The AI Daily Brief: Artificial Intelligence News and Analysis - ChatGPT Just Became a Work Agent

Starting point is 00:00:00 Today on the AI Daily Brief, more new models plus a big harness update from OpenAI. And before that in the headlines, Cursor also appears to be developing a harness to go after the larger knowledge work sector. The AI Daily Brief is a daily podcast and video about the most important news and discussions in AI. All right, friends, quick announcements before we dive in. First of all, thank you to today's sponsors, KPMG, Rackspace, Blitzy, and Airtable. To get an ad-free version of the show, go to Patreon.com slash AI Daily Brief. Or you can subscribe on novel podcasts. And of course, to learn more about sponsoring the show, head on over to AIDailybrief.ai slash sponsors,

Starting point is 00:00:38 or send us a note at sponsors at AIDailybrief.A.I.D.? Quite appropriately, given that our main episode is about a big harness update, we kick off our headlines with News that Cursor is planning a general purpose agent to compete with Claude Co-work. The information reports that work began on the project in April, shortly after Cursor signed their deal with SpaceX. The agent is expected to use GROC 4.5 and will be Cursor's first project aimed at anyone other than professional coders. Called Sand, it's designed to function as a personal assistant performing standard office tasks like dealing with email or working with spreadsheets. It sounds like it could also eventually become a unified platform, with the information's reporting suggesting that the agent will also be functional at AI coding.

Starting point is 00:01:18 Sources said the platform was rolled out internally in June, however, it's still unclear whether it will get the greenlight for a public release or when that would happen. Certainly, the product suggests that Cursor, now part of SpaceX AI, is looking to grow beyond their traditional user base of software engineers. This makes sense, as we'll see, a huge theme of all of the announcements today with OpenAI are all about taking what has been working in coding and bringing it to a broader set of knowledge work. Now, speaking of what's working with coding and what's not working, OpenAI has captured the zeitgeist and declared that the leading coding benchmark is bunk.

Starting point is 00:01:50 In a new report, OpenAI audited Sweet Bench Pro and found the benchmark to be sorely lacking. In their testing, they found that 30% of the tasks on the benchmark were broken and are now formally retracting their support of the benchmark. Many of the issues stem from some of the tasks being public, which can distort results by having those specific problems be included in training data. Others had hidden requirements, contradictory instructions, overly strict tests or incomplete grading criteria. Their conclusion was that SweetBench Pro, quote, no longer reliably measures frontier coding capability. And to be clear, the shift away from SweeBench was already well underway. Cursor has been using their own proprietary benchmark for months, while cognition and data bricks

Starting point is 00:02:26 also launch their own benchmarks this week. I think we're officially at the point where there's going to be lots of introduction of new benchmarks. Companies are going to present all of them, including, I guarantee they will still present Sweet BenchPro, and mostly just wait for people to have the vibes that confirm whether the benchmarks are bunk or not. Now, staying on OpenAI for a minute but going to a very different area,

Starting point is 00:02:45 the company has published a new statement that explains their approach to government and military partnerships, announcing their new national security principles. Open AI writes, We believe democratic society should be able to use AI to protect people, defend critical infrastructure, deliver public services, and respond to emerging threats. But increasingly capable AI systems must be deployed in ways that reinforce democratic accountability, meaningful human judgment and the rule of law, and strengthen democratic institutions

Starting point is 00:03:09 rather than concentrate power. Distilling the principles into a few key points, OpenAI explicitly stated that they will not support the use of their technology for mass domestic surveillance, high-stakes decisions, including decisions over the use of force without appropriate human judgment and accountability, or uses that evade legal obligations oversight and accountability. Now, these you might recognize are essentially the same as Anthropics' red lines, and given that that created such chaos with the government, it's not clear what the goal of this document is supposed to be. But I guess at least we have a clear articulation of what they stand for and something

Starting point is 00:03:38 to build from in their conversations with the government. Speaking of the overlap between the government and AI, Anthropic has appointed former Fed Chair Ben Bernanke to the board of their long-term benefit trust. Bernanke, of course, was appointed Fed Chair by the W. Bush administration in 2006, and served until 2014, presiding over the events that led up to and followed the global financial crisis. He is one of the more controversial Fed shares in recent history, with views on him differing depending on whether you think the bank bailouts in 2008 saved the world from a depression, or you view them as the original sin that accelerated the wealth divide in the United States. When Claude was asked

