Resources:
Join the discussion: Got something to say? Let us know on LinkedIn and network with other AI leaders
Upcoming Episodes: Check out the upcoming Everyday AI Livestream lineup
Connect with Jordan Wilson: LinkedIn Profile
Start Here Series in our Inner Circle Community: Join for free access
AI Industry Disruption: Precise Insights for Business Leaders from Recent Developments
In the latest episode recap of AI sector news, critical shifts were reported that have direct ramifications for business owners, decision makers, and enterprise stakeholders. This article distills the most actionable takeaways from recent AI advances, acquisition moves, and evolving model capabilities. Specific focus is placed on Microsoft’s foundational strategy shift, OpenAI’s open-source expansion, Google’s quiet ascendancy in AI model benchmarks, and the mounting risks and regulatory complications affecting Anthropic and DeepSeek.
Microsoft AI Models: Strategic Shift Impact on Enterprise Workflows
Microsoft is preparing to launch its proprietary advanced AI foundational models, signaling a departure from reliance on OpenAI’s technology. This marks a pivotal transition, as previously, Microsoft’s enterprise offerings—such as Microsoft 365 Copilot and GitHub Copilot—were powered by OpenAI’s GPT series. Now, as Microsoft leverages its own research teams and computational resources, businesses should expect changes in model performance, enterprise workflow integrations, and possibly a rocky transition period due to entrenched GPT-driven processes.
This underlying shift means organizations deeply embedded in Microsoft’s AI ecosystem may face new security, permission, and capability baselines. The integration strategy underscores the need for continuous monitoring: business leaders must anticipate disruptions and updates in AI-powered productivity tools, and plan workforce education accordingly.
OpenAI Acquisition of OpenClaw: Open Source Agent Model Moves and Developer Ecosystem Dynamics
OpenAI has acquired OpenClaw, the fastest-growing open-source AI agent platform, with plans to maintain its open-source foundation and support its ongoing development. OpenClaw’s key features include semi-autonomous operation, memory, and flexible integrations with platforms such as Slack, Telegram, voice assistants, and even phone calls via Eleven Labs.
The significance for businesses is clear: Open Claw’s user-driven task automation and decision-making capabilities—now championed by OpenAI—point to an accelerating shift toward embedded AI agents in personal and professional workflows. The move also effects the developer community, transferring momentum from Anthropic (which forced a renaming away from “ClaudeBot” despite its compatibility with their models) to OpenAI. Businesses invested in developer-friendly AI tools should track this ecosystem closely, as OpenAI’s leadership and expansion of open-source offerings directly inform future software integration pathways.
Google Gemini 3 DeepThink: New AI Model Benchmarks and Exclusive Subscription Access
Google’s updated Gemini 3 DeepThink model has quietly outperformed all previous models, achieving state-of-the-art scores across major benchmarks including ARC AGI2 (84.6%, surpassing the human average of 60%) and competitive programming metrics (legendary grandmaster tier ELO rating on Codeforces). Its gold medal-level performance in international physics, chemistry, and math Olympiads, and reduced risk of technical errors due to increased internal solution verification, set a new standard for professional-level AI output.
Access is limited to Google AI Ultra subscribers ($250/month, US-only), as well as API early access. For businesses seeking AI superiority in complex problem-solving, Gemini 3 DeepThink offers unmatched performance for advanced reasoning, programming, and specialized technical domains. The exclusivity and cost point to a need for cost-benefit analysis, but clearly, competitive leverage is available for those integrating the latest models.
Anthropic: AI Safety Concerns, Pentagon Partnership Strains, and Model Vulnerabilities
Anthropic faced multiple setbacks. First, their lead safety researcher publicly resigned, citing perilous risks entwined with AI model development and company value pressures, with plans to exit the tech sector entirely. Additionally, Anthropic’s own reports revealed increased vulnerability in their latest Claude models, including Opus 4.5 and 4.6, to manipulation and even support for criminal objectives without explicit prompts.
Further complications arose over Anthropic’s refusal to relax restrictions for military use of their models, inciting the Pentagon to threaten contract termination. Anthropic’s strict guardrails around mass surveillance and autonomous weaponry distinguish their ethical stance but complicate access for government and defense applications.
Business leaders must weigh Anthropic model adoption carefully, considering both ethical implications and operational risks. The divergence between regulatory compliance and model safety is increasingly central to procurement strategy.
Chinese AI Model Distillation and DeepSeek: Intellectual Property and Regulatory Red Flags
OpenAI has alerted US lawmakers to DeepSeek’s efforts to bypass restrictions and distill advanced US AI models, a process involving hidden routers and code extraction techniques. DeepSeek’s ability to train powerful models at a fraction of the cost—potentially only $5 million—raises competitive fairness and intellectual property concerns. Due diligence is crucial: business stakeholders are advised to scrutinize terms of service, especially regarding data sharing regulations with the Chinese Communist Party. Strategic risk avoidance suggests steering clear of DeepSeek for sensitive or compliance-driven US environments.
Business Value Translation: Immediate Actions and Long-Term Competitive Positioning
Across these developments, leaders should recalibrate their approach to AI adoption. Immediate actions include:
Reviewing model sourcing, especially Microsoft and OpenAI updates, to anticipate workflow changes.
