Ok, then how do you approve/take the first 2 commits and not the last one? Or how do you insert a commit in between 2 commits or address feedback on a given commit?
The person making the change is now going to have to run multiple confusing interactive rebases and git shenanigans, the rewrite the history of the PR branch on every feedback, then you have to re-review all the commits again because they are all different.
It’s possible of course to push all that complexity on the tooling. Have GitHub and git provide tooling for doing all that within the context of a single branch/PR. But why is that better? Multiple branches are easier to manage in git, and as long as they don’t conflict on the merge. Obviously if a feedback on PR#1 causes a conflict in PR#2 which causes a conflict in PR#3 it’s still tedious, but it’s a lot more doable than managing interactive rebases on every feedback comment.
Jujutsu makes the author-side of it all fairly nice, indeed. But the reviewer-side has been so horribly abandoned by github that it has been frustrating me for a long while now.
Of course it is. If I write a movie review and sell it to a magazine or whatever, it's derived from the movie, and it's fair use, and I don't need to ask the movie owner for permission first, or give them a cut of my sales. Even if I use some reasonable number of screenshots and video clips, as long as the resulting work is "transformative" i.e. actually a new work, a movie review instead of a copy of the movie.
Do you want this to work any other way? I constantly see people in the AI debate working themselves into wildly copyright maximalist positions. I actually don't think that we should give every author veto power over a book review!
>I constantly see people in the AI debate working themselves into wildly copyright maximalist positions
I really dont get this. I know its that conflation fallacy or whatever, but I was under the impression we had sort of gotten over copyright maximalism as a society after Napster etc.
Whats worse is that, meaningful reform in this space has basically been waiting on a multi billion dollar corporation to come along and push it forward. So now that we have an opportunity to expand and globalise fair use, the sudden and quite angry opposition weirds me out to no end.
1. author owns the right to distribute copies of the work
2. this right goes on for faaaaaar too long.
I don't have an issue with 1. You had a good idea, you implemented it, you deserve something for it. Given some people got sued into oblivion with ridiculous dollar value outcomes on a per unit basis - why doesn't this apply here? Sure 1.5 billion is a lot. But the number of infringments is insane and the company is approaching a trillion in valuation. You could make it ten times that number.
I do have an issue with 2. Sure, you had a good idea, you implemented it, you deserve something for it. But after 20 years, you should be able to come up with another idea or just work like the rest of us. Going for 50, 70, 90+ years with the rewards going to estate heirs? Fuck that.
So yeah, I am both against copyright AND surprised at the slap on the wrist for what happened here.
>Given some people got sued into oblivion with ridiculous dollar value outcomes on a per unit basis - why doesn't this apply here?
I mean, it feels to me like one or both of:
1. The class action lawyers werent 100% certain they could win in court.
2. The class action lawyers smelled an easy payday.
They get ~100 million out of this.
I also think that the 1500 bucks going to most of these authors is going to be more than they ever saw in royalties. I read somewhere that 500 - 1500 bucks is roughly what a self pub book makes in its lifetime. Why push the envelope? Anthropic hasnt done anything that deserves to pay for the entire lifetime royalties of most books. Their legal alternative is to cut the spine off and scan the book in. In which case the author and publisher will be splitting 20 bucks instead, assuming Anthropic isnt buying used.
This seems like a donation tbh.
>slap on the wrist for what happened here.
Its not a punishment at all because this is a civil case that has been settled out of court.
IANAL but as an IP creator I have not heard of "derivative products" in the copyright context. There are "derivative works", which are covered by the same copyright as the original. For example, a translation to another language is a derivative work, a novelisation of a movie, a screen adaptation of a book etc. If some author could have proven that any Anthromic model is a derivative work of theirs then they had the copyright on that model and made mad bucks licensing it back to Anthropic.
I think PIT debuggers have their use especially with regards to altering state mid run, but this is the first time I've heard of reverse debugging and it seems incredibly useful, especially for multithreaded software I'm often debugging
Each genre has a fairly tight envelope within which to operate. Regardless 90% of tracks never make it to the finish line because hobbyists haven't learnt them well enough to groove them out. If with a little help these tracks were all finished then bedroom producers will over time learn what works and be able to explore more.
I think the parent comment was saying that the problem is not quantity, but quality.
Warping my mind back into a hobby-enthusiast music producer mindset:
an MCP that generates presets for a limited pipeline with many sweet spots sounds... interesting?
To me, the idea of being able to have, say, a chain of a simple VA synth + delay + compressor and a very simple step sequencer, combined with prompting and a genAI model that spits out patches, sounds very endearing and interesting.
Much more interesting than Gemini or Suno for example.
Depends on the training and input space of course.
I deliberately described a limited setup, the controls of which could be described in less than a kilobyte.
Many dance music synth patterns could be described by simple means (tracker/step sequencer, looping, a few knobs).
That's what makes a lot of music interesting.
I can easily imagine a producer creating very individual and interesting output by unleashing the right models.
I think, just like with human producers, constraints liberate.
An AI controlling a very limited synthesis chain is more interesting than a very complex synthesis chain controlled by a human with no musical "vibe".
While I can get behind the sentiment I hope bad writing doesn't become the standard for anti AI. A simple grammar check would have greatly improved this post.
AI has plenty of training data on poor writing. If people start looking for bad grammar and typos to identify human articles, generative AI is certainly capable of spitting out prose that looks poorly edited.
I kind of hope the anti-AI-writing stuff passes and we can focus on what makes writing good or bad again instead of “this is clearly AI” posted in response to every blog. I actually don’t care if it’s AI but I do care if it’s worth reading and pleasant to read.
When most of the AI written blog posts are just written via a "write a blog post about X" prompt. I don't see any value in reading any of it. If I want to know what ChatGPT thinks of a given subject I'd just ask it directly.
Good code has always been written with a reader in mind. The compiler understanding it was assumed. The real audience was other engineers. We optimized for readability because it made change easier and delivered business value faster.
That audience is changing. Increasingly, the primary reader is an agent, not a human. Good code now means code that lets agents make changes quickly and safely to create value.
Humans and agents have very different constraints. Humans have limited working memory and rely on abstraction to compress complexity. Agents are comfortable with hundreds of thousands of tokens and can brute-force pattern recognition and generation where humans cannot.
We are still at the start of this shift. Our languages and tools were designed for humans. The next phase is optimizing them for agents, and it likely will not be humans doing that optimization. LLMs themselves will design tools, representations, and workflows that suit agent cognition rather than human intuition.
Just as high-level languages bent machine code toward human needs, LLMs let us specify intent at a much higher level. From there, agents can shape the underlying systems to better serve their own strengths.
For now, engineers are still needed to provide rigor and clearly specify intent. As feedback loops shorten, we will see more imperfect systems refined through use rather than upfront design. The iteration looks less like careful planning and more like saying “I expected you to do ABC, not XYZ,” then correcting from there.
Given how precious the main context is would it not make sense to have the skill index and skill runner occur in a subagent? e.g. "run this query against the dev db" the skills index subagent finds the db skill, runs the query then returns the result to the main context.
reply