Starting point is 00:04:12 to provide an objective viewpoint on Bernanke, it responded that he's generally well regarded but has critics on both the left and the right who view him as, quote, emblematic of an unaccountable technocracy, protecting elite interests. So what then is Anthropics' long-term benefit trust? In their own words, they write, Anthropic is a public benefit corporation, meaning the company was created to balance commercial success with generating social and public good. The benefit trusts exist to help the company responsibly maintain that balance over the long term, providing a check on how Anthropic develops and deploys AI. Essentially, this is an independent oversight and advisory board that sits on top of Anthropics' normal board of directors. Importantly, the benefit trust has the power to elect or remove one

Starting point is 00:04:50 member of the corporate board, which escalates to two and then three according to time and funding-based milestones. By next year, they will have majority control over the corporate board, albeit with a shareholder override that requires a supermajority vote against the actions of the trust. Unlike normal corporate boards, no one on the board of the trust is allowed to be a shareholder. There is a lot of discourse on this one, especially from folks on FinTwit, but I think Tom Bruny got at the key question that many of us are asking, will he bring token prices to zero like he did with rates. Moving over to the chips world, Meta is heading north for their next big data center project with a $10 billion facility planned in Canada. Meta announced the groundbreaking for the

Starting point is 00:05:27 Alberta site on Wednesday with plans to build a total of one gigawatt of capacity over the coming years. Interestingly, community pledges were front and center in the announcement, with Meta planning to contribute $60 million in Canadian dollars to local infrastructure upgrades, including roads and water. On electricity, META pledged to pay full infrastructure costs, and officials expect local rates to actually decrease as a result of the project. They will also provide direct funding to local nonprofits. In addition, they plan to hire 3,000 construction workers at the peak of the build and support 300 ongoing operational jobs. Now, as I continue to scream from my bully pulpit, these sort of community contributions can be on the one hand incredibly valuable to those communities

Starting point is 00:06:04 and are a rounding error in these data center budgets, making it one of the easiest potential alignments between hypers in the communities they operate in than they could possibly imagine if they only take the time to actually do it, so it's great to see Meta actually going down that path. Now, the other big implication here is, of course, that Meta's big AI buildout is not slowing down. When news broke last month that Mehta was planning to sell extra capacity, many drew the implication that they'd wind back on CapEx. Adding another gigawatt suggests Meta's plans are still ticking along at a rapid pace, and I think when you get to the main episode, you'll kind to start to understand a bit more why. Meanwhile, meta's in-house chip program is back from the

Starting point is 00:06:41 dead and set to ender production in the coming months. According to an internal memo, cited by Reuters, Meta is on track to begin producing the first chips in September. The report states that at least one chip design sailed through the testing phase in just six weeks. Meta is working with Broadcom on design and will contract with TSM for processors and Samsung for memory. Most semiconductor analysts assumed this generation of chips would be shelved, just like all the others.

Starting point is 00:07:03 Meta has been working on their own advanced AI chip since at least 2022, but each iteration hit Roblox and was never put into full-scale production. This generation had a similar vibe up until now. It was first announced in March just one month after Meta had canceled the previous generation, with that announcement describing four configurations across the chip family tuned for different inference and training workloads. Then last month, it was reported that Meta had asked Samsung to temporarily pause development on their custom chip set, and many assume that was the beginning of the end.

Starting point is 00:07:30 To the contrary, it now seems that Meta has a finalized design and just needs to wait for some foundry time with their chip-making partners. The chips are expected to be installed in Meta's data centers and allow them to cut down on their spending with Nvidia and AMD. meta also plans to design a new chip every six months beginning next year, which would be significantly faster than the standard industry cadence. The memo viewed by Reuters also reaffirms Meta's plan to scale compute. The company said that they plan to deploy seven gigawatts of capacity this year and double

Starting point is 00:07:56 that pace in 2027. So yeah, doesn't seem like we're getting a cutback for meta anytime soon, but again for why, I think we need to end the headlines and move on to the main episode. One of the most important AI questions right now isn't who's using AI. it's who's using it well. KPMG in the University of Texas at Austin just analyzed 1.4 million real workplace AI interactions and found something surprising.

