Software Engineering at the Tipping Point (opens in new tab)

It is indeed at a tipping point; we are going from engineered deliverables to throwing stuff at the wall and keeping what sticks.

I wonder if the public appetite for the inevitable quality crash is there, though.

The page is blank.

  Uncaught TypeError: can't access property "parentElement", t is undefined
      i https://io.google/2026/explore/workshop-2:29
      <anonymous> https://io.google/2026/explore/workshop-2:29
      <anonymous> https://io.google/2026/explore/workshop-2:29

Evidently the title is correct: we're beyond the tipping point of software quality.

I'm sick of the people, these so called builders, how they are parasitizing on the time of the actual engineers to get their PRs into the codebase.

People who have never coded, for who we now need to spend an insane amount of time validating their design, asking questions they don't even have the answers to, figuring out if the code written was needed for the prompt or whether it was just thrown in there because ... the AI had it as part of the tokens.

Remember, their design isn't built from the ground up, it's focussed on the outcomes, like the movie Bedazzled.

And then you spend time validating their code, you give them review feedback, instead of internalizing it they just shove it into the AI and let it fix it, meaning they don't learn.

Next PR will have the same issues.

"Learn to use systems thinking to understand how developer ecosystems guide the evolution of your software systems. Improve your intuition for the systemic impacts of AI-driven software development and understand how you can better prepare for the exciting changes coming to our industry."

This reads like a bunch of marketing speak that says nothing.

I can't help but feel that this (not this specific article but the whole trend) is a little bit misguided. Nothing has fundamentally changed in engineering? We aren't committing raw markdown files in Github and having Jenkins re-inference entire systems at each push. The new AI tools still deal with the underlying mess and at some points the abstraction will leak.

It feels like Kubernetes? Yes, it allowed more complex setups but a decade later the leaky abstractions have become apparent and there are constant tradeoffs to be made with the problems it brings vs the problem it solves. But the awkward point is why was Kubernetes not a trillion dollar unicorn if it unlocked so much productivity gains?

Someone has to clean up the slop and explain why their product got more outages than before. Reputation and responsibility cannot be vibe-coded.

Case in point: GitHub.