A (0:01)

Today on the AI Daily Brief, not one but two main episodes crammed into one. The first story is that XAI is being acquired by SpaceX, with the second main being All About OpenAI's new Codex app and the shifting paradigm of how we work with agents. The AI Daily Brief is a daily podcast and video about the most important news and discussions in AI. Alright friends, quick announcements before we dive in. First of all, thank you to today's sponsors, assembly robots and pencils, Blitzy and super intelligent. To get an ad free version of the show go to patreon.com aidaily Brief or you can subscribe directly on Apple Podcasts. To learn about sponsoring the show or pretty much anything else regarding the show, you can go to aidailybrief AI. That's my vibe coded command center where I keep everything with frequent changes. Frankly, right now the thing that I would like to point you to, which I have been pointing you to all week, is this January AI Pulse survey. I I'm going to leave it open for probably just a couple more days. I basically want to give people better access to actual data around what people are doing. In aggregate, it's things like which AI models are being used most and for what use cases. Give you a little preview. Right now, Anthropic is cleaning up in the most used model use case. In any case, anyone who fills this out will be thanked with access to the results a week before I share it with anyone else. Mostly though, you're just contributing to better overall knowledge of what's happening in the world of AI. Now the last thing that I want to point out today is that we have the rarest of AI Daily Brief episodes, which is the double main. Simply put, there were two stories which were too big to be shoved in the headlines, and I'm pretty convinced that we have some more big stories coming down the pipeline later this week, so I didn't want to wait on them either. The first is about Xai and SpaceX merging. The second is about the Codex app and with that said, let's dive in. When I was doing my 2026 predictions, when one of the things that we were looking at was the competitive race between the big foundation labs. In a lot of ways it was kind of easy to see where all the different players were relative to one another. Coming into this year, Anthropic ahead on coding and had started to use that to get ahead on Enterprise. OpenAI still having the best consumer recognition by far and seeing pretty steady gains in how consumers were switching. For example from Google search to ChatGPT search, but at the same time really losing out on some of the momentum or on the enterprise encoding side to Anthropic and on the more general side to Google's Gemini. Meanwhile, as Google Gemini was coming into the year with the wind in its sails and then on the other end of the spectrum there was Meta, who really needed 2025 to be a rebuilding year, but at least they had the Ray Bans and the biggest question mark in some ways was XAI and Grok. How are they going to carve out and differentiate their place in this whole race? Now one of the possibilities that we discussed was that they could start to leapfrog performance because of the investments that they had put into major infrastructure development in the form of the Colossus clusters. So that was one thing we were keeping an eye on. The other thing that we were looking for though is whether we were going to see synergistic alignment, let's say between Elon Musk's various companies. There were certainly indications that this was a strategy for XAI given that it had already combined with X slash Twitter and now the next stage in that muskie and consolidation has come as SpaceX and Xai are officially merging. The deal values the new entity at 1.25 trillion, with SpaceX valued at a trillion and XAI at 250 billion. The previous valuation on XAI was rumored to be 230 billion from their 20 billion Series E that closed earlier this month. So this deal is more or less flat from that. SpaceX wrote in its announcement post that it had acquired XAI to quote form the most ambitious vertically integrated innovation engine on and off Earth with AI rockets, space based Internet, direct to mobile device communications and the world's foremost real time information and free speech platform. This marks not just the next chapter but the next book in SpaceX and XAI's scaling to make a sentient sun to understand the universe and extend the light of consciousness to the stars. Musk, for his part, said that the merger was an acceleration of sci fi technologies that live at the intersection of AI and space. In the announcement note he wrote, my estimate is that within two to three years the lowest cost way to generate AI compute will be in space. This cost efficiency alone will enable innovative companies to forge ahead in training their AI models and processing data at unprecedented speeds and scale scales, accelerating breakthroughs in our understanding of physics and invention of technologies to benefit humanity. Musk said that orbital data centers would unlock further advances including self growing bases on the Moon an entire civilization on Mars, and ultimately expansion to the universe. He said