The Next Two Years of Software Engineering

Addy Osmani frames five open questions on where AI takes software engineering through 2026, each with two contrasting scenarios rather than a single prediction. Junior hiring could collapse under automation or rebound as software spreads into new industries. Core skills could atrophy as AI writes most of the code, or become more valuable as engineers shift into an oversight role. The job itself could shrink into auditing AI output or expand into orchestrating AI agents. Narrow specialists risk obsolescence while T-shaped generalists get amplified. University degrees could keep their default status or lose ground to bootcamps and portfolios.

He pairs each question with separate advice tracks for junior and senior developers, and closes on the idea that none of the scenarios are mutually exclusive: some companies will cut junior hiring while others expand it into new domains, and the same developer might spend a morning reviewing AI output and an afternoon on architecture.

Fichamento

#

• Junior hiring: a Harvard study of 62 million workers found that genAI adoption drops junior developer employment by about 9-10% within six quarters, while senior employment barely moves. Big tech hired 50% fewer fresh graduates over the past three years. The Bureau of Labor Statistics still projects about 15% growth in software jobs from 2024 to 2034.
• Skills: 84% of developers now use AI assistance regularly. The entry-level skill is shifting from implementing algorithms to prompting and verifying AI output. The counter-scenario has humans handling the hardest 20% (architecture, integrations, edge cases) while AI covers the routine 80%, making deep expertise more valuable, not less.
• Role: one path narrows developers into auditors who review and approve AI-generated code (“maker becomes checker”). The other expands them into orchestrators, an architect or “composer” role deciding which tasks go to which AI agent or service.
• Specialization: betting a career on a single stack carries more risk as tools rise and fall fast, with COBOL and Flash developers cited as precedent. T-shaped engineers, deep in one or two areas and broad elsewhere, are favored, partly because AI tools make it easier for generalists to cover components outside their core skill. About 45% of engineering roles now expect proficiency in multiple domains.
• Education: CS degrees may stay the default credential while lagging behind industry needs, or lose ground to bootcamps, certifications, and employer-run training. In 2024, close to 45% of companies planned to drop bachelor’s degree requirements for at least some technical roles.
• Across all five questions, the advice to juniors centers on using AI as a learning tool without skipping fundamentals and building a portfolio; the advice to seniors centers on mentorship, architecture, and judging when to trust or override AI output.