Agentic coding is burning me out
8 hours ago
- #LLM Oversight
- #Agentic Coding
- #Decision Fatigue
- Agentic coding compresses traditional development pacing, removing routine tasks that provided natural breaks.
- LLM-generated code lacks manual development's gradual context-building, leading to 'cold starts' and constant oversight.
- Managing agents introduces variable rewards and cognitive fatigue, unlike the deep focus of traditional programming.
- Developers must trust LLM output, but can't fully rely on it, creating a productivity-trust paradox.
- Scaling to multiple agents amplifies decision fatigue, requiring constant judgment and architectural oversight.
- Decision fatigue limits developers to fewer productive hours compared to traditional coding sessions.
- Relying on more agents isn't sustainable; human cognitive load can't match automated output.
- Building reliable verification systems for LLM-generated code raises circular trust and validation challenges.