May you live in interesting times
Maksim Lin
12 Jan 2026
Over the last year, the software industry has been upended by LLMs. In the space of less 12 months, the capabilities of models have become really impressive. From being barely good enough "super autocomplete", to now, where at the beginning of 2026, the best models paired with the best agents are now able to create whole (if still simple) GUI applications from scratch and perhaps more importantly can debug even complex and subtle bugs like race conditions and obscure memory corruption issues in low level c++ code running on microcontrollers.
This is a step change in the order of going from writing machine language routines on punch cards to a 2000's era Java GUI IDE with none of the teletypes, assembly, C, glass teletypes or multitasking OS's in between and in the space of 2 years vs 4 decades!
I have no idea where this is all going to end up or even if there will be a paying job in industry or a role in society of "software engineer" the like of what I do today. At times recently I have had moments of pessimism about the whole outlook in general of where this will lead. And yet, LLMs are successful because they have "fitted in", providing powerful capabilities for us to work with the infrastructure we have, the infrastructe built over the last half century since the advent of mass computing. This global investment is beyond comprehension and is not going away or being replaced by LLMs anytime soon or potentially ever. All the existing computing infrastructure has been built up on what came before, most of what we have today is layers and layers built on even older layers.
So unexpectedly, I end here optimistically, because what LLMs have given me is the return of a spark I had completely long forgotten over the last few decades as a professional software engineer: that amazing feeling I got when I first learnt to program and realised that, within the confines of the machines capabilities, I could literally create anything I could imagine. I could write programs to solve any problem in the world that I wanted to work on.
Compared to those days, today our machines are very capable indeed, so, even if its for a brief twinkle in time, I hope my (and your) imagination is up to the task!
