The author, a staff engineer, believes that AI agents will likely displace most software engineering jobs within ten years, despite historical precedents where such doom predictions were wrong.
- The author contrasts the secure feeling of being a software engineer in 2021 with the uncertainty of 2026, predicting the industry may not survive another decade.
- He notes the irony that software engineers are now facing automation, the same force they used to eliminate other jobs.
- Staff engineers like him will be among the last to be replaced, while junior engineers will suffer first.
- The outcome depends on whether tech companies undershoot (keeping humans longer) or overshoot (scrambling for senior talent) relative to AI agents' capabilities.
- He rejects the optimistic Jevons effect argument that total demand for engineers will rise, because AI is already improving at maintenance tasks.
The changing outlook
In 2021, being a good software engineer felt great; the world was full of software and jobs seemed secure. By 2026, the author is uncertain the industry will survive another decade, expecting far more change than in the previous twenty years. He foresees either carving out a niche supervising AI agents or leaving the industry entirely.
Unseemly grief and irony
Grieving the loss is unseemly for two reasons. First, software engineers historically used code to automate away other jobs, so it's cosmic justice that automation now targets their own industry. Second, as a staff engineer, the author expects to be among the last to go, since his work already involves supervising others. Junior engineers will suffer first, as companies can replace their hands-on coding with cheap AI agents.
Overshooting and undershooting
The author acknowledges that similar doom predictions (like high-level languages or outsourcing) never materialized, but argues that industries do die when made obsolete. The optimistic Jevons effect—that total software demand rises even with fewer engineers per line—is unlikely because AI can fix bugs and maintain code as well as write it. Personal experience shows AI tools have progressed from hopeless to often faster and more insightful at code maintenance. The author concludes that no new AI capabilities are needed to take his job—only incremental improvement in reliability.
Final thoughts
The author laments the loss of job security and the shift from internal struggles like burnout to external ones. He finds it silly for engineers to complain when automation finally affects them. He remains grateful for having recognized the good times and acknowledges being in a better position than very junior colleagues. He hopes he is wrong, but accepts that he and his colleagues may need to find something else to do.
Read this at any depth.
Install Depth and pick your level — Glance for a sentence, Summary for the gist, Read for the full take. Free daily quota, no signup needed.
Add to Chrome