darktable-org / darktable-org/darktable
RFC: What happens when we run out of/can't afford AI tokens
Nobody has claimed this yet.
- Dominant language
- C
- Stars
- 13.1k
- Forks
- 1.4k
- Avg merge
- 22h 14m
- Merged PRs (30d)
- 198
Description
Historical cost of development
darktable development and maintenance has been performed by humans donating their time and expertise. Adding new developers is a slow process because of the time it takes to understand the code base and gain expertise. In terms of cost code has been expensive in time and maintenance has been expensive in time. However, since time is donated freely the cost in currency is 0.
The recent 9 months
We have seen an influx of AI assisted developers who leverage AI to replace darktable knowledge and expertise. The result is code has now become cheap in terms of time. Maintenance has become cheaper in terms of time. The last release contained the largest number of commits ever. Remember that we used to only release once a year. The next release is on track to have more commits than the last.
The eye opener
@jenshannoschwalm asked copilot to do something in a PR and copilot returned with a message worded something like "you have no tokens left".
In a "discussion" in another PR I had made the statement that because of AI code was now cheap and maintenance was now cheaper than before.
Now I realize code has a cost in currency and, darktable has no currency. So, maybe open source isn't free after all.
The future
A month ago I reached the conclusion that darktable would be unmaintainable in 2 years. My reasoning:
- A lot of code written with AI by people that have little or no coding skills (not an assumption as several have stated that) so they don't know if the code is any good or not and they rely on AI to tell them it's OK. Kind of like giving the prison inmates the keys and asking them to keep themselves locked up.
- Overwhelmed old developers and maintainers. The sheer amount of code being introduced every day has increased the workload of developers at least 10 fold. For me, I have no time to do any development because all my time is spent reading PRs and issues, reviewing code, doing occasional testing, and the endless discussions trying to reason with AI devs. I don't know how @TurboGit does it because he is also reviewing code at a much deeper level than I am plus a whole lot more management.. We are going to burn out and walk away for our health and sanity. I've been thinking about leaving for the past couple of months because it's not really fun anymore.
- The volume of code written by AI will need to be maintained by AI since no one understands the code including the ones who "wrote" it. As the code base keeps growing we'll quickly exceed any "free" tier of tokens, if we haven't already. Once we can't afford the tokens necessary to maintain the code, darktable is not maintainable any more.
When I had the eye opener I realized that my projection a month ago was probably very optimistic. 😢
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
No files, tests, or entry points are named. Start by reading the RFC and its discussion, then identify whether the project can agree on a concrete response to AI-token costs and maintenance workload. Done cannot be defined until the discussion produces a specific, actionable proposal.
Written by the indexing model from the issue text.
Assessment
- Domain
- developer-experience
- Issue type
- Feature
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Active
- Clarity
- Needs clarification
- Newbie friendliness
- 15/100