AI is repricing software engineering identity and skill

Three threads today circle the same uncomfortable question from different angles. The 'LLMs are eroding my software engineering career' post is the most direct signal: engineers are watching their leverage shrink and they are not sure what replaces it. The Harness post about shipping a million lines of code in weeks with Codex provoked a mix of awe and disgust, with commenters calling it 'half-baked CRUD' at industrial scale. The tokenomics paper asks whether the next generation of senior engineers will be people who optimize token consumption in agentic workflows rather than people who optimize algorithms.

The pattern here is not just about job losses. It is about what competence means. Google famously hired on systems-thinking and optimization. The tokenomics thread explicitly asks whether companies will soon hire on how well you can reduce the token cost of an AI workflow. That is a real shift in what expertise looks like, and it is happening faster than most engineers expected.

The counterpoint in the comments is real though: the Harness million-line codebase drew heavy skepticism. People noted that volume of AI-generated code is not the same as quality, and that someone still needs to understand what the code is doing. The debate is not settled, but the direction of travel is clear.