that in the long term, lunar manufacturing could allow SpaceX to put 500 to 1,000 terawatts of AI satellites into deep space orbit annually. For reference, current data center capacity on Earth is around 200 gigawatts. And for Elon, this is more than just talk. The short term focus certainly seems to be on getting the first orbital data centers off the ground. Earlier in the week, SpaceX filed an FCC application to launch a million AI satellites into orbit. Thus far, no one has launched more than a single satellite as a test case, meaning that it's questionable whether the FCC will approve an orbital data center of this scope. There are currently around 15,000 active satellites in Earth's orbit, so this network would increase satellite density by two orders of magnitude. I'm sure many of you are having visions of Wall E when they return to Earth to see the orbit is entirely covered in space junk. At the Same time, though, SpaceX has been known to begin negotiations with regulators at a high level so they have room to compromise. They were recently approved to double the number of starLink satellites to 15,000 by 2031, and their original ASK had been 30,000 satellites. The application itself is a dramatic statement of Musk's ambitions. It suggests that the orbital data center network would be the first step towards becoming a Kardashev Level 2 civilization, meaning one that can harness the sun's full power. SpaceX is also arguing that orbital data centers will be cheaper and more environmentally friendly than Earthbound facilities. Satellites will be able to vent heat into the depths of space, reducing the need for active cooling and water use. They can also harness solar energy at a much greater efficiency due to the lack of atmosphere. In other words, if future AI technology really does require COMPUTE at the hundreds of terawatt scale, orbital data centers might be the only politically viable solution. So everyone is just accepting this explanation for the merger on face value, right? Yeah, not so much. As always, any news having to do with Elon Musk is as much a Rorschach test about what people think about Elon Musk as it is an interpretation of the actual news. And it's not just individuals, but many of the media outlets that think that this isn't just about orbital data centers. The information wrote a piece called what's really driving Musk's SpaceX XAI merger. The piece noted that SpaceX didn't explain how XAI would actually contribute to the ambition of orbital data centers. They're obviously a customer, but it's not obvious how an AI company would help bring that technology to fruition. Instead, argues the Information's Martin Peers, there's no question the move is financially motivated. Musk may be the richest man in the world, but he's facing the same financial realities the leaders of other AI startups face. It's very difficult to compete in AI development with deep pocketed tech giants like Google and Meta, which own cash machines in their advertising businesses now. He also noted that the merging of the two startups together won't necessarily solve issues of profitability. SpaceX investors said the company generated 15 billion in revenue last year with a profit of 8 billion. Some noted, however, that this is Ebida profit, meaning that it excludes taxes and depreciation. While EBITDA is a fairly well accepted as a non standard accounting measure for tech companies, that makes sense where depreciation is fairly low. The reason that some have an issue with this is that when your main business is rocketry and satellites, depreciation is likely a much larger factor. XAI remains a Cash incinerator Bloomberg reported this month that XAI recorded a Net loss of 1.46 billion for the September quarter on 107 million in revenue. Revenue did double quarter over quarter, but it will need to double four more times to catch up with their burn rate. Reports state that Xai spent 7.8 billion over the first three quarters of last year. All of that means that the combined entity is valued somewhere north of an ADX revenue multiple. That would obviously be on the upper end for public markets, although certainly isn't the most extreme valuation in private markets. But you gotta think that now that these companies are entering the hinterland kind of between private and public, maybe those numbers start to make some investors nervous. At the same time, as peers point out, the numbers don't really matter. What matters is the strength of Musk's aura. So let's talk about some of the themes in the conversation. Investor Ross Gerber summed up one frequently shared skeptical take. He tweeted x was out of money merged with Xai Xai out of money merged with SpaceX. SpaceX out of money merged with Tesla when they are all out of money compound 248 asked how SpaceX investors are going to feel about this, they wrote an 8020 split for SpaceX and xai seems like a horrible deal for SpaceX giving the number 5 if that AI company with limited real revenue and Twitter ownership of 20% of a truly world changing rocket company is ridiculously bad for SpaceX shareholders. SpaceX doesn't need XAI, XAI does need SpaceX. On the inverse side, there was a lot of