I would disagree. The size doesn't matter that much. The main metric is just number of commits, because that is what drives the flow. Not saying size doesn't have an effect, just saying number of commits is the primary signal.
Average number of files changed per commit is going to have a huge impact on performance since that’s a huge part of what the “git push” handshake is trying to figure out regarding what to upload.
> It is very unlikely that LLM had access to original source code of Photoshop.
It doesn’t matter. It’s likely had access to the knowledge of somebody who worked on that source code.
I have worked at companies that have done clean-room implementations. I was not permitted to look at their source or interact with those teams, because I had been exposed to the source of what they were re-implementing in a prior job.
As a human with a mushy brain I was never going to remember the source, but the fact I had been proximate to it was enough to lock me out. You have no idea what the LLM trained on, so you can’t prove a LLM derived clone is a “clean-room” implementation.
Some questionable idea that the LLM "had access to the knowledge of somebody who worked on Photoshop source code" seems like a poor basis to allege it's not a clean room implementation.
Wouldn't the burden of proof here be on Adobe if they wanted to make such an argument?
That's not the argument in the article. Not only that but coding is deterministic up to a certain point. Once the project becomes complex, it becomes harder and harder to account for all edge cases. Sometimes it requires heuristics that are not deterministic, etc. There are infinite ways to implement the same requirements.
That argument is like saying that weather is deterministic. It is if you know every single variable, but that's impossible.
The Director's Cut (released in 1992) drops the narration and adds the dream sequence, but despite the name is not entirely approved by Ridley Scott.
The Final Cut (released in 2007) is the only version where Ridley Scott had full final cut approval. It re-masters the audio, fixes some of the effects (replacing faces on stunt doubles, removing wires, etc), and changes a few lines (most notably "I want more life, father" instead of "fucker"). The color is also re-graded.
Some people prefer the Director's Cut over the Final Cut, especially the color grading, although I saw the Final Cut in 35mm when it was released and it looked pretty spectacular. In any event the changes in the Final Cut are nowhere near as egregious as the Star Wars special editions or the AI remasters of James Cameron's films.
> the Star Wars special editions or the AI remasters of James Cameron's films
I really wonder why some creatives are so determined to sabotage their earlier works later in life. Do they actually believe that they are making them better and are so used to being told how great they are that they can too easily dismiss anyone telling them otherwise? Or are they perhaps (consciously or subconsciously) threatened by a standard that their current self can no longer match up to?
It's sad that with the way copyright and movie distribution work that these butchered version become the only version new audiences get to see.
> All of my kids spent hours pouring over a book we had called 'The Way Things Work' with a delightfully funny Mammoth and a good description of how things actually work
I grew up with this book - I have vivid memories still of the pages about a nuclear reactor - and I was pleasantly surprised to visit a bookshop recently and find it still in print, updated with new things like LIDAR, 3D Printers, MoCap, etc.
The author David Macaulay taught illustration at my college. His original series started with buildings like “Castle”. The way things work was a breakout hit.
This book is probably 90% of my understanding of how why so many transistors can add up to a CPU, and I didn't learn any programming for a good ten years after I pored through that book as a kid (for me it was "The new way things work", updated for the computer age).
> Usually, companies like Apple are largely to blame
Just as a point of fact, Micron and its peers have in the past operated an illegal cartel that engaged in a price fixing conspiracy to manipulate the cost of RAM.
> mega-buyers like Apple caused various issues, and the production equipment that was scaled back during that time is now leading to supply shortages in the upcycle [2]. Apple fans probably won't admit it, though.
That's Micron's problem. Nobody held a gun to their heads and made them accept the prices Apple was offering. Micron willingly took those deals.
It's true that memory companies formed a cartel in the early 2000s, and I'm not saying the memory companies did nothing wrong.
It's just that in the context of the current price surge, the biggest factor varies depending on how you look at it, but hardware manufacturing itself has problems with fixed-cost structures and demand monopolies. It's a problem of sunk costs. If fab utilization drops, losses pile up. In other words, a fab has to run 24/7.
So the problem is that if a fab stops running, astronomical losses begin, so it's natural to hesitate in investing in production facilities. And that hesitation is driven by fundamental issues of supply and demand. That's also why memory became expensive.
Ultimately, it seems you agree that the current prices were formed by these voluntarily created conditions, right?
>That's Micron's problem. Nobody held a gun to their heads and made them accept the prices Apple was offering. Micron willingly took those deals.
Then they risk losing Apple as a customer, with empty Fabs running which cost them most money. And pile of stock unable to be covered by any other player in the market.
The power is over at the buy side. The sell side have very little leverage. Hence they some times come together, and then they were called price fixing.
> Many such IDs are designed for local physical verification, like proof that the mobile phone owner is above certain age or has a valid driving license, they are not designed for remote verification.
This is incorrect, the Digital Credentials API[1] is designed so that identity information can be remotely verified in a cryptographically secure manner.
There is no reason Anthropic could not use the DC API for this in countries and states that support digital identity, I assume they simply aren't because they threw this together at the last minute and simply out-sourced it to Persona.
> they threw this together at the last minute and simply out-sourced it to Persona.
They could have vibe-coded their own verification system that uses DC APIs. It shouldn't take long, assuming that Anthropic still has access to Fable. /s
Google is competing with nvidia (TPU), AWS (GCP), Netflix (youtube), Tesla (waymo self driving), OpenAI (Gemini), Microsoft (Workspace), Apple (Android)....
I don't think so, their article says they were seeing 30,000 commits a day in September, when I was at Meta six years ago they were easily doing that.
reply