Starting point is 00:08:25 The highest impact users aren't better prompt engineers. They treat AI like a reasoning partner. They frame problems, guide thinking, iterate, and push for better answers. And the good news? These behaviors are teachable at scale. If you're trying to move from AI access to real capability, KPMG's research on sophisticated AI collaboration is worth your time. Learn more at KPMG.com slash us slash sophisticated.

Starting point is 00:08:49 That's KPMG.com slash us slash sophisticated. One of the more interesting shifts in enterprise AI right now is how quickly the conversation is moving towards infrastructure and operations. As AI moves into core workflows, regulated data environments, and agentic systems, enterprises need governed infrastructure and inference that can operate reliably day-to-day with clear operational accountability built in from the start. As those systems scale, the operating model increasingly becomes part of the AI strategy itself. Rackspace technology is the operator of the full Enterprise AI stack, from agents to infrastructure

Starting point is 00:09:23 across private cloud, hybrid cloud, and edge environments. Rackspace builds and operates governed AI infrastructure inference and production AI systems for organizations where sovereignty, compliance, and uptime are non-negotiable. Therefore, deployed engineers stay embedded beyond deployment to help operationalize and run AI in live environments. To learn more about where enterprise AI runs and outcome scale, go to rackspace.com. Blitzy is driving over 5x engineering velocity for large-scale enterprises.

Starting point is 00:09:50 A publicly traded insurance provider leveraged Blitzy to build a bespoke payments processing application, an estimated 13-month project, and with Blitzy, the application was completed in live in production in six weeks. A publicly traded vertical SaaS provider used Blitzy to extract services from a 500,000-line monolith, without disrupting production, 21 times faster than their pre-Blitzy estimates. These aren't experiments. This is how the world's most innovative. enterprises or shipping software in 2026. You can hear directly about Blitsey from other Fortune

Starting point is 00:10:18 500 CTOs on the modern CTO or CIO classified podcasts. To learn more about how Blitsey can impact your SDLC, book a meeting with an AI solutions consultant at blitzie.com. That's BLiTZY.com. This episode of the AI Daily Brief is brought to you by HyperAgent, where you run fleets of agents your team can manage together. New users get $1,000 in inference. Forget local agents and chat workflows waiting on your laptop to be prompted. Hyperagent deploys always on agents in the cloud, doing real work across the tools your team already uses. Marketing's agent turns competitor moves into landing pages. Sales as agent enriches leads, drafts emails, and updates the CRM. Ops agent chases the paperwork and tracks the budget. Every agent has access to shared context

Starting point is 00:11:00 and follows your rules about scope and approvals. It's time you add agents that feel like teammates. Hire yours at Hyperagent built by the team at Airtable. Claim your $1,000 in inference at hyperagent.com slash AI Daily Brief. Welcome back to the AI Daily Brief. This week, obviously, the entire story has been new models. I mean, it has just been a barrage. We've got GBT 5.6, the entire family, which are finally officially out, available for everyone with all the related public release information.

Starting point is 00:11:33 But then also the fairly unexpected GROC 4.5, at least unexpected in terms of its apparent performance, as well as a new one, Muse Spark 1.1, out of left field that actually has people talking about meta again. And yet, in a year where people have come to understand just how significant not only the model is, but the harness in which it operates, it is so appropriate that we are also closing out the week, with in some ways the biggest news being big changes to the OpenAI harness through which you are going to use the model. OpenAI has released chat GPT work, and we are going to talk about what that means in first impressions. But first, let's finally catch up on the now fully released GPT 5.6 model family. Now, this is the first time that OpenAI