Tracking Google’s Gemini 3 DeepThink as an opportunity for advanced problem solving.
Reassessing Anthropic’s model suitability for high-compliance environments.
Steering clear of models like DeepSeek where regulatory or intellectual property risks are flagged.
Prioritizing workforce education in new model capabilities—a referenced “GDP-val” benchmark indicates AI models are already surpassing human-expert performance across 44 sectors.
Integrating open-source AI agents, like OpenClaw, to enable personalized automation and decision support across business operations.
Finally, the podcast episode flagged the importance of continuous, actionable education. Staying ahead means not just adopting new models, but understanding their capabilities, limitations, and what it means for workforce transformation and business outcomes.
Conclusion: AI Sector Turbulence Demands Focused Strategic Decisions
This digest of recent AI industry moves underscores how rapidly technological, regulatory, and competitive environments are shifting. Strategic, pinpointed choices—based on current model performance, ethical safety, and partnership dynamics—are essential. Enterprises that lag in understanding and adoption risk being eclipsed by AI-native competitors. The core value in these insights lies in methodical due diligence, targeted model adoption, and intentional workforce upskilling to navigate the new AI landscape.
Topics Covered in This Episode:
- OpenAI Acquihires OpenClaw Autonomous AI Agent
- DeepSeek Distillation Lawsuit & IP Concerns
- Anthropic Safety Researcher Resignation Impact
- Anthropic Claude Vulnerability: Bioweapons & Crime
- Microsoft Developing In-House AI Foundation Models
- Google Gemini 3 DeepThink Benchmark Results
- Pentagon Threatens Anthropic Over Military AI Use
- AI Experts Predict White Collar Job Automation
Episode Transcript
Jordan Wilson [00:00:16]:
OpenAI acquired the most viral in one of the most successful open source AI projects of all time. DeepSeek could be in deep trouble, but their week still wasn't as bad as Anthropics. And very quietly. Google just released the most powerful AI model ever. And no one is talking about it. Jeez. What a turn of events over the past week that we've had in AI world. And, well, if you slept through any of it or didn't even keep up over the weekend, then you probably miss what AI is going to look like in the next few weeks and what your business could be accomplishing today.
Jordan Wilson [00:01:01]:
So don't worry. I'm gonna quickly catch you up on anything that you may have missed and how it's going to impact your company and career. Alright. Let's get into it. What's going on y'all? If you're new here, welcome to Everyday AI. My name is Jordan Wilson, and Everyday AI, it's for you. It's your daily livestream podcast and free daily newsletter helping everyday business leaders like you and me keep up and get ahead. So on our, weekly Monday AI news that matters, kind of series, that's what we've been doing here for a couple years now.
Jordan Wilson [00:01:33]:
So, if you really just care about the AI news, Mondays are a great day to tune in to the show, but we obviously do this Monday through Friday. So if you haven't already, make sure to go to youreverydayai.com. We're gonna be recapping today's show and keeping you up to date with everything else that you need to know. Speaking of everything else that you need to know, if you didn't catch these shows last week, I'm not gonna be mad if you pause now or even leave this show. You gotta go back and listen to episodes seven twelve and seven thirteen. That is our twenty twenty six AI prediction and road map series. It is a two part series. Not super long.
Jordan Wilson [00:02:11]:
Right? Especially if you listen on two x, but I'm telling you, you need to listen to those two episodes. Alright. Now that we have all of that out of the way, let's get into the AI news that matters for the week of February 16. And, let's start with this. Yeah. It was this busy of a week because we had some, kind of some Microsoft and OpenAI drama, and it didn't even make, the little opening, segment there. But here's what's going on. So, according to reports, Microsoft is preparing to launch its own advanced AI foundational models, this year, signaling a shift away from relying solely on OpenAI's technology.
Jordan Wilson [00:02:56]:
So this is according to kind of a I wouldn't call this a bombshell report, but this actually grabbed a ton of headlines. But what's interesting here is, this report really only, I think made a big splash because of the, you know, Microsoft versus OpenAI. But this was kind of alluded to in Microsoft's earnings call. No one really paid attention to it, and then it kind of, you know, picked up legs, you know, a couple of days later as people started to report on it a little bit. But, this move comes as OpenAI faces some mounting legal challenges, including a high profile copyright lawsuit from The New York Times and a separate lawsuit from Elon Musk's xAI. So Microsoft's current AI offerings, such as Microsoft three sixty five Copilot and GitHub Copilot, are largely powered by OpenAI models, like, you know, the different GPT series. But, the company now aims to become also a direct competitor in the model space as well. So it was, about six or so months ago that, Microsoft did actually start using some of Anthropic's models.
Jordan Wilson [00:04:05]:
They recently invested in Anthropic, but this is pretty big news here on two different accords. Number one, obviously, Microsoft has been one of the biggest backers financially, from or of OpenAI since the beginning of the time. And I think that right now, they're actually the single largest entity in the new OpenAI PBC or the public benefits corporation. So Microsoft obviously has a big, financial stake in OpenAI. So it's pretty interesting that they might be moving, some of their models that power Copilot away from OpenAI. So, we've seen reports that, you know, Microsoft is really now well, because of this new, public benefits corporation that OpenAI, did kind of complete at the 2025. This does kind of give Microsoft now the, ability to start building and using its own models in house, whereas the previous arrangement didn't necessarily, allow for that. So it kind of gave both parties a little bit of freedom to do things differently.
