00:00:00quite a lot is happening these days in the developer tooling and runtime space.
00:00:07I mean, you know, developer tooling, that stuff that you're using under the hood when you're
00:00:12building a website, something like Vite, or if you're into Python, something like UV for managing
00:00:19Python versions, or something like Bun for installing packages, and also, of course,
00:00:25for executing JavaScript code outside of the browser, that kind of stuff. A lot is happening
00:00:31there. For example, today, OpenAI released an announcement that they are acquiring Astral,
00:00:38which is the company behind the UV tool, which is a Python tool. And I'll get back to that,
00:00:44because even if you're not into Python, and I'm not very much into Python, that's quite interesting.
00:00:50Now, there have also been other developments recently. For example, the Vite team released
00:00:55their own full framework, or announced their own full framework, but I'll get back to that in a
00:01:01future episode. I have not been able to use it yet. I want to stick to these recent developments
00:01:08regarding to OpenAI, but also, Anthropic to some degree. Now, here's the thing. OpenAI acquired
00:01:18Astral, and Astral, as I mentioned, is a company that's behind popular Python tools like UV,
00:01:25which is, I'll admit it, the only tool. I've used Python, but I'm using it all the time when I'm
00:01:30working with Python. Now, I'm not primarily a Python developer. I rather work with TypeScript,
00:01:36JavaScript, but here and there, I, of course, have to write a tiny Python program,
00:01:42maybe a little web scraper, some utility program, whatever. So, of course, I do work with Python here
00:01:49and there, and I'll be very honest, I found it always quite frustrating before I discovered UV
00:01:59one year ago, or something like that. Because UV is much faster than PIP, it makes it much easier
00:02:06to spin up multiple virtual environments with different Python versions. It's super quick to
00:02:11switch between them. It makes version management, package management, a breeze. It makes working
00:02:18with Python so much more fun. I can't tell you how amazing this tool is. Now, yeah, it is really,
00:02:28really great. That's the tool, in the end, OpenAI bought, acquired. And the question, of course,
00:02:35is why? Why would they bother? I mean, this is a tool you can use for free. It's not something
00:02:40you have to pay for. So why would OpenAI buy it? Well, I would say for the same reasons why
00:02:48Anthropic acquired Bun last year in early December. Bun, of course, is a JavaScript runtime similar to
00:02:56Node.js, faster than Node.js. It supports TypeScript out of the box. And here's the first interesting
00:03:03clue. Anthropic, of course, wrote Claude Code. Claude Code uses Bun or is written in Bun, so to
00:03:12say. So it's written in TypeScript, of course, but it's based on the Bun runtime. As the Bun team said
00:03:18in their announcement of the acquisition by Anthropic, Claude Code ships as a Bun executable
00:03:25to millions of users. So Bun, in case you don't know, has that feature that you can compile your
00:03:30project, your files, your TypeScript files into a single binary, which you can then ship to upper
00:03:35users, which can run it without having Bun installed. So that's what Anthropic did. And they, in the end,
00:03:40bought, they acquired the runtime. One of their main tools, Claude Code, and I guess Claude
00:03:48Co-work as well, is based on. Now, we can already see why they did that in a recent Twitter post,
00:03:57a very recent indeed. It's also from yesterday when I'm recording this. There, Jared Sumner,
00:04:04the head of Bun, so to say, posted that in the next version of Bun, there will be a Bun web view
00:04:11API built into Bun that makes it easy to programmatically control a headless web browser.
00:04:17So no more playwright. You don't need to install extra tooling. It's built into Bun. Why would they
00:04:22do that? The original vision of Bun was to be an alternative to Node.js. So you could argue mostly
00:04:29a server-side runtime. A web view built in was probably not one of the highest priorities back
00:04:36then. Well, of course, it makes a lot of sense if we're thinking of tools like Claude Code,
00:04:42which are agentic tools, which of course benefit from having certain tools built in. I mean,
00:04:48if it's easy for Claude Code or Claude Co-work to spin up a browser and take a look at the website
00:04:55it built or interact with your online banking account if you want to, that of course helps
00:05:02Claude Code. That makes it more useful because, of course, all these companies have the vision
00:05:08of building a generic AI agent. We have those coding agents right now. They are very useful,
00:05:14and that's why I have courses on codecs and Claude Code, and you'll find links below if you want to
00:05:19learn all about them. But that's of course not the end goal of those companies. Those companies
00:05:24want to build versatile agents. So it makes sense that Anthropic bought the runtime that
00:05:31powers its main agent, Claude Code, and that they extend it in ways that makes Claude Code
00:05:37more versatile, that gives Claude Code easy access to certain tools it might need. So that
00:05:43makes sense, and I would argue it's the same reason for why OpenAI is acquiring Astral.
00:05:48Here we're talking about some Python tooling. Codecs, their main coding agent is written in
00:05:55Rust, not in Python, but these Python tools are also written in Rust. So they are of course
00:06:02acquiring a team of people that have the knowledge needed to continue developing codecs,
00:06:08and they acquire tools that can of course also be used to run one-off tasks or execute some
00:06:15utility scripts quickly through Python via those tools. So I would say the reason is pretty much
00:06:21the same. They are acquiring tools that hopefully from OpenAI's perspective make codecs or the AI
00:06:29agents they are working on more powerful, more versatile. And I mean, they're saying that in
00:06:35their announcement OpenAI, expanding codecs beyond coding. The goal is of course not to be stuck at
00:06:42coding. Code of course is the entry door to automating everything on a computer in the end,
00:06:49and that is clear to you, to me, and of course also to them. So of course they want to leverage
00:06:56tools like the tools built by Astral and continue evolving codecs into a general agent that can do
00:07:04all kinds of stuff. And I'm already doing that, actually. I'm using codecs, or actually I'm mostly
00:07:10using Pi, which is an amazing open source coding agent which you can use with your codec subscription
00:07:16if you want to. It's super minimal and extensible. I love it. I created a separate video about it,
00:07:21might create more content on it. I'm using that, but I'm also using it for more than just coding.
00:07:26I'm using it to analyze certain documents and automate certain tasks on my machine. So
00:07:33that is what I'm doing, that is of course what they want to do, and that is I think the reason
00:07:38behind this acquisition. Now with all these acquisitions, we always hear that the tools
00:07:44of course will stay open source, and I have no doubt that they will. Of course you can definitely
00:07:50argue or doubt whether the future focus of these tools will be in line with the motivation or the
00:08:01things that are important for the users of these tools. So for example you could argue whether
00:08:07people that are building web applications on top of BUN, if they need this web view they probably
00:08:13don't. But yeah, that is what I mean. The tools, the runtimes and so on, they stay open source,
00:08:18but of course the future direction, the future roadmap might not closely align with the priorities
00:08:25or the wishes of the people that used these tools in the past. But of course we should also give them
00:08:30the benefit of doubt, we don't know what will happen here to the the astral tools to UV and so
00:08:36on. And yeah, that's just my two cents on this and what I think of that and what I think where this
00:08:42entire ecosystem in the end is heading here and what the goal of these AI tools is. But as always,
00:08:49please also share your thoughts on this and let me know what you think. Is there a reason and what we
00:08:54can expect from acquisitions like this?