Darko Tomic is a Unity developer of about 11 years, and the essay starts with the feed rather than the figures: LinkedIn is now mostly “Open To Work” rings, including friends with a decade of experience. His argument is that treating the contraction as a single AI event misses what it actually is — the tail of a bubble, priced in years, with AI compressing the headcount that survived it.

The chronology he lays out is the piece’s real contribution:

  • 2020 — lockdown demand. He had 15 interviews lined up in a market with roughly three Unity studios, companies approached him first, and studios staffed up on anything.
  • 2021 — NFT and metaverse money peaks. Recruiters daily, salaries bid up between studios, and switching jobs as the standard answer to any complaint.
  • 2022 — two signals land together: the metaverse funding peak, and a large social platform halving staff and still functioning, which taught other companies that deep cuts were survivable.
  • Late 2022 onward — capital rotates overnight into AI, and the layoffs that began in 2022 have not stopped.

The part worth stealing is his account of what AI-caused job loss looks like from inside a studio. It is not an announcement. At a VR company in 2024 he shipped a Python feature despite never having written Python beyond tutorials — work that previously required hiring a Python developer, absorbed by a generalist with good tooling. His summary: that is how a Python developer lost a job he never had.

He is not anti-AI and does not pretend to be neutral either. He uses an AI editor daily, only writes code by hand when learning a language, and says a task that took two programmers three weeks now takes him about an hour. He is also candid that the pace is punishing — he programs all day because the tools move weekly.

Where he draws the line is a human bottleneck: audiences register inauthenticity, so output gains do not convert one-to-one into market value. His evidence is personal and unquantified — his human-written blog posts outperform his AI-assisted ones — and he labels his prediction that new human-content roles will emerge as a guess rather than a finding.

The advice he lands on is mostly to wait out the cycle and, meanwhile, sharpen what AI cannot copy: real experience and programming intuition. The waiting half assumes the cycle turns and that time is not itself the scarce resource for someone mid-career. The second half does not depend on the cycle at all.