GitHub spent $13B stealing our code. It failed.

at this point, I don't even know what's voice. The fact that GitHub and Microsoft were allowed to use millions of lines of open source code to build GitHub copilot without ever paying anything for it, or the fact that this is not even the worst thing they did. It all started when Microsoft stunned the tech world and acquired open AI for 10 billion in its article, but actually checking Wikipedia, I it's about 13 billion. And the 13 billion promise was that copilot will increase developer productivity, many speculated, it will actually replace junior developers, but in the end it was built on open and stolen code. GitHub Open AI Codex was trained on a selection of English language public GitHub repositories. This includes a filtered data set of 159 GB of Python code from 54 million public GitHub repositories. Someone ring the bell and there was a lawyer who actually uh got GitHub and Microsoft to court. But as of today, the case has been dismissed, meaning that any AI company, not just Microsoft, can use your code to train their AI models without you being able to claim a copyright infringement. Meanwhile, GitHub copilot became the market leader AI coding tool. The estimate was that 90% of US-based developers already were using AI-powered coding tools. I remember the words of one Big Tech CEO, he estimated that engineers now complete their coding tasks twice as fast as they did before adopting the assistant in March. Even Microsoft CEO himself, Satya Nadela said that more than a million people have used copilot up to date in their earning calls in 2023. But then things started to crack. Articles like this say why copilot is making programmers worse at programming, with one of the most significant risk of relying on tools like copilot being the gradual erosion of fundamental programming skills. It also turns out that GitHub has been kind of lying with statistics

And this article explains it very well. Does GitHub Copilot improve code quality when uh GitHub Copilot was claiming that people who that developers who use Copilot write 13% more lines of code on average without error? It turn out we are talking about 2.2 lines of code. Which is not a lot, it can be luck, it can be the fact that they only studied 200 developers for this study. In reality, the code written with Copilot doesn't leave too long a report from give clear about AI Copilot code quality. We realized that after analyzing 211 million change lines of code, they realized that all percentages of code churn, like new lines churn, meaning lines that we are, you know, we're putting in in a certain sprint and we have to delete. increased at a almost exponential rate. When using Copilot, and I think it points to a bigger problem, the main challenge with AI assisted programming is that it becomes so easy to generate a lot of code, which shouldn't have been written in the first place. Because in the real world and even the most junior developer knows that, developer productivity, it's inversely proportional to the number of lines of code written. Funny enough, it was Bill Gates himself, the founder of Microsoft, who said, measuring programming progress by lines of code is like measuring aircraft building progress by weight. So why people are busy claiming their AI coding tools are generating so much code, senior devs are actually busy thinking about how can we write less code to solve this problem? Out of a sudden, copilot starts getting worse. People started reporting that suddenly GitHub copilot suggestions and performance was getting worse. Some people said it's it's it's horrible. Um that copilot it's it's rubbish and I'm tired. pretending that it's not, like these developer mentions, I've been using this tool for a month. Far from aiding my development process, it's slowed me down, making me a less effective developer. Microsoft was losing $20 for every 10 copilot AI subscription use.

and Microsoft fixed it by actually making it dumb boy. They added limits, they added enterprise plans, they added a context limits, which actually resulted in worse performance unless you are paying a lot or your company pays a lot. Now, losing money to gain user adoption is pretty common, in fact. It's there's not a problem with that. But I believe degrading the performance of your product while competition is pulling out better tools, it's a strategical mistake.

And this is exactly what happened. Because meanwhile, there were new kids on the block. And the new kids were actually better. We are talking about Anisphere, the company behind Cursur AI, who went from raising 8 million to raising 105 million into a 2.5 billion valuation. Then another AI code editor popped up. Windsurf was released. So, the Hunger Games began because people started comparing the performance of Cursur, of Copilot, of Windsuf.

They even started comparing with more generic models like ChatGPT and Claude. So, despite the huge investment Microsoft did, it wasn't clear that they were going to be the winner or that the investment is going to pay off because there were a lot of players in the marketplace. Because while Copilot was a quick add-on, Cursur and Windsurf reinvented the IDE by integrating all AI capabilities directly into the IDE, which is the native environment where software engineers are actually working in.

And giving even more flexibility to the developer. Microsoft tries to play dirty, and then one developer sounds the alarm and realizes that Microsoft, it's actually blocking Cursor and and Codium, the company behind Windsurf from basically accessing VS Code and building on top of VS Code. I mean, it was pretty obvious that Microsoft can do this, you know, VS Code, it's open, but it's theirs. So, if you compete with them, it might be less open.

We still have to see how this whole AI code editors battle will unfold. But it seems like, you know, after fighting with it for a while, Microsoft kind of accepted defeat. And as actually in talks to buy Windsurf. Not only that, but Microsoft just laid off more than 6,000 employees to supposedly focus on AI. I feel like all these investments are not paying off and the company is taking a toll. Actually, this is from the New York Times. Microsoft moderates AI spending as profit increases 18%.

So, despite this huge bet with AI, which might turn out okay, might turn out not so okay. The company is doing pretty well, they are laying off people, but they are reporting more profit. Which might not be the most ethical thing to do, but hey, you know, it's business, it's tech, that's what everyone it's actually doing. What I'm actually more worried about, it's the acquisition that Open AI is trying to do, the acquisition of Windsurf, which I wasn't able to find a lot of information on, and this is the answer that Perplexity gave me.

It turns out it's a high stakes move that will face intense antitrust scrutiny. I don't believe so, I think it should, I don't think it will happen. Why? Because Open AI has been one of the first investors in Cursor AI. When Cursor raised their CSA, the first round, that was about 11 million, and if I remember correctly, 8 million of that was coming from Open AI. Now, this might be a non-controlling stake, meaning Open AI might own part of Cursor, but they cannot control the company. Nevertheless, it's a bit shady, and I hope that DOJ and the US government actually keeps an eye on this, not to to damage the company or anything like that, but make sure that the market is competitive and that antitrust is not breached. All see you folks in the next one.