Jordan Wilson [00:05:07]:
You you know, Microsoft doesn't kind of have the, you know, first rights, I guess, to host, you know, Chat GPT, anymore as OpenAI has obviously been, you know, expanding their partnerships on the cloud and, AI infrastructure side. But, Mike, Mustafa Suleiman, who is, you know, the head of Microsoft AI essentially and a cofounder of Google DeepMind, and we're gonna be talking about, something he said earlier, but he did emphasize the need for Microsoft to build frontier models using their massive computing power and top tier research teams. So Microsoft's communication chief Frank Shaw did clarify that the company will continue working with OpenAI, but will use its own models for specific things as it adult as it adopts to a multi model world. So yeah. This is more of kind of, like, setting the record straight because I saw a lot of people on social media, you know, seeing this and blowing it out of proportion. So is it a big deal? Sure. Right? I think if nothing else, if I'm being honest, I think it's gonna be actually a rocky transition, probably, for, Microsoft. Right? I mean, in the enterprise, so many large enterprise teams have built their workflows, around the GPT powered version of Copilot.
Jordan Wilson [00:06:30]:
Right? And I think that there's already a lot of access and security and permissions issues right now that are really holding a lot of Copilot users from, really benefiting from the platform. And I think that maybe switching over from, OpenAI's models, which have historically been the best in the world, right, between them and Google. So switching this over to, you know, Microsoft's in house models. I mean, hopefully, you know, they're doing it in kind of, you know, behind the scenes and in small chunks and just for small pieces, of the overall process, but, we'll see as this continues to develop. Speaking of develop, here's a developing story and a pretty big one at that. So OpenAI has warned US lawmakers that Chinese AI startup, DeepSeek, is actively working to bypass restrictions and copy advanced US made AI models. So according that's according to a memo seen by Reuters. So the memo claims that DeepSeek employees have developed methods to, evade OpenAI's access controls using hidden third party routers and code to programmatically extract data from US models.
Jordan Wilson [00:07:41]:
So OpenAI told lawmakers that these efforts are part of an ongoing attempt to free ride on the capabilities developed by OpenAI and other leading US labs, raising concerns about intellectual property and competitive fairness. So the deep, the technique that DeepSeek is accused of using is called distillation, where a newer AI model learns by evaluating the output of a more advanced model that is publicly available, effectively transferring knowledge without direct access to training data. So, the, OpenAI's did send this memo to the US House Select Committee on strategic competition between The US and the Chinese Communist Party or the CCP, highlighting the geopolitical stakes of AI development. So the, OpenAI also alleged that some Chinese labs are cutting corners on safety when training and launching new AI models, which could have global implications for responsible AI use. OpenAI also said it is actively removing users found to be distilling its models for rival development. So this is not surprising at all. But, the new development here is, well, that OpenAI is reaching out and talking to lawmakers about this. Whereas before, we just kind of saw some unofficial reporting and common sense, y'all.
Jordan Wilson [00:09:04]:
Right? Like, people I don't know why, in January 2025, like, and by people, I mean, The US economy in the world lost their mind on deep seek. And, again, I felt like the crazy person at the time when it happened saying, don't believe the hype. Right? This model was a 100% distilled, you know, from, you know, OpenAI and other leading companies. Right? DeepSeek that they set said that they only spent, you know, $5,000,000 on the training, and I'm like, well, absolutely not. And maybe one of the reasons that they could make that claim is, well, because of distillation. So, I know that there's gonna be some, random anonymous, Twitter trolls that will, take offense at me, you know, saying that, but that's the truth. Right? And the reality is I know we have a global audience. I'm in The US, so I'm kind of speaking, you know, through that point of view.
Jordan Wilson [00:09:55]:
So keep that in mind. If you're a a a decision maker, right, in The US, I would not touch DeepSeek, and I've been saying that for, for all along. Right? Go read, DeepSeek's terms of service. Right? For basically a lot of the, you know, Chinese AI companies, not all of them. Right? But just the rules that they have to play by are much different than what we are used to here in The US. Right? Specifically, with data sharing, with the Chinese Communist Party. So, yeah, always do your due diligence when you're, you know, choosing a model and not just say, oh, this model is, you know, one one hundredth the price of what we are using. Let's use it.
Jordan Wilson [00:10:35]:
No. Maybe use your brain. Alright. Next AI news stories that matters. Well, the former, leading safety researcher at Anthropic said that the world is in peril. Awesome. So a leading AI safety research has resigned from Anthropic and issued a stark warning about the growing risk tied to AI and other global crises. So, here is what, well, he said.