memeing about how well Twitter employees have done in all of this. Nate McGrady posted the mom how did we get so rich? Meme with the answer your dad worked at Twitter, which got acquired by X, which got acquired by Xai, which got acquired by SpaceX. But what about people's interest in the big technology vision that Elon's presenting? Steve Howe of Bloomberg tweets, elon says that within two to three years the lowest cost way to generate AI compute will be in space. What's the Elon to reality multiple on that estimate? Nick Carter shared the portion of the post where Elon wrote, in the long term, space based AI is obviously the only way to scale. To harness even a millionth of our sun's energy would require over a million times more energy than our civilization currently uses. Nick added, does anyone believe this nonsense? On the flip side are the optimists and the accelerationists. Beff Jesos summed up that optimism in his post SpaceX can turn fuel into solar energy. Intelligence turns energy into economic value. SpaceX and XAI is the ultimate way to turn rocket fuel into unlimited revenue. Believe it or not, there are even some folks who are trying to navigate outside the Elon Rorschach test to explore just how possible this is. Engineer Andrew McCallop wrote a blog post called Economics of Orbital versus Terrestrial Data Centers. His summary in the subtitle it might not be rational, but it might be physically possible. It is way beyond the scope of this episode to explain, but I will include a link in the show notes because it is a really interesting resource for exploring this more now of course, for many, all of this really comes down to a planned ipo. The Kobayisi letter writes, the biggest IPO in history just got even bigger. The financial lot account, meanwhile, explored what it might mean for OpenAI. They write, did Sam Altman commit the biggest theft in world history? Elon Musk finally got revenge today by bringing SpaceX into XAI. All investment liquidity will be redirected from OpenAI into XAI, leaving SAM with nothing. A 5D chess move to destroy Sam and Elon made the final blow. I'm not convinced that that's exactly how it plays out. Even if they get XAI as a bonus is going to be about the uniqueness of SpaceX as a company, I think it's far more likely that whoever goes public first between anthropic and OpenAI has a bigger impact on the other than a SpaceX IPO has on either of the two Pure Labs. Meanwhile, as all this happens, it does feel like XAI is on the verge of some big GROK updates. Just a couple of days ago they officially Launched Grok Imagine 1.0. The update brings 10 second video generation, 720p resolution and as they put it, dramatically better audio. And a lot of people are impressed. Last week, on the same day that Moat Book started going off and Google launched Genie 3 Swix and latent space argued that the bigger story was XAI launching the new state of the art image and video generation model. We'll come back and look at that in more detail in the future, but the point is that Elon's not just playing the financial engineering game. It's very clear that GROK and XAI are still out to win on their own terms as well. So how to feel about all of this? I think on the one hand it's incredibly easy to be cynical about something that seems as sci fi as data centers in space. Elon himself is of course a hugely polarizing figure right now, which has certainly not been made less complex by the release of the Epstein files. And yet I do think that there's a broader question here. Do we want all of our AI efforts to be about better short form video and more automated ad units, or can we still get ourselves excited about big crazy ambitious things which seem so insane that our first instinct is to plumb for ridicule? I actually think Tom Nash's simple tweet sums it up. SpaceX and the idea of an orbital data center When SpaceX talks about infrastructure, I listen, even if it sounds extreme. Big systems usually start as uncomfortable ideas. And frankly, this isn't just theoretical. Remember when Starlink started, its economics did not make sense at all. Since then, the cost to launch satellites has come down 20x and made what many believed was a total pipe dream into something economically valuable. That is not an argument a priority that orbital data centers are going to work, but simply to remember, to slightly paraphrase Arthur C. Clarke, that sufficiently advanced technology is indistinguishable from what appears to be the batshit crazy ravings of an online lunar if you're building anything with voice AI, you need to know about assembly AI. They've built the best speech to text and speech understanding models in the industry, the quiet infrastructure behind products like Granola, Dovetail, Ashby and Clulee. Now, as I've said before, voice is one of the most important modalities of AI. It's the most natural human interface and I think it's a key part of where the next wave of innovation is going to happen. Assembly AI's models lead the field in accuracy and quality, so you can actually trust the data your product is built on and their speech understanding models help you go beyond transcription, uncovering