Show HN: AI that edits your files directly, no approvals
5 comments
Instead of implementing your own command loop, why not just have a command that prompts the LLM, and uses the result to edit and commit a file (using hard-coded logic for the commit rather than parsing a tool request)? Then you're always in "stay at the terminal" mode anyway and you can use Git normally to "restore". If you want to loop anyway to re-use a process (and avoid persisting context between queries, or whatever), experienced terminal users can just background the process anyway. For me, everything aside from the actual prompt interface here is window dressing.
Anyway, safety for these tools is only minimally about having backups. The much more important part is the amount of agency extended to the tool. The LLM part itself can only generate instructions; it's up to the rest of the agent to actually follow them. So really all depends on what commands you recognize (is there anything beyond a "write text X to file Y" command?) and what is hard-coded (e.g. invoking `git commit` after each write).
Anyway, safety for these tools is only minimally about having backups. The much more important part is the amount of agency extended to the tool. The LLM part itself can only generate instructions; it's up to the rest of the agent to actually follow them. So really all depends on what commands you recognize (is there anything beyond a "write text X to file Y" command?) and what is hard-coded (e.g. invoking `git commit` after each write).
That's a fair question, and you're right that you can do this with a single command plus git.
The difference I'm exploring is defaults, not capability. In most tools the safe mode is "suggest, then approve", or "set a dangerous flag and remember to recover later". Here the default is the opposite: apply immediately, snapshot automatically, and make undo trivial. There are no flags to remember and no mode switches.
That sounds subtle, but in practice it changes iteration speed a lot. You stop deciding whether it's safe to let the tool act, and instead decide only whether the result is useful.
On agency: the LLM itself only produces text diffs. File writes and shell execution are fully deterministic and constrained on the client side. There is no autonomous command execution beyond writing files.
The difference I'm exploring is defaults, not capability. In most tools the safe mode is "suggest, then approve", or "set a dangerous flag and remember to recover later". Here the default is the opposite: apply immediately, snapshot automatically, and make undo trivial. There are no flags to remember and no mode switches.
That sounds subtle, but in practice it changes iteration speed a lot. You stop deciding whether it's safe to let the tool act, and instead decide only whether the result is useful.
On agency: the LLM itself only produces text diffs. File writes and shell execution are fully deterministic and constrained on the client side. There is no autonomous command execution beyond writing files.
you know there is `--dangerously-skip-permissions` for claude right?
Good comparison.
The idea of writing automatically by setting some flags and then recovering with git is not new: I agree with that; and it's not just Claude - others are doing it too.
What's different is putting it as the default user flow: automatic updates (no flags) - in combination with instant restore, not as a configuration flag to set and then a set of commands to recover.
That shift sounds subtle, but it changes how fast you iterate, especially when refactoring or exploring ideas.
The idea of writing automatically by setting some flags and then recovering with git is not new: I agree with that; and it's not just Claude - others are doing it too.
What's different is putting it as the default user flow: automatic updates (no flags) - in combination with instant restore, not as a configuration flag to set and then a set of commands to recover.
That shift sounds subtle, but it changes how fast you iterate, especially when refactoring or exploring ideas.
I built this because I got tired of the "suggest -> review -> approve" loop in existing AI coding tools. As models improve and generate proper code more often than not, manual approval started to feel unnecessary as long as there is a strong safety net to allow easy rollback of the changes.
Aye Chat applies changes automatically, but every AI edit is snapshotted locally, so you can instantly undo any change with a single command. This automatic file update with a safety net is the core idea.
In the same session, you can run shell commands, open Vim, and ask the AI to modify your code.
It supports multiple models via OpenRouter, direct OpenAI API usage with your key, and also includes an offline local model (Qwen2.5 Coder 7B).
You can watch a ~1-minute demo here: https://youtu.be/i-vGI6-kP4c
Basically, the typical workflow goes like this (instead of a chat window, you stay in your terminal):
I use Aye Chat both in my work projects and to build Aye Chat itself. Recently, I used it to implement a local vector search engine in just a few days.
Lower-level technical details that went into the tool:
The snapshot engine is a Python-based implementation that serves as a lightweight version control layer.
For retrieval, we intentionally avoided PyTorch to keep installs lightweight. Instead, we use ChromaDB with ONNXMiniLM-L6_V2 running on onnxruntime.
File indexing runs in the background using a fast coarse pass followed by AST-based refinement.
What I learned:
The key realization was that the bottleneck in AI coding is often the interface, not the model.
I also learned that early users do not accept a custom snapshot engine, so to make it professional-grade we are now integrating it with git refs.
What I'd love feedback on:
- Does the snapshot safety net give you enough confidence to let the AI write files directly, or does it still feel too risky?
- Shell integration: does the ability to execute native commands and prompt the AI from a unified terminal interface solve the context-switching problem for you?
There is a 1-line quick install:
Homebrew and Windows installer are also available.
It's early days, but Aye Chat is working well and is legitimately the tool I reach for first when I want to iterate faster. I would love to get your feedback. Feel free to hop into the Discord (https://discord.gg/ZexraQYH77) and let me know how it goes. If you find it interesting, a repo star would mean a lot!