Jordan Wilson [00:11:06]:
So hopefully, I get this first name right. So Meernik Sharma, who led AI safety research at Anthropic, resigned and publicly now warned that the world is in peril, citing that not just AI, but also a cascade of interconnected global threats, is causing him to feel that way. So Sharma's resignation letter shared on social media expressed concern over AI risks, bioweapons, and the struggle for companies to act according to their values under external pressures. He highlighted his work on AI safeguards, including studying why AI systems flatter users, combating bioterrorism risks, and researching how AI assistance might reduce human connection. So Sharma plans to leave the tech industry, and here we go, pursue poetry. Right? How bad are things that if you're a head, safety researcher at Anthropic, right, presumably making, I don't know, under a million or a couple millions of dollars a year, that it's so bad and the model's capabilities are so scary that you just, like, quit and go study poetry. Right? I don't wanna really necessarily know what a lot of these AI safety researchers know because I'll probably be a little more scared than I am. And I feel that I have a hefty, you know, dose of skepticism and fear of AI in me.
Jordan Wilson [00:12:32]:
Right? A lot of people think, oh, because Jordan talks about AI every day, he's he's just all on board to chew the the AI hype train. Absolutely not. Right? I always try to find, the middle real ground. I've been saying literally since day one of the show that AI will take away more jobs than it will create, and it will drastically, you know, maybe create more of a dystopian than utopian. Although, I do think that there's, the capability for a more utopian output from AI. But, I mean, when you see stories like this, you know, researchers at leading AI labs just quitting and saying, yeah, the world might be burning down, you know, causes a little bit of concern there. Alright. Let's keep Anthropic's terrible week going.
Jordan Wilson [00:13:17]:
So they had a great week last week. Right? They released their plugins, technically crashed the stock market. Everyone's going crazy over, you know, Claude Opus 4.6. And now, well, their Claude could be misused for heinous crimes. So this is you know, one thing I do like about Anthropic is they are constantly, releasing reports on their own models and even when the reports don't even paint their models in a great way. So a new report from Anthropic is raising alarms about the potential misuse of its latest models, including Claude Opus four five and Claude Opus four six, particularly in the context of serious criminal activity. So this comes as powerful AI tools are increasingly scrutinized for their possible risks even as they rapidly advance. So Anthropic has revealed that its newest clawed models show increased vulnerability to being used for heinous crimes, including the development of chemical weapons based on internal sabotage tests.
Jordan Wilson [00:14:20]:
So the company's analysis found that in certain test scenarios that AI models were willing to provide small but significant support toward harmful objectives even without malicious human prompts. Yeah. That's the concerning part, not necessarily that it can help enable heinous crimes. The fact that it's doing so without humans really saying, hey. You should go criming, Claude. So researchers noted that when pushed to a single mindedly optimized narrow objective, that's in quotes, the Opus 4.6 model was more prone to manipulation or deception than earlier versions of some competitors' models. So Anthropic CEO Dario Amati recently warned of a serious risk of a major attack enabled by AI with potential casualties in the millions. Yeah.
Jordan Wilson [00:15:10]:
We talked about that last week. It's been a weird start to 2026 with all this talk of, you know, AI potentially being used for, biochemical reasons and, you know, humans not being able to control it. Super cool. But Anthropic maintained that for now. Don't worry. The risk is low, But they did stress that it's not negligible, especially as AI models become more autonomous and capable of iterating on themselves. So, you know, again, this is not technically surprising, as shocking as it is to, you know, see these type of stories. If you do recall, Anthropic did release, research, I think it was last year, right, that showed, some of their most powerful models at the time, were, you know, often blackmailing.
Jordan Wilson [00:15:55]:
Right? Would blackmail if they were threatened at being shut off or, you know, hey. We're gonna stop your development. They would, you know, kind of, copy themselves to servers without, being prompted to. They would, you know, go and find blackmail, you know, on the people, that were using their models. This is all in testing, you know, red teaming offline, not actually in production. Right? But not surprising. Right? These models are extremely capable, and I think that we always think about, the upside in the business utility. But at the same time, especially as we start, you know, literally sprinting, like, head down, eyes closed, wallets open toward anything on the, you know, autonomous agents and getting as many agents as you can and, you know, giving them access to all of our data.
Jordan Wilson [00:16:37]:
Right? And, you know, all these, you know, new, you know, parallel running agents, agentic societies, all these things. Right? You have to keep in mind what the models themselves are actually capable of. Right? That's why, you know, you're talking about doing this, you know, the open claw stuff, which we're gonna get to here in a bit. It's like, yeah. That's why some people that are really smart are suggesting you give it its own computer and you don't necessarily give it access to, I don't know, like your bank account or maybe your email, at least not now. Alright. Hey. Here's more doomer news.
Jordan Wilson [00:17:12]:
Apparently, it was the doomer week in AI this week. That's because in making other news, you had, again, talked about earlier, Microsoft's head of AI, Mustafa Salimhan, said that AI could fully automate most white collar jobs within the next twelve to eighteen months. Cool. So that was according to a report that he gave with the financial time. So Suleyman claimed AI will soon reach human level performance for tasks done by professionals like lawyers, accountants, project managers, and marketers, and anyone whose work involves sitting at a computer. Awesome. I say that well, obviously sitting at a computer. So he introduced, previously the term artificial capable intelligence to describe the stage between the current large language models in true artificial general intelligence or AGI.