Starting point is 00:12:17 has split their model family into a set of classes. In addition to the flagship model, Seoul, GPT56 also has a mid-sized version in Terra and a small cost-efficient version in Luna. Now, we have started to talk about these models throughout the week, as OpenAI gave clearance to people to start talking about their early adopter impressions of it, but we just now, with the official announcements yesterday, got the benchmarks. And interestingly, the benchmark story as it's presented is very different than the way that OpenAI has done this in the past. Sam Altman tweeted obviously the best model we've ever produced, but also one of the best blog posts we've ever produced, and part of that is that instead of just a simple, easy table of a bunch of numbers for their

Starting point is 00:12:53 benchmarks, OpenAI is now highly focused on charts that show performance per cost. For example, on the agent's last exam test, which is long horizon agentic workflows across a number of professional domains, the chart is actually presented in three ways. The score on the y-axis and some other variable on the X-axis, with the premier chart being the API cost. They show the comparison not only to GPT-5-Claude Opus 48 and Claude Fable 5 in terms of the score, but also in terms of the cost, with these strong emphasis being that not only do the GBT-56 series of models perform better, but they do so at a much-reduced cost. Now, you can also vary that chart to be organized by latency or by output tokens used, which is, of course, another measure of efficiency. You see a

Starting point is 00:13:38 similar chart presented this way for the artificial analysis coding agent index. Again, cost, latency, and output tokens. And this is pretty much now the standard for how they present the benchmarks. Sam Altman reinforced this focus with one of his announcement tweets where he said, We have heard enterprises on their concerns about AI costs and 5.6 sole is a huge step forward for dollars per task, as are Terra and Luna. Now, rather than trying to explain all of these different charts on a podcast that many, if not most of you still consume via audio, the big overview is basically that 5.6 sole on max settings, benchmarks ahead of Opus 4-8, near or above Fable 5, and most importantly, at a significantly lower cost to each. For the artificial analysis

Starting point is 00:14:22 index run, GPT 5.6 was a close second to Fable 5, falling short by a single point, but it completed the run at a third of the cost of Fable and was even 40% cheaper than Opus 4-8. Meanwhile, on the coding agent index, GPT 5.6 is the new state of the art, beating Fable 5 by 3rd. three points. In fact, at least on that artificial analysis benchmark, the mid-sized Terra variant performed at the same level as FABEL and would of course be significantly cheaper as a daily driver coding model if those benchmark results translate to real-world experience. Now, bringing it back to the model that had so much attention in the Fable 5 shutout period, Simon Smith pointed out that 5-6 Luna actually matches GLM 5.2 on the artificial analysis intelligence index and comes in at 43% cheaper.

Starting point is 00:15:06 getting at the idea that we were talking about in our episode a couple of days ago, of the new market opportunity for more efficient Western models. Simon continued, I'm glad openweight's models exist because they push Frontier Labs to innovate and release, but I don't think enterprises will wholesale shift to open weight models strictly for cost reasons. I think Frontier Labs will optimize for both intelligence and efficiency, offer models at multiple performance and price points, and train their best models to know how and when to use their cheaper models to maximize impact for cost, and that will make for a compelling value proposition that negates the need to shift to open weight models strictly to save money. Now, as I mentioned, since early testers were allowed to break

Starting point is 00:15:44 their silence from the beginning of the week, we've been able to cover a number of first impressions over the last couple of shows. And to recap, the big takeaway is that we now have two frontier models that perform very differently. Broadly speaking, the early consensus is that Fable 5 is the big model for massive long-running tasks, especially ones where there can be a lot of a autonomy, while GPD 5.6 sole is a huge upgrade not only as a daily driver, but for large tasks where you want to be more involved in the intermediate decisions. Reinforcing some of the early takes, every CEO Dan Shipper wrote, 5.6 is powerful, fast, half the price of Fable, and my default for almost everything. He noted that 56 is not as good as Fable for coding, but also pointed out that

Starting point is 00:16:25 Everry's benchmark for coding is a massive long-running code refactoring task, which Fable is naturally more suited to than 5.6. Fascinatingly on writing, he said that 5-6 is the model that every has liked more than anything else, arguing that it's clear and more concise than anthropic models, and throughout his review, Shipper continuously comments on the speed. Fable is a big, slow, incredibly powerful model, while GPT5.6 is very fast, making it more of an active collaborator rather than a model you leave running and come back to. In one of the biggest green flags for a lot of you listening, Shipper said that the real leap is around knowledge work. Soul, he wrote, is the first model I've trusted to run whole loops of

