Not GP, but in my experience, those small changes are usually contrived to be small. Like saving snapshots instead of being atomic and self contained. And if they are the latter, they should have been separate PR.
That’s why I usually look at the diff of the whole PR and prefer to merge by squashing. Commits inside the PR are usually noise.
What you are saying is true if created manually. It is a chore to create sensible commits messages when you are committing often.
But LLMs appear to be pretty good at making small commits with sensible commit messages, that greatly help to make sense of the whole PR.
Regarding large monolithic commits, how do you even know where to start reviewing in a large PR? Smaller commits at least give you some sense of ordering.
> Regarding large monolithic commits, how do you even know where to start reviewing in a large PR?
What I’m saying is that large commits shouldn’t be a thing without out of band discussions (formatting, design refactoring, features merging). With such knowledge, it’s easy to glance at the code and understand the implementation.
You expect to send a big PR to someone without him being aware of the context or without being available to discuss your solution?
> I really doubt that this is true for large, legacy code bases, where one is only familiar with a small part of the system.
Reviewers should already be knowledgeable about the part being modified. If no, the review is just for checking for antipatterns in the code, or badly written tests. A bit useful, but can be unnecessary in high trust (or keen on accountability) workplaces.
It is not that I expect it. But instead, it is just about being realistic about how real world projects differ in workflows and conventions. If what you describe is not available, then the next best thing is to make the reviewer aware of the context and solution by means of small commits with sensible descriptions.
Maybe someone in Germany needs your help with the AFD Email leak; right now journalist are going through the data; your solution sounds like something that need
More relevantly, in a modern war you want experts. Even grunts need to know a lot to be effective, they have drilled and drilled until they react correctly without thinking about it - and somebody in charge of a handful of them needs to know how to use their skills to achieve larger strategic goals which then trickle up.
Needing to use a draft is a bad sign - the more of your soldiers are volunteers the better quality you'll get out of that.
I agree it's a luxury position, I'm just suggesting that "We should draft trust fund kids" isn't actually some clever fix to ensure alignment from the ultra-wealthy and your military forces will hate it.
There is the real thing you do. You ensure your kids are drafted. However, even in war, there are places that are peaceful that have to be protected. Your kids do have to go through basic training, which is grueling, but once they do that, you just make sure that they're the ones selected to man the supply chain back home. Or maybe they stand in front of an embassy of a third party not involved with the war because you still have to have guards in front of that embassy even though it's probably a ceremonial position.
Note that it often isn't the powerful doing anything. Anyone with sense in military chain of command is going to ensure the kids of the powerful people get moved to important but non dangerous positions without being asked. Even if asked to put the kids in a dangerous position it won't happen.
People are literally doing it right now since about a month ago, and the industries are losing their minds. Most notably some guy just decompiled all the Adobe products, made them free to use and without bullshit, and published them under the name ArtCraft. There've also been successful "port this Mario game to a Skyrim mod", "port this PS5 game to run on PC", and so on.
In the upcoming lawsuit, Adobe is either about to destroy the AI industry by collecting massive damages, or Anthropic is about to destroy the software industry by effectively making software non-copyrightable. There's no other outcome.
And making your software SaaS won't help because the AI tools can infer a server-side implementation based on enough examples of behaviour.
From what I've gathered, and I'm still feeling like I'm missing something, most people don't seem to be doing that. They seem to pass the binary in, say to decompile it, and out the other side is something that looks similar but not exact. https://youtu.be/qFEuYUfauNY .
On the recent WAN show LTT said they tried the recent Photoshop clones and they were buggy, feature missing messes. If they were truly reverse engineer decompiles (minus needing a subscription) they would work just as well. But they don't so they're not.
the adobe suite clones are vibe coded clean room implementations, decompiles on the other hand should compile 1:1 to the existing software, and we have ways to verify that's the case. recompiles from my understanding of the term are somewhere inbetween.
That's the view you hear from doomers - the view you hear from boosters is that it's basically Adobe quality. I haven't tried it so I don't know which is actually true.
I also heard a lot of people rave about Affinity and all kinds of Lightroom alternatives, but so far I found it all terribly disappointing and am back to CS5 and Lightroom 4 (because I own those on disc)… There is a lot of useless hype around.
No chance the AI industry gets destroyed. China wouldn't care if Adobe somehow sued and won (so really it would just be America destroying its economy), and this point nvidia could buy Adobe and shut it down mid lawsuit just to make it go away.
Meh I don't think Adobe has anything there. AI is just a tool. Are they going to sue Microsoft and Dell just because those two were involved in the laptop used to instruct the agent to do the thing?
Are they gonna sue the user's power company because they supplied power for a nefarious activity without making the human agree to use it for innocent purposes only?
With the right harness, it's doable today. Obviously quality will be what you end up actually spending time manually testing and what surfaces when testing this, because agents won't be able to even get 80% there without strict hand-holding. But with the right individual driving the right harness, definitely doable.
I've reverse engineered multiple proprietary things around the house thanks to a local Qwen-3.8 running under Strata and Hermes.
Some have been USB, some have been BT, and some were android binaries. I explain what's going on, how they should work, investigate, and write compatibility layers or hack the binaries to do my thing.
And right now, we're seeing flagship consoles getting emulators and compatibility layers to run the games on computers. And those are machine assisted as well.
Just go look at how bad Sony fucked up with 'no ownership of PS5 games' heh. Everybody's attacking and downloading and pirating PS5. LLM tooling makes it not just doable, but actually easy.
Before everything looked like bootstrap or one of the other 5 design systems.
Website? Hero Image, 3 Colums, footer. This was peak design half a decade ago.
reply