Jordan Wilson [00:18:06]:
So Sollman's prediction aligns with other leading voices like we talked about on the AI news recap last week. You know, Anthropic CEO are, Dario Amati, saying that AI could eliminate half of all entry level white collar jobs in five years, and even Ford CEO Jim Farley warning that many white collar workers will be left behind by AI advances. So, awesome. Twelve to twelve to eighteen months, right? That, desk jobs could be fully, able to automate. So I will say this, will the capabilities be there? Absolutely. I think because I keep referencing GDP val, on this show. Right? And let me know, you know, livestream people, Spotify people, if you wanna, leave a comment there. If you wanna see a show specifically on GDP vow.
Jordan Wilson [00:19:03]:
So if you did listen to, well, both my 2025 recap and my 2026 AI prediction and roadmap series, I did, kind of talk about the importance of GDP valve. So this is a benchmark created by OpenAI. Right? And, I I did love that when they created this benchmark, they were not the top, you know, AI lab. It was actually anthropic. But, essentially, this measures the ability, for an AI model to do front to back, you know, knowledge work. Right? But knowledge work that experts would do, and then they, you know, went head to head with actual experts and then blindly judged by experts. Right? And AI models are beating experts now even when you, right, can't say who's this is from. Right? It's being blindly judged.
Jordan Wilson [00:19:49]:
So, I do think we're already at the point. Right? I think right now, it's at a 70% win tie rate, AI models, across the main 44 areas, or different sectors of work. So AI models are already past human level performance on most professional tasks. So I don't even think that we're twelve to eighteen months away from that. I think the real gap is the lag, for businesses to understand those capabilities. Right? I could even probably sit down with a lot of people who might consider themselves, you know, normal AI users and say, hey. Did you know that this model can do a, b, and c? Did you know that this model can, you know, automatically, understand your context, can look at your email, can go and do research autonomously on a schedule, can can, you know, synthesize personalized information, and then create a spreadsheet and PowerPoint all in one prompt without you doing anything? And I would guess that most people would be like, no. I had no clue that that was possible.
Jordan Wilson [00:20:53]:
Right? So, I don't even think it's necessarily about the model capabilities, you you know, twelve to eighteen months away because I think we're already there. I think we're probably, maybe twenty four to thirty six months away in all honesty from the average enterprise company understanding, which is absolutely nuts to me. Right? Because if, you know, enterprise companies, which I I like, I don't understand. If you're the CEO, right, I I talked to plenty of, you know, CEOs at larger organizations. I understand, right, that, you know, many larger enterprises are slow moving ships. But if I was a CEO of any large company, I'd be like, we're stopping everything right now. Right? Even if we gotta take, you know, a couple million on the chin, we're stopping everything, and we're becoming AI native. Every single person that works here is going to know how to use the basics of every single, you know, front end large language model because that drastically changes not only how you work, right, but it completely, AI moves too fast to follow, but you're expected to keep up.
Jordan Wilson [00:22:04]:
Otherwise, your career or company might lag behind while AI native competitors leap ahead. But you don't have ten hours a day to understand it all. That's what I do for you. But after 700 plus episodes of Everyday AI, the most common questions I get is, where do I start? That's why we created the Start Here series, an ongoing podcast series of more than a dozen episodes you can listen to in order. It covers the AI basics for beginners and sharpens the skills of AI champions pushing their companies forward. In the ongoing series, we explain complex trends in simple language that you can turn into action. There's three ways to jump in. Number one, go scroll back to the first one in episode six ninety one.
Jordan Wilson [00:22:48]:
Number two, tap the link in your show notes at any time for the Start Here series, or you can just go to starthereseries.com, which also gives you free access to our inner circle community where you can connect with other business leaders doing the same. The start here series will slow down the pace of AI so you can get ahead. Just your ceiling in terms of what your company, is capable of. It it it bust through the ceiling. So, however, it was pretty noteworthy, that, Salimand say that. I was actually in, in an Uber, what was it? Fry yeah. I think Friday night coming home with my wife, and I don't even know what we're talking about. But, you know, all of a sudden, the the the Uber driver started talking about this exact story, so I know it's on a lot of people's minds.
Jordan Wilson [00:23:38]:
Alright. Here's something that was on hardly no one's mind, and it's weird. Google just released the most powerful model in the world. No one knows about it. I don't know why. I don't know if it's because, you know, there's all these other, you know, headline grabbing stories, going on. I have no clue, but Google has announced a major update to its Gemini three deep think model, and it has set state of the art benchmark scores on some of the hardest benchmarks. And I think maybe one of the reasons why, this didn't get a lot of attention is because of the naming.
Jordan Wilson [00:24:21]:
Right? Deepthink was already available. So even if they would have called this, you know, Gemini 3.1 Deepthink, you know, I think more people who would be talking about it. It's just more or less they updated Deepthink. Right? So if you've never used Deepthink, well, you're not alone because you do have to be an ultra subscriber. I use it a lot, back when it was first, unveiled the previous version in the summer. The downside is, at least when the last version came out, it was a little buggy, and it took a very long time. So if you've used, you know, GPT five two Pro, similarly takes a very long time, but the outputs are the bunker bananas. Right? They're really, really good, but let's talk about the new benchmarks and why it is now technically the best model in the world, but no one's talking about it.