Starting point is 00:17:01 knowledge work, not just help with individual tasks. It has shifted my job from doing the work to tending the system that does it. Now, probably many of you have spent the last couple of months, listening to all of these new conventions like slash goals and discussions of loops, not exactly sure how they relate when you move outside of the coding domain and into knowledge work, and so the fact that Dan and others are starting to experience some of that with 5-6 soul is pretty notable. Summing up, he wrote, if I really had to put my finger on it, I'd say Fable has way more big model smell.

Starting point is 00:17:29 But that means it's a skill in itself to get value out of it, and 99% of people are still not there yet. GPD 5.6 is almost as powerful, but it's easy to use, fast and relatively cheap. It should give you an early sense of where model work is going. Theo presented his review in the form of a portfolio posting. Over the last month, I burned over $200,000 in tokens with GPD 56 Sol, I built a lot. And if you listen to the way that Theo describes his building process, it once again reinforces the idea that whereas Fable is a model that you let it go off on its own, to do things,

Starting point is 00:17:58 5.6 is one you interact with. Now, overall, the sentiment towards 5-6 soul is extremely positive. And it feels to me like, to some extent, it's also being the beneficiary of the fact that heading into the Fable 5 ban, Open AI kind of had all the momentum when it came to the most enfranchised AI users. Codex had increasingly become the default harness that people were looking for, and before Fable at least, a lot of folks has shifted over from Opus 48 to GBT5.5. Now, given that we've been so excited to have Fable 5 back, we haven't talked.

Starting point is 00:18:30 for a while about some of the constraints that were placed on the model when it was announced, but certainly for some buyers, those concerns haven't gone away. Gurglia Rose tweeted, Interesting take on Fable versus GPT 5.6 Seoul, from a dev at a large and AI bullish company spending lots of money on AI. They told me, Anthropic has not changed their data retention policy on Fable, meaning they would store our data, so we cannot use it. We're going hard on GPT 5.6Soul as a result. But as I said at the top of this, the story of 2026 is not just a model story. In fact, in many ways, it is equally, if not more, a harness story. As we move into the agentic era, the systems that we put around our models to allow them to

Starting point is 00:19:11 access tools, interact with other models, spin-up subagents, access context, etc., are every bit as important as the underlying model themselves. It's pretty clear that part of the reason that OpenAI started to let people talk about GPT 5.6 early is that they wanted the emphasis on the official announcement day to be on the harness updates that we got in a new user interface that they call ChatGPT Work. OpenAI wrote, ChatchipT Work is an agent that can take action across your apps and files, stay with a project for hours if needed, and turn a goal into finished work. Now, at a surface level, this is OpenAI's answer to Claude Co-Work, but it's also clearly the next step of their superapp strategy. This is a new agentic harness that takes the functionality

Starting point is 00:19:52 of Codex and extends that approach out to all knowledge work. OpenAI. The best way to learn how to use chat GPT work is to give it tasks you already know well. Analyze a month-end budget variance, turn source materials into a marketing campaign brief, or prepare for a sales meeting. You can follow its progress, answer questions, change direction, and improve important actions. Now, the interface supports everything you would expect. It has connectors for all of the typical tools, including Notion, Google Drive, and Microsoft 365, allowing easy access to full work context.

Starting point is 00:20:22 It also supports scheduled tasks and functions on a cloud instance, so your agents can work even when your laptop is closed. It's configurable for enterprise-grade security and access controls, as companies are coming to expect from Leading Edge AI tools. The early testimonials sound pretty compelling. Angela Ferrante, the head of enterprise that Zapier wrote, used ChatGPT work to build a repeatable system for reviewing thousands of leads each month. It traced customer touchpoints across Zapier's CRM, email, and other tools,

