aider
aider copied to clipboard
Aider-AI
Where is Paul?
Issue
The most urgent issue currently for this project is clarifying Paul's status.
He has gone dark on twitter (I've pinged him) and in the discord.
Paul are you OK??? We're all here because we love aider and the work you have done and want to see aider continue to grow. If anyone know's Paul, can they reach out to him and advise what his intentions are.
In the meantime, the most active fork of this project seems to be https://github.com/dwash96/aider-ce/ where Dustin is doing some great work.
Version and model info
No response
Phew! I think we were starting to worry.
Realistically do you think you'll be back? Or ave you had any thoughts about a succession plan or models where you can share the load with the community?
Eg appointing/anointing a/some trusted codevelopers? Appointing a discord admin etc etc? Putting out an expression of interest?
I watched a fascinating video (cultrepo) on the history of python a few days ago (as an example of a successful community that grew from having a "benign dictator for life" to an ongoing community).
I suspect I'm like many users, loving aider but watching how this space is evolving and starting to feel a bit left behind.
I don't know if it changes anything, but I would actually pay monthly for aider. I don't know if it's possible currently. but I would.
Phew! I think we were starting to worry.
I do it now, after Paul's remark. A single-man project in an area with lightspeed evolution in general is demanding per se. I think, this will spiral. We have a working fork now (aider-ce), but I'm not too fond about this kind of development. I would like to see things kept together. Naming and all. At this point, this is still possible.
Phew! I think we were starting to worry.
I do it now, after Paul's remark. A single-man project in an area with lightspeed evolution in general is demanding per se. I think, this will spiral. We have a working fork now (aider-ce), but I'm not too fond about this kind of development. I would like to see things kept together. Naming and all. At this point, this is still possible.
~~Sadly, at least for me, aider-ce has a few too many bugs to be be usable.~~ [Update: CE is usable, with some minor display glitches.] The best outcome would be if Paul can find some time to express his intentions. If the not the second best option is that those that love aider can see if there is enough energy in the community to advance in an orderly and sustainable manner.
Phew! I think we were starting to worry.
I do it now, after Paul's remark. A single-man project in an area with lightspeed evolution in general is demanding per se. I think, this will spiral. We have a working fork now (aider-ce), but I'm not too fond about this kind of development. I would like to see things kept together. Naming and all. At this point, this is still possible.
Sadly, at least for me,
aider-cehas a few too many bugs to be be usable. The best outcome would be if Paul can find some time to express his intentions. If the not the second best option is that those that love aider can see if there is enough energy in the community to advance in an orderly and sustainable manner.
I think at this rate we are heading towards a fork - if Paul is too busy then that's just how it is. Personally I have found a version of aider-ce that works for me (not the latest which is broken for now) and am sticking with that while keeping up-to-date with changes. It's a complex project but there is real progress happening over there. I just wish I was good enough to contribute more.
Perhaps @paul-gauthier could take on co-maintainers, such as the maintainer of aider-ce? It is clear that this project is stagnating, and I've really not found a viable alternative. I use tools like Gemini CLI which are agentic but I keep coming back to Aider for many of my development tasks due to the control it offers, and it can be a lot faster with less token usage. I have literally spend millions of tokens to end of with a few thousand of usable lines of code with tools like Gemini CLI. I love the control that Aider offers but I am very concerned with the lack of progress lately with fixing issues and integrating models.
I would hate to see Aider break apart into a bunch of lightly maintained forks. Instead I think Aider needs to grow a community of maintainers like many of the popular open source projects. I think since it has been mentioned on the documentation that much of Aider was written with AI that it would be good to formalize the structure and processes with a community that could guide AI development.
I still think even with all the editor and CLI based agent projects Aider is really important and this type of project is underpresented now in the AI development options. I'd also love to see growth in the scripting and programmatic use of Aider use case, not just for coding, but for any type of project.
What do others think?
What do others think?
I'm just a small time user, but I care about the longer-term evolution and success of aider.
Context:
- With (so called) AI-assisted coding tools evolving very fast, it is probably essential -- at this time -- for aider to keep up with the latest and greatest performance while staying extremely close to the original usage/intents/visions.
- To be realistic: aider repo has the stars, got the early love and links, and is the place people will end up. No matter what, it will be hard for aider-ce to pick up the stars, which are important.
Therefore, proposals include:
- Less clean?: Make aider-ce maintainer of aider; merge all of aider-ce into aider. (Aider inherits aider-ce).
- Cleaner?: Since """When you transfer a repository, its issues, pull requests, wiki, stars, and watchers are also transferred.""" [0], just transfer aider repo proper to the Github account of aider-ce maintainer. Let that individual take it from there. (Original aider maintainer always welcome back any time. Future forking can be done, but meet that bridge when if/when needed.)
In summary, main concerns are:
- Keep aider going -- this involves some kind of cleanup of original-repo ownership and/or maintainership. Something beyond huge text at the top of the README saying "go look aider-ce".
- This must involve stars continuity somehow.
- This must result in lack of confusion of "what the current flagship aider is that is being developed, inherited directly from the original one", and also "there is the huge starcount", and also "there is a clear ongoing development and way to discuss and be involved".
Good luck and godspeed.