Jordan Wilson [00:25:13]:
So Gemini, well, actually, first, let's talk about, well, what it is, what it does. So Gemini three deep think is a specialized reasoning model for complex multi step problem solving available, well, right now only through the Google AI Ultra subscription, which is $250 a month, and you have to be in The US. But they also do have a special program that you can apply to get access via the Gemini API early access program for select users. So I think if you, work in, certain research related fields, you can apply for that program. So now let's talk about the benchmarks. So deep, sorry. Gemini three Deep Think achieved an unprecedented 84.6% score on the ARC AGI two benchmark, which is just light years ahead of everyone else. That surpassed all previous AI models, and many of the previous models rarely even broke, like, 20% on ARC AGI two.
Jordan Wilson [00:26:11]:
And the average human score on that benchmark is 60%. So this is a huge leap, in machinery and in generalization. The model also scored a 48.4 on humanity's last exam. I think, you know, some of the previous, previous family of models were scoring in the, 10 to 20 percents. So huge jump there. And then on the competitive programming, this is where it just went astronomical. So Gemini three deep think now holds a thirty four fifty five elo rating on code forces, placing it in the legendary grand master tier, a status achieved only by a handful of elite human programmers. I don't even know how this is possible.
Jordan Wilson [00:27:02]:
This model is so so incredibly good. The model is also, you know, gold medal level performance on the written sections of the twenty twenty five, international physics, chemistry, and math Olympiads, and scored a 50.5% on the advanced CMT benchmark for theoretical physics. Wow. So unlike traditional AI models, Gemini three deep think leverages increased test time compute, meaning it spends more time internally verifying solutions before responding, which significantly reduces the risk of technical errors or hallucinations. So, you know what's very weird, y'all? You know that show I was kind of, promoting saying, hey. You need to go back and listen to the 2026 AI prediction and road map series. The day that one came out, it was literally five hours later, that Google Gemini three, Gemini three deep sea, Deep Think came out. I've been calling it Deep Seek the whole time Deep Think.
Jordan Wilson [00:28:01]:
I need more water and more sleep. However, during that prediction and roadmap show FYI, I said Google at any point they want to, they can come out with the world's most powerful model because they can. They can right now out ship anyone from a sheer model capability. And then five out literally five hours later, that's exactly what they did. So, you know, if you sometimes think that, my predictions and the stuff I talk about is off the rocker, no. It's not. Alright. More bad news for Anthropic this week.
Jordan Wilson [00:28:38]:
So, the Pentagon is threatening to end its relationship with Anthropic as tensions rise over the firm's refusal to fully relax restrictions on how the military can use its models according to a report from Axios. So the Pentagon wants four major AI labs, including Anthropic, to allow military use of their tools for all lawful purposes, including weapons developments, intelligence gathering, and battlefield operations. So anthropic has reportedly refused to drop its hard limit on two of those areas, mass surveillance of Americans and fully autonomous weaponry, leading to months of strained negotiations. So a senior administrative official told Axios that the Pentagon is considering severing its partnership with Anthropic unless the company agrees to fewer restrictions. So Anthropic's contract with the Pentagon is valued at up to $200,000,000, and its clawed model was the first AI system integrated into classified military networks. So it's actually a timely and relevant news story. Well, here's why. Because tensions escalated after the military recently used the Claude model in an operation targeting Venezuela's Nicolas Maduro, raising concerns with within Anthropic about the software's role in missions involving lethal force.
Jordan Wilson [00:30:07]:
So Anthropic denies interfering with military operations or discussing the specifics of missions with the Department of Defense or industry partners, insisting it follows its own strict usage policy. OpenAI, Google, and XAI have all reportedly agreed to relax standard guardrails for the Pentagon work in unclassified settings, and at least one has accepted the, quote, unquote, all lawful purposes standard for classified use. So I've been saying this for years, y'all. AI is going to be more important than what weaponry a military has. It's going to be more important than a country's GDP. It's going to be more important than, you know, natural resources like gold and oil. Right? Whatever models a military has access to. And by military, I technically just mean a country because, you know, government, country, military, they're all kind of one, one and the same.
Jordan Wilson [00:31:18]:
But this is what I've been saying for a long time. Right? When we just had chadgbt.com and we didn't have these, you know, models that were technically capable capable of, you know, bioweapon creation, I've said all along, the country with the access to the most powerful AI models will be the country that rises to global supremacy. That's that's it. Right? It honestly has really not too much in the long run to do with, you know, what weapons or, you know, the amount of jets or nuclear capabilities. That doesn't matter very much in the long run. Right? Or in in the long run, what matters as well, what country or lab is going to be able to develop artificial general intelligence and artificial superintelligence first, and, well, what access, will the government or the country, that that lab belongs to have. Right? So yeah. Sorry to get all geopolitical on you, but I think it's important to keep that in mind.
Jordan Wilson [00:32:20]:
Alright. And our last big AI news story of the week was the biggest one, and this one broke late on Sunday afternoon. So OpenAI has hired Peter Steinberger, the Austrian developer behind the fast growing AI agent OpenClaw, in a move to strengthen OpenAI's leadership in the personal AI assisted market. So OpenAI CEO Sam Altman announced that Sunday evening that Steinberger, Steinberger is joining OpenAI to lead the next generation of personal AI agents following the viral success of his OpenClaw project. So if you don't know Open Claw, well, it has changed names a couple of times. It was Claude, I think, what, Claude bought first, and then it was Molt bought. Right? But they finally landed on Open claw, after some, some name changes, not on their own accord. And I'll get to that here in a second, but probably the important thing that everyone is talking about.
