6 months ago I wrote on Tools, thinking and people where I explored how interacting with AI is a bit similar to interacting with Google Search 20 years ago.
Today I want to talk about a different thing that bothers me in the tech industry.
Recently Sunil wrote this longer tweet - exploring how some senior tech people were publicly skeptical of AI and how after AI improved dramatically they painted themselves into the corner.
There is one more related phenomenon. Even before AI I worked with a few very senior engineers who stayed at the company for many years. They were setting up all the technical agenda and making key decisions. But if you look into company software and tools and processes - all were 10 years old: slow UI, ugly interfaces, deprecated libraries - no preview environments, no easy way to deploy a prototype.
10 years ago I worked at a company called Runnable - we tried to do staging/preview environments for each pull request. Very similar to what every single sandbox company tries to do for agents (same isolation, per agent instead of per pull request). Even large enterprises tried to build similar in-house solutions using kubernetes.
Then of course we hit the age of AI and all the vibe coding tools changed the math here. Ability to quickly spin up a new app became so much more important.
So the flow became clear (with Lovable and all the companies in the space) - you start a new app, work on it and it should be live (publicly or internally).
We knew what the process should be for 10 years. But here come gatekeepers. I also worked with engineers who were not aware how much the world changed (not even saying in the last 4 years with AI). They just wouldn't understand how important it is to let people easily create internal live apps and demos. Friction is not acceptable anymore.
I worked at one company where a senior engineer mandated one specific Python library and one blessed server technology for all internal apps. Long deploy process, internal bureaucracy just to get a “Hello World” app live. Something I last experienced 16 years ago working on government projects!
It was annoying before AI. In 2025-2026 it became unacceptable - AI raised the bar.
By the way, that same engineer didn't even try AI at all for the whole of 2025!
People who didn't revisit their technical choices before AI have trouble revisiting them now. Not all of them, but some! If you are not willing to improve processes for your teammates or customers, if you want everyone to use the tools you learned 20 years ago - you are the gatekeeper.