One of the things that kicked off the era of Agentic Engineering was just robust file editing.
- One of the things that kicked off the era of Agentic Engineering was just robust file editing.
- As recently as a few months ago, LLMs editing files was error-prone and slow.
- Streaming out an edited file, token by token, and hoping the LLM doesn't get confused or try to 'improve' something unrelated.
- LLMs producing diffs would get subtly confused.
- All of the agent harnesses do file editing in slightly different ways.
- But they all give the LLM a mechanistic tool.
- They typically require the LLM to provide the exact text it wants to modify, and treat multiple matches as an error.
- Some also require the LLM to have read the file since it was last written before they allow a write.
- If you turn these off, Agentic Engineering gets orders of magnitude less smooth.
- One of those minor catalyzing innovations we take for granted!