear by just throwing AI at them. You’ve got to fix the roadblocks to realize the full value of AI tools.

Developer experience was about making developers happy. AI experience is about shortening the loop without loss of safety.

We still need experts to recognize that it did a good job, though. It might produce something extraordinary with security problems, and we still need to be experts to recognize that.

Internal developer portals are all the rage. And like most shiny tools, they’re being oversold.

History is the teacher of life. There are those who have been shipping software for decades, and there is much we can learn from them.

Instead of adding yet another hot take on whether vibe coding is real or if AI is about to replace software engineers, I wanted to take a shot at predicting what software engineering might look like in 2027.

Code migrations are like cleaning the house. We know we have to do it, but we wish we didn't have to. Yet, we are still only using AI to create new code vs cleaning the existing mess.

“So, where are you doing your doctorate?”

That was the third time someone asked me that and it wasn’t even 11 AM on the first day of IUI, a research-focused conference about Intelligent User Interfaces.

We all know that in many cases the phrase "we will improve this later" means "we will never improve this".

Great programmers are not born; they are made - says Daniel Terhorst-North, the author of the viral Twitter thread on the best programmer he knows.

Great programmers are not born; they are made - says Daniel Terhorst-North, the author of the viral Twitter thread on the best programmer he knows.

Striving for excellence is not just a matter of professional pride; it is necessary to prevent the endless cycle of suffering.

A few thousand lines of code everyone is satisfied with and understands used to sound like science fiction – now it was reality.

Despite zero TypeScript and MERN experience, we pulled off deploying Backstage in our platform, with a much better user experience. Only 5 % of our engineers use it.

Picking your (tech) poison wisely

What to do and what not to do when you consider adding new tools and technology to your tech stack.