Uber's AI adoption numbers were wrong. A bug inflated the figures. The corrected data is more striking: 84% of Uber developers are agentic coding users, 65-72% of code is AI-generated inside IDE tools, and Claude Code usage nearly doubled in three months, jumping from 32% in December to 63% by February. IDE tools like Cursor have plateaued. That gap is the story.
The main piece tackles three pressure points hitting engineering orgs right now. First, what the Staff+ engineer role looks like when agents write most of the code. The argument is counterintuitive: demand for Staff engineers may increase, not collapse. Second, CTOs are sounding alarms about AI token costs rising faster than budgets anticipated. Two engineering leaders are on record. Third, Atlassian cut 10% of staff and named AI investment as the destination for the savings. The piece interrogates whether that framing holds up.
The Staff engineer section draws from a 50-person workshop in Utah organized by Martin Fowler, where practitioners self-organized sessions on the future of software development. That sourcing alone makes it worth reading in full. The licensing debate around an AI-reimplemented library and Anthropic's new $15-25 per-review code review product round out the industry items. The token cost section is the one to watch: it is the earliest signal of a spending correction cycle forming inside engineering budgets.
[READ ORIGINAL →]