Starting point is 00:20:48 found where follow-ups broke down and generated a weekly executive dashboard that highlighted missed pipeline and revealed seven figures in potential sales. And really, across the entire set of announcement materials, the biggest part of the pitch is about taking work patterns in Codex and applying them more generally. Rather than micromanaging small components of tasks, OpenAI is encouraging knowledge workers to define a goal, load context, and use chat GPT work to complete large multi-step tasks. What's more, they said that their teams have already adapted to this new approach. All of their teams are now using ChatGPT work in Codex with OpenAI writing, in sales, chat chiptiept work turned a discovery conversation into a tailored proof

Starting point is 00:21:24 of concept for a mission-critical problem within 24 hours, a process that normally takes weeks. ChatGPT structured the notes, routed the request to a solutions architect, and collaborated with the technical team, freeing the lead to focus on the customer and serve as a high-value consultative partner. In finance, ChatGP2 work reduced month-end close and forecasting from days to hours by helping teams find source data, move it into Acceler Sheets, reconcile it, create slides, and verify the results. This lets the finance team spend more time understanding what's changed in the forecast,

Starting point is 00:21:51 explaining why it changed, and advising leaders on what the company should do next. Now, when it came to consumer response, it was a little bit more muted than I think OpenAI might have hoped. Peter Yang summed up one set of feelings, I think, what he wrote, I think that chat chip ETT work versus codex thing is confusing. It raises questions like, so developers aren't working, and what if I'm using it to plan my vacation? Is that work? I think the whole thing should just be called chatchipt, Codex or chat chiptee T codex, and there should be no tabs or toggles.

Starting point is 00:22:18 In my opinion, codex is not that complicated for a normie to understand and use. You're still chatting except now it can do stuff. The sooner this is all unified, the better. Ethan Mollock wrote, This is confusing. Claude Co-work was a purposefully more secure and slightly weaker alternative to code designed so non-coders couldn't cause too much trouble. I got that, but I don't understand what chat GPT work is

Starting point is 00:22:38 or what I am gaining or giving up using it versus codex on work. Now, some were even actively worried about the blending of the apps. Dan Shipper again wrote, I was extremely worried about this because I love the Codex app. OpenAI was caught in an interesting position, How to Make an Agent Orchestration app for regular ChatGPD consumers, coders, and businesses all in one app. They now split the interface between ChatGPT Work and ChatGPT Codex. They're basically the same except Work hides code, and Chad has been demoted to second-tier status for quick questions in either one.

Starting point is 00:23:08 It's not a big leap, but it's not a huge setback either, and it remains my favorite of the desktop agent orchestration apps. Still, when you're coming with a big new announcement, having the response be the merged app is fine, which is what Dan's headline was, is not necessarily exactly what you're want. I think that there are confusions and the naming conventions, and to get a real feel for this, it's just going to take some time for people to actually experience it and figure out what the benefits or the particular challenges are. Now, one additional feature that came alongside chat GPT work was their updated sites feature. This basically allows you to turn any knowledge work into a website or web app that you can share with other users inside your company, even if they're not using chat GPT, which I already did a whole show about how many different knowledge work use

Starting point is 00:23:49 cases I think you should be building websites for instead of the traditional artifacts. And so I think that making that easier and integrating better hosting is going to actually be a pretty significant change in how people output work with chat chvety. Ultimately, all of this is still first impressions. What's clear is that the OpenAI team is really excited about all of these things and seemingly fairly confident that as users get more reps on with both 5-6 and with the work harness, our second and third impressions are going to be even better than our first, but for that we will just have to wait and see.

Starting point is 00:24:18 Now, we knew that the official release of 5-6 was coming and at least 24 hours before OpenAI started teasing that a new harness was coming as well. What I don't think anyone expected was for Mark Zuckerberg from Meta to tweet for the first time in three years, announcing that Meadow was also releasing a new model today, this time called Muse Spark 1.1. Now, the first edition of Muse Spark landed in April, without much fanfare, but this updated version seems like an entirely different beast. On the benchmarks, the model looks competitive with Opus 48 and GPT-5-5. It beat both by a significant margin on Humanities. last exam, which tests web search, tool execution, and knowledge. On coding, it's in the

Starting point is 00:24:57 ballpark, a couple of points behind Opus and 55 on Terminal Bench 2.1, and between the two rival models on Swee Bench Pro. On DeepSwee, which is increasingly important as a standard, Muse Spark 1.1 was a little further behind with a score of 53.3% compared to 59 for Opus and 67 for GPT-55. Still, for Meta, who have struggled to produce a viable LLM since Lama 3, this is a big improvement. And where the model shows signs of actually leading the pack is on personal agentic tasks. Mew Spark beats both Opus 48 and GPT-55 on Jobbench, which tests the ability to complete real-life professional work.