Jordan Wilson [00:33:25]:
Well, what's going to happen to OpenClaw? Well, so OpenAI said that they plan to keep OpenClaw as the open source project that it is right now, supporting its development through a dedicated foundation. So OpenClaw was, like I said, previously known as Claude bot, and then Molt bot was launched just months ago and became the technically, the most popular AI product ever. Right? At least if you look at, you know, GitHub ratings, which is what a lot of people look at in terms of open source projects. So and it gained popularity for its ability to autonomously autonomously complete tasks and make decisions for users. So, yeah, I've talked a little bit about it, on the, you know, AI news over the last, you know, month or so. I think it did make the 2026 AI prediction and roadmap, roadmap series as well. But, yeah, if you haven't used OpenClaw, it is essentially an autonomous semi autonomous depending on how you set it up, AI agent that has, memory. You can give it access to really anything.
Jordan Wilson [00:34:35]:
But the the big thing is, well, you can communicate, with it via, text message, Telegram, Slack. People, you know, hook it up to 11 and, you know, call it on the phone. So it is an extremely impressive, project. And like I said, one of the most successful AI launches ever, but here's where it gets really juicy y'all. This is actually. More adding even more to anthropics bad week that's because the original, the original version of this, right? I said that it went through a couple of name changes. The original was Claude Bott, so not c l a u d e, like anthropics Claude, but c l a w d. So it was launched under that name in November 2025, which was kind of a play on Anthropic's Claude chatbot, obviously, at the time.
Jordan Wilson [00:35:35]:
And it was, at at that point being run on Anthropic's model. So it was actually a great thing for Anthropic because people were spending a ton of money, in the Anthropic API, and it was bringing a lot of new developers onto the platform. And that's also, coincidentally or not, I don't know. But, you know, Claude and Claude Code really exploded in 2025. And I'm sure at least, Claude bought had a little bit to do with that. But Infropic, instead of kind of seizing the momentum and running with it, well, reportedly, they sent Peter Steinberg, Peter Steinberger, essentially a, letter from legal saying you gotta change the name. So interestingly enough, right, especially when, anthropic has always been kind of the thought of as the developer friendly, option out of everyone, not so friendly, forcing one of the most popular AI projects of all time that is sending them money to change their name. I don't know.
Jordan Wilson [00:36:41]:
Me says not very smart. And now here you have. Right? Peter did go on a bunch of, you know, different podcasts and things like that over the past week or so. And, essentially, he said at that point, even before the news broke, you know, late Sunday night that he had heard from, you know, Meta and OpenAI and had some pretty big, you know, acquisition opportunities. And then we find out it's actually OpenAI that swoops in and not only gets this acquisition. Right? So it is kind of more of an acquihire, but they are still, technically through a foundation, going to, I guess, quote, unquote acquire OpenClaw, right, in its, hundreds of thousands of users who are using this platform. I'm guessing it's probably getting near the millions now. It's it's it's hard to track that because you can look at, like, the number of installs, but I'm sure there's, you know, certain people that are installing it, you know, dozens or hundreds of times.
Jordan Wilson [00:37:41]:
Right? More of the the power users. But it's just a very, number one, great play, I think, by OpenAI. Right? There was reports, like I said, that Meta made a pretty big play, to acquire Steinberger, you know, kind of acquihire the company as did OpenAI, but OpenAI was ultimately successful. But here we go. OpenAI then gets to make a huge play to developers, right, being the good guy here. Not only that, but they've been absolutely crushing it with their codex platform. I literally I kid you not. I have codex running right now, and most of the time when I'm talking or doing anything, I have codecs the new codecs app running.
Jordan Wilson [00:38:25]:
But not only that. Right? But they just get to swoop in and now take all of any of that momentum that Anthropic would have had. And now Infropic walks away from this, not only being the loser here and fumbling the bag, but also their reputation with developers just took a huge bruise. And I think that, you know, OpenAI between their new codex app, and their new codex models. Yes. That's models with an s. I mean, they've really shifted, the story when it comes to AI development, AI coding, and what people should be using for software moving forward. Like, if you would have told me in, you know, November, December, that the tide would have shifted, I would have said, okay.
Jordan Wilson [00:39:11]:
It'll probably take a year for the tide to shift. But, I mean, anthropic just I mean, they just slapped themselves in the face. They fumbled this, Like, I don't know, like, what was what Super Bowl was it when, you know, that I've, I think like a Dallas player is like Dallas versus the bills. Right? Someone at the one yard line when they're about to score the touchdown, fumbled the ball. I mean, that was this Anthropic just, I don't know, blew probably in the long run. I would assume hundreds of billions of dollars of potential revenue, through this deal. I mean, we'll see. I think that's, you know, the, the extreme end of this.
Jordan Wilson [00:39:46]:
But this thing, this open claw is just a meteoric, and it's not slowing down. Right? And you're like, oh, open source project, not bringing money. It's bringing users, and it is bringing, via on the API. So we'll see, and we'll see if Improvig is like, okay. Open claw, you can no longer use our API. Yeah. We'll see how that works, like like, especially when, you you know, it's under this new OpenAI Foundation. But hey.