insights, identifying speakers and surfacing key moments automatically. It's developer first. No contracts pay only for what you use and scales effortlessly. Go to semblyai.com brief, grab $50 in free credits and start building your voice AI product today. Most companies don't struggle with ideas. They struggle with turning them into real AI systems that deliver value. Robots and Pencils is a company built to close that gap. They design and deliver intelligent cloud native systems powered by generative and agentic AI with focus, speed and clear outcomes. Robots and Pencils works in small, high impact pods. Engineers, strategists, designers and applied AI specialists working together to move from idea to production without unnecessary friction. Powered by RoboWorks, their agentic acceleration platform teams deliver meaningful results, including initial launches in as little as 45 days, depending on scope. If your organization is ready to move faster, reduce complexity and turn AI ambition into real results, Robots and Pencils is built for that moment. Start the conversation@rootsandpencils.com aidaily brief that's robotsandpencils.com aidDaily Brief Robots and pencils impact at velocity. You've tried in IDE copilots. They're fast, but they only see local silos of your code. Leverage these tools across a large enterprise code base and they quickly become less effective. The Fundamental Context blitzi solves this with infinite code context. Understanding your code base down to the line level dependency across millions of lines of code. While copilots help developers write code faster, blitzi orchestrates thousands of agents that reason across your full code base. Allow Blitzi to do the heavy lifting, delivering over 80% of every sprint autonomously with rigorously validated code. Blitzi provides a granular list of the remaining work for humans to complete with their copilots. Tackle feature additions large scale refactors legacy modernization greenfield initiatives all 5x faster. See the Blitzi difference at blitzi.com that's blitzy.com. Welcome back to the AI Daily Brief. Coming into this year, Anthropic and Claude have seemed like an unassailable juggernaut when it comes to coding related use cases. It's been this way for quite some time, all the way going back to Sonnet 3.5 and yet it's been very clear ever since the launch of GPT5, frankly that that OpenAI was not going to give up without a fight. Sam Altman even recently recognized that part of the reason that 5.2 wasn't very good at writing is that they just deprioritized that use case versus everything relating to code. We've also for the last two models 5.1 and 5.2 gotten specific versions that were optimized for the coding use case. And now OpenAI is moving the competition away from simply the model and into the realm of the product. One of the things that happened with the launch of Claude code about a year ago now is is that developer behavior shifted away from graphical user interfaces and towards the terminal. Now many people have become terminal pilled. They have gotten into the efficiency of just using the command line to tell CLAUDE code what they want and having it happen without having to mess around with specific interfaces. There is also, frankly, I think, a feeling of coolness factor, especially as folks go from non technical to semi technical or whatever we want to call this in between stage where you're commandeering technical agents without necessarily being super technical yourself. For those folks, myself included, it feels rad to have the terminal open and be doing things in it. But there are also obviously limitations for that and a lot of things that GUIs give you access to. Well now OpenAI is making a bet once again on graphical user interfaces with the release of the Codex app for macOS. They write Codex app is a powerful new interface designed to effortlessly manage multiple agents at once, run work in parallel and collaborate with agents over long running tasks. And the need, they say, comes from the natural progression of how developers are working. Since we launched Codex in April 2025, they write the way developers work with agents has fundamentally changed. Models are now capable of handling complex long running tasks end to end, and developers are now orchestrating multiple agents across projects, delegating work, running tasks in parallel and trusting agents to take on substantial projects that can span hours, days or weeks. The core challenge has shifted from what agents can do to how people can direct, supervise and collaborate with them at scale. Existing IDEs and terminal based tools are not built to support this way of working. This new way of building coupled with new model capabilities demands a different kind of tool. Which is why we're introducing the Codex desktop app, a command center for agents. So right away you can hear that this is not just Codex code or chatgpt code. This is a bet on where the paradigm is heading effectively OpenAI is saying we've moved from the autocomplete era to the IDE and command line era to now the agentic and sub agent era, where quote unquote developers are actually orchestrators of agents doing a bunch