Starting point is 00:25:33 It's state-of-the-art on MCP Atlas, which is a tool-a-thong-style benchmark that test the ability to gather information through MCP servers, and on DeepSearch QA, which is an agentic search benchmark, it's slightly behind GPD-5-5 and slightly ahead of Opus 48. Basically, this is a model that has glimmers of frontier performance, especially in the context of meta-focusing on more consumer-firm. friendly personal assistant style AI. Now, we haven't seen a ton of interaction with the model yet,

Starting point is 00:25:58 but it also doesn't fully appear to be one that just released with no one ever having had a chance to build with it. The team from Julius, for example, used Muse Spark to build a Minecraft clone inside Julius, which it was able to do in about five minutes for 73 cents and tokens. On the Vowse AI public LLM evaluation, Muse Spark 1.1 landed at number four, ahead of both GPT-5 in Grock 4-5, and as Val's AI pointed out, running three times faster than the top three models. Rayon from Val's wrote, wow, this model is fast. Across our benchmarks, we found it to be one-fourth the latency of Opus 4-8 and one-half the latency of GPD-5. I would expect meta to have incredible web infra, but really don't know what witchcraft they're pulling to host the model for such fast

Starting point is 00:26:39 inference at high rate limits. Still, the biggest thing that Rayon talked about, and that was really key to this announcement, was the cost. Chubby writes, MetaMews Spark 1.1. 1 is a very good agentic model, but above all, it is incredibly affordable and cost-efficient. For example, on VibeCodebench 1.1, it came in fifth overall, but the cost to test it was 92 cents compared to, for example, $5.9 to $0.05 for Opus 48 and $12.51 for Fable 5. Ray on again from Vowls wrote, The model is so cheap, I almost don't believe it. In practice, we see it's one-tenth the cost of both Fable and GBT-5-5.

Starting point is 00:27:13 If you thought open-source models would compete away margins, just wait till you see this. it's somehow cheaper to use Muse Spark 1.1 than host your own OS model. Leo really summed up the feeling of many when he wrote, for my second LMAO WTF moment of this week, meta just announced Muse Spark 1.1, and it's also a frontier level model competing with Opus 48 and GPT-5. Zuck and Elon are back. And I think as we zoom out to broader implications,

Starting point is 00:27:39 every single model this week, GROC 4.5, SWE 1.7 from Cognition, Muse Spark 1.1, and even GPT 5.6 had an incredibly strong emphasis on cost and efficiency in its announcement. Even 5.6, which was also competing for the state of the art, so much of the emphasis was about its performance in practice and the difference that that was going to make to overall token budgets. What this all means is not only that the AI race has shuffled,

Starting point is 00:28:05 which it clearly has, but we are now officially at the point where the labs themselves realize that they are competing on an entirely different vector than just frontier performance alone. At the beginning of this week, X-AI, now SpaceX AI and Meta, were not functionally in the conversation at all when it came to model selection, especially for enterprises. And they are now ending the week back in that conversation. In fact, semi-analysis even wrote, at the simplest level, there are three things you need to build

Starting point is 00:28:33 a true frontier model, data, talent, and compute. We believe meta is the only hyperscaler slash neolab on track to be world-class at all three, and therefore has the best chance at catching up with Anthropic and Open AI. The AI Tectonic plates have shifted once again, and in this case, we are all the beneficiaries. I am excited to dig more into these models over the weekend and in the coming weeks. For now, though, that is going to do it for today's AI Daily Brief. Appreciate you listening or watching as always. Until next time, peace.

The AI Daily Brief: Artificial Intelligence News and Analysis - ChatGPT Just Became a Work Agent

There aren't comments yet for this episode. Click on any sentence in the transcript to leave a comment.