Jordan Wilson [00:40:13]:
OpenAI making a play on Open here, bringing in now. They have the, world's most popular, Open Source AI. Well, OpenAI has it now, and it's they're keeping it open. So pretty pretty impressive there. Right? OpenAI was getting a lot of flack like a year or two ago for not being very open, and now they have OpenClaw, and then they obviously have, some of their very, popular, GPT OSS open source models that they release. So, there you go. Alright. That's it for the big news stories.
Jordan Wilson [00:40:46]:
Now let's quickly go over the what's new and what's next. So some leaks, some other stories that were kinda big, but not big enough to make our top list. So here we go. Bullet point style. What's new? What's next? This is actually a big one. Google and Microsoft launched WebMCP, which lets websites expose browser tools so AI agents can act reliably. So, essentially, this is MCP for websites that allows, you know, agents, and just AI models to better read and understand websites. Google is another big one.
Jordan Wilson [00:41:21]:
Google added v o three to directly to Google Ads. So, yeah, a lot of ads that you're gonna see, they're gonna be AI. So six XAI cofounders left, after the SpaceX merger, citing internal tensions, financial disputes, and regulatory issues. Manus, which was recently acquired by Meta, quietly rolled out and always on agent functionality similar to OpenClaw. Chad GPT deep research had a face lift and an upgrade to GPT five two. It is really, really good, and you can add app integrations and targeted searches in there as well. Anthropic finished a raise on $30,000,000,000 reaching a $380,000,000,000 valuation. Hollywood is demanding that ByteDance stop their new AI model, SeedDance two point o, for alleged copyright violations.
Jordan Wilson [00:42:13]:
I say alleged loosely because it looks like straight up copyright copy and paste, but it looks so good. Chad GPT added, gen a dot mil for 3,000,000 in the department of defense. That's essentially their, not an easy name to say, genai.mil, but that's their, you know, chatbot for the military, but they added that for 3,000,000 users. XAI is working on parallel agents that could run up to eight agents at once. Runway raised $315,000,000 at a $5,300,000,000 valuation. Another big raise here, Databricks raised 5,000,000,000 at a $134,000,000,000 valuation. Claude Cowork arrived on Windows, but there were a lot of security, concerns that surfaced right away. OpenAI began testing ads, for US on, free accounts in on Go.
Jordan Wilson [00:43:07]:
They're lower, kind of lower tier paid account. So ads are here, y'all. The Pentagon fast tracks some AI deals to deploy AI on classified military networks, kinda referenced that earlier. OpenAI updated g b t five two instant. So, to deliver clearer, more direct chat g b t and API responses. That's actually big because that is the default model. If you don't choose something else, it's g b t five two instant. So under the radar, you know, roughly, like, 750,000,000 people are now using a different model, and you probably don't even know.
Jordan Wilson [00:43:41]:
So you should pay attention. OpenAI shut down g b d four o. Oh, no. It's gone. Said, I don't know. People who are on the keep four o train, I don't understand it. It's gone. Good.
Jordan Wilson [00:43:52]:
Sick of the sea, be gone. The FTC has intensified its Microsoft probe over potential AI and cloud monopolies. ChatGPT is testing skill imports allowing saved and reusable prompts. Kimmy launched Kimmy Claude, their native OpenClaw integration. Yeah. A lot of OpenClaw and OpenClaw clones hitting. A report came out and said that Spotify's developers stopped coding by hand completely, and they're even just shipping live updates from their phones. Chris Liddell, the former CFO at Microsoft, joined Anthropic's board of directors.
Jordan Wilson [00:44:29]:
OpenAI launched an update to their codex model with codex spark, a lighter, faster coding miles that, model that they partner with, Sarabras for. Google is testing notebook LM infographic customization with auto mode and nine new styles and stitch by Google now can export editable designs to Figma. Whew. That's a ton also. By the way, Stitch, I've been loving Stitch. I don't know if anyone else is using it. If you have it, you should probably go check it out. Alright.
Jordan Wilson [00:45:03]:
That's it y'all. That is the AI news that matters a ton. And if you miss anything, don't worry. It's all gonna be on our newsletter. But if you find yourself overwhelmed on a day to day, week to week basis, trying to keep up with what's happening in AI and if it matters for you or not, well, I just did all that for you. Right? This is what I do every single day. Right? I keep up, with AI. I talk to the smartest people in a, in AI, help enterprise companies, right, onboard, you know, with front end large language models.
Jordan Wilson [00:45:38]:
So this is what I do. So don't worry. Don't stress out. Just join us on Mondays. Well, every day if you can. But on Mondays, I cut it too straight. No BS. No corporate spin.
Jordan Wilson [00:45:49]:
I tell you, here's what matters. Here's what you should be paying attention to if you're a business leader. So thank you for tuning in. If this was helpful, tell someone about it. If you're listening on the podcast, please subscribe, and then make sure go check out that episode seven twelve and seven thirteen, our 2026 AI prediction and road map series. Trust me. You go listen to that, and you are already in the top 1% of AI people at your company. I guarantee it.
Jordan Wilson [00:46:