of things all at once. And with the Codex app they are arguing that behavior needs a new interface and a new set of tools. Pretty much everything that they point to as what you can do with the Codex app is for that new approach to building work with multiple agents in parallel, seamlessly switch between tasks without losing context. Built in support for work trees so multiple agents can work on the same repo without conflicts. Now similar to Claude Cowork, the team behind codecs at OpenAI shared that a lot of the code for Codex had come from Codex itself. Andrew Ambrosino writes, today we're introducing the Codecs app, our flagship codecs experience work on multiple things in parallel, extend codecs with skills and automate repetitive tasks. The most exciting part for us has been using the app to build itself. Thibault on the Codex team writes, Codex now pretty much builds itself with the help and supervision of a great team. The bottleneck has shifted to being how fast we can help and supervise the outcome. Not that you would expect any different, but the team and leadership at OpenAI definitely seems excited about this. Sam Altman tweeted, Codex app is out for Mac. I'm surprised by how much I love it. It's a bigger step forward than I imagined. President Greg Brockman wrote, I've been a die hard terminal and Emacs user for many years, but since using the Codex app, going back to the terminal has felt like going back in time. Feels like an agent native interface for building and other people agree. Prominent AI encoding YouTuber and content creator Theo dropped a 22 minute video about how much he loved the product first, caveating all the reasons why he had to dislike it. For him, the comparison is clearly not just to Claude code, but also to cursor with him, even calling it a cursor killer. Nick Farina writes, yeah, the new Codecs app is the best UI for AI assisted coding that I've used so far. It's incredibly intuitive and manages to provide a ton of features that reveal and unfold naturally as you use the product. Swix and the team at Leyton Space said, we almost did not give OpenAI the title story today. XAI technically got acquired by SpaceX and after all it's just air quotes a desktop app UI for the already existing CLI and cloud app and VS code extension and it's quote unquote just OpenAI's version of Conductor and codecs monitor and antigravity's inbox in December they pointed out that the integrated developer environment would die and here we are in 2026 they write an OpenAI which once offered 3 billion for windsurf is out here shipping a coding agent UX that is not a VS code fork. Bears some thought on truly how far coding models have come that serious coding apps are shipping without an ide. There was a time when app that lets you write English and build without looking at code was equivalent to Vibe coding, but these non technical audiences are not the ICP for codecs. This is very seriously marketed at developers who historically love code and identify strongly with handwriting every line of code. Now OpenAI is saying let's looking at code is kind of optional. They also say that Codex's reliance on multitasking and work trees is quote in hindsight, the perfect natural UI response to the increase in agent autonomy. The team at EVERY did one of their classic Vibe checks and basically found Codex was good. In perhaps understated fashion, their headline says OpenAI's Codex app gains ground on CLAUDE Code, but if you dig in more, there's definitely a lot of behavior shift here. Every founder Dan Schipper writes, previously I was using Claude code 80% of the time and Codex 20% of the time. Over the last few weeks in the app, that Percentage has become 50:50. For large production apps, codecs is slower but smarter and more reliable than Claude code. Opus 4.5 is still my daily driver for the rest of my work and for programming tasks that require taste, empathy and speed. But the reversal is significant indeed. Most of the team at Every gave it a green or psyched about this release rating, and in fact, it seems like the only reason that they didn't mark it as more significant is that they echoed what SWIX and the team at Leighton Space said, saying that this is built for hardcore engineering. And honestly, for the first time in a very long time, you're starting to see some chinks in Anthropics armor. On Sunday, Yu Chen Jin wrote, The creator of Claudebot Multbot Openclaw, Peter Steinberger, pushes 144 commits per day on average. Pre AI. This was impossible. He ships codes he never reads. He's a conductor. GPT and Claude are his orchestra. Five to ten AI agents run in parallel under his command. One person is now an army. Peter bit back kind of harshly, I don't like Claude code on my code base. It's all Codex would be too buggy with Opus now for Peter, this is not just a brand new opinion. This is something he's thought for a while, but more people are taking notice of it, in part because of the success of openclaw. Daniel Hassan points out that even with just the early reactions, it does seem like interface makes a difference. It's interesting, he writes, how the Codecs app has increased people's usage of codecs. The