The next bottleneck will be deciding what to build, and the
difficulty of trading-off conflicting needs. A software paradox of
choice.
Professional software developers will still produce the vast majority
of software in the world. Developers will continue to have jobs.
Developer productivity will only get a 10-15% boost on average by
most measures.
There will be outlier wizards: machine whisperers who produce
terrifying amount of software.
Companies start forcing people to adopt coding agents top-down, and it
doesn't go well. AI usage will start becoming a 'goal' and will get
Goodhart-ed.
There will be a movement to reject coding agents, reject LLMs and go
back to the good old days of writing code by hand.
It will fail to get traction in the workspace, where most software
is written for instrumental goals.
Knowledge work
Most white collar work will continue as-is. AI diffusion will remain
low.
Ham-fisted attempts to force AI use will lead to a growing backlash.
There will be multiple Claude Code equivalents for knowledge work, but
no clear winner.
By the end of the year, we'll reach a state where the case for using
agents for knowledge work simply can't be ignored.
Economy & Society
Anti-AI sentiment will be much higher in 2026.
The EU will seriously try to legislate AI use, to predictable
consequences.
There will be at least one major stock market crash and it will seem
like the bubble is bursting. This won't last though, and the bubble
will keep on bubbling.
Generative videos will be really popular even though no one really
thinks it's a great idea.