[0] https://docs.github.com/en/repositories/creating-and-managing-repositories/transferring-a-repository?utm_source=chatgpt.com
I personally don't think moving the repository or kicking out Paul from maintainership is a necessity of keeping this project alive unless Paul just doesn't want to be active anymore and refuses to allow others to co-maintain it. After all, Paul was the visionary behind the project and has a lot of knowledge and skill. I think it would be a lot better if other forks and maintainers coalesced around this GitHub organization and project unless Paul has abandoned it (which it does not appear that is completely the case). What this project needs is a vibrant collection of co-maintainers that are working together with formal process, not a transfer to another single maintainer in another orgaanization in my opinion. I think it also needs to formalize the AI development process so that Aider can continue to be largely built with AI. Pretty must all successful open source projects have many co-maintainers to ensure the project survives, and I think Aider could be a really great project that could fill a lot of existing gaps in this fast moving world if it had that.
@paul-gauthier Would you mind elaborating on your plans for this project here and letting us know if you would be open to bringing on co-maintainers in this GitHub organization?
Are there any alternatives to aider and the guy that forked it with aider-ce?
That's a shame. I just tried opencode and while it looks prettier and appears to run faster/sleeker than aider, its just buggier when you actually use it to, well, actually code with agents. Nice themes though...
That's a shame. I just tried opencode and while it looks prettier and appears to run faster/sleeker than aider, its just buggier when you actually use it to, well, actually code with agents. Nice themes though...
i also used claude code and gemini cli, but i like the granularity of aider. the agentic stuff is a bit too much code spewing, scanning etc.
Yeah, OpenCode is downright unusable for me using Konsole on KDE (Ubuntu) and it's buggy as hell. I also use agentic tools, particularly when I need to take a specification and plan and have it generate a initial draft, but I find myself coming back to Aider for a lot of my work, specifically because it is not agent based. It's more like a tool than an attempt at a developer, and as a developer for over 20 years I still like being in th drivers seat and I don't like having something develop so much code at once that it becomes hard to follow changes. Aider gives me that. I have also done work integrating Aider into programmatic workflows, both with shell scripting and at the Python level (creating simple Aider Python SDKs), which I would love to see a lot more of with this project.
The biggest strength of Aider in my mind is that it is not trying to be a do everything agent, but a CLI tool and scripting framework that actual developers can use that allows us to tightly control the context window and the flow. In a world where every CLI system out there is trying to build better agents this tool is badly needed, and in short supply.
I have yet to find any viable alternative and I have looked far and wide, particularly because it seems like integrating with new models and fixing issues has grind to a halt. My biggest issue by far myself is that writing markdown files is pretty much not possible if it has code blocks with the ``` fence. This drives me absolutely bonkers, but I still use it pretty much daily. My favorite way to use it at this point is to have it help draft guides and specifications for agentic CLI systems, and evaluate and find issues and draft issue lists to guide the agents to fix their work. Like I said I also like using it programmatically.
@paul-gauthier Once again since you are following this issue as of 16 hours ago, can you let us know what your plans are for this project and if you are open to building up a community of co-maintainers in the Aider-AI GitHub organization? Even one sentence that takes less than a minute to draft would be super helpful.
Hi @paul-gauthier, Even a co-maintainer can help with normal reviews and mergers, which is very beneficial for the long-term development of the AIDER project.
Just tried aider-ce. I really appreciate random people taking time to contribute and improve aider but, i immediately noticed more bugs/issues than aider, still less than opencode lol. I'm currently back on OG aider and happily using it. Not too surprised to see the aidermacs issue. All I'll say is, I don't envy people that have worked on medium+ software projects. I don't have too much experience working on massive codebases that support a plethora of platforms and I'm sure the regression testing needed for it be successful is on another level
That's a shame. I just tried opencode and while it looks prettier and appears to run faster/sleeker than aider, its just buggier when you actually use it to, well, actually code with agents. Nice themes though...
i also used claude code and gemini cli, but i like the granularity of aider. the agentic stuff is a bit too much code spewing, scanning etc.
Yes!! I love how simple and straightforward aider is. I really feel in control when I use aider, whereas when I use stuff like opencode I feel like I'm forced to be a product manager talking to a really stubborn junior dev that wants to do everything his way and I cannot control or influence much how he goes about solving the problem haha
Just tried aider-ce. I really appreciate random people taking time to contribute and improve aider but, i immediately noticed more bugs/issues than aider, still less than opencode lol. I'm currently back on OG aider and happily using it. Not too surprised to see the aidermacs issue. All I'll say is, I don't envy people that have worked on medium+ software projects. I don't have too much experience working on massive codebases that support a plethora of platforms and I'm sure the regression testing needed for it be successful is on another level
That's a shame. I just tried opencode and while it looks prettier and appears to run faster/sleeker than aider, its just buggier when you actually use it to, well, actually code with agents. Nice themes though...
i also used claude code and gemini cli, but i like the granularity of aider. the agentic stuff is a bit too much code spewing, scanning etc.
Yes!! I love how simple and straightforward aider is. I really feel in control when I use aider, whereas when I use stuff like opencode I feel like I'm forced to be a product manager talking to a really stubborn junior dev that wants to do everything his way and I cannot control or influence much how he goes about solving the problem haha
Just report those bugs on aider-ce, personally I use aider-ce since it works out of the box with archlinux.. and doesn´t need python 3.12, what issues are you having?
@paul-gauthier completed? In the very least you could respond to the requests in this issue. I'd like to see some concerns addressed. But if you are just going to close issues like this as completed it seems that forking the project is the only way forward.