same model exposed under a different interface can make it so much more useful. Now, a couple more features that I think are worth noting. First of all, they have something called Automations, which is basically a cron job or a scheduled task. At OpenAI, they say we've been using automations to handle the repetitive but important tasks like daily issue triage, finding and summarizing CI failures, generating daily release briefs, checking for bugs, and more. There's also a big emphasis on skills, and you begin to see that OpenAI is clearly not just thinking about code generation, but also the way that code generation leads to everything else. Or as someone recently put it, code AGI is functional AGI in any case, the blog announcement writes, Codex is evolving from an agent that writes code into one that uses code to get work done on your computer. With skills, you can easily extend Codex beyond code generation to tasks that require gathering and synthesizing information, problem solving, writing, and more. The anonymous I rule the world account on Twitter writes, the pattern keeps repeating. Claude code dropped and everyone filed it under DevTools for months. Then people realized it was a general purpose computer agent wearing a coding hat. Codex just launched a proper desktop app. Same energy. It's not a coding tool, it's an agent command center. Parallel threads across projects Browser Automation Skills Library 30 minute autonomous runs the coding angle is just the entry point. If you're waiting for someone to explicitly announce this is the everything agent, you're going to keep being late. And when you go back and read the actual words that OpenAI is saying with that in mind, there's certainly a lot of evidence for that. The first sentence of the introduction to the Codex app calls it a powerful new interface designed to effortlessly manage multiple agents at once, run work in parallel, and collaborate with agents over long running tasks, the word code is implied but not present in any part of that. Many people think that we're likely to see all of these features shift over to everything very soon, krosinko writes. The features will easily be copied to all other agent managers in weeks. Gavin Purcell writes curious how all the Vibe code startups are doing with the Claude code explosion and now a very user friendly Codex app. I know some people are cursor hardcores, but overall the big seem to be eating the entire space. One thing that's for sure, whatever model people are using is that software engineers are settling into a very different reality. Also yesterday Sam Altman tweeted I'm very excited about AI, but to go off script for a minute, I built an app with Codex last week. It was very fun. Then I started asking it for ideas for new features and at least a couple of them were better than I was thinking of. I felt a little useless and it was sad. I am sure we will figure out much better and more interesting ways to spend our time and amazing new ways to be useful to each other, but I'm feeling nostalgic for the present Signal reposted that and said Nick St. Pierre echoed something similar Big identity crisis in many engineering circles right now, he writes, people who've historically considered themselves builders now realizing they aren't the ones building anything anymore. AI is the moral superiority of I build things, you just talk. Mentality is irrelevant now that the coding language is English and anyone can build things by talking. The skills that made them so economically valuable are almost fully commoditized and they're being forced to adopt a new identity, an identity most of them despise and have mocked their entire careers. To remain relevant, they must become the idea Guy. Signal, meanwhile, has put this consideration on an even larger scale. Resharing that post from Sam, they write, AI is going to sever the deepest identity loop in the west. That is, who you are is roughly equal to what you do for money. For centuries, trade was dignity, meaning social rank and even morality. Capitalism wielded cognition to occupation. What do you do? Became shorthand for what are you worth now? Machines are almost done encroaching this almost sacred layer for almost all white collar work. The next few years will be like watching God being forced to retire in real time. A little bit heady for this episode, but interesting that this is the conversation. I think the big takeaways right now are first, that OpenAI is making a bet on a different type of interface being necessary for a parallel agent swarm type of future. And second, for the first time in a long time, they're actually winning back some conversation equity and mindshare from Anthropic when it comes to this set of use cases. Alas, that may be very short lived. All signs continue to point to Claude sign at five coming very soon. In fact, perhaps even before you listen to this show. In between when I finish recording and when it goes live. So we might be once again in a whole different paradigm of capabilities in just a few short hours. For now, that is going to do it. For this episode of the AI Daily Brief. Until next time, be safe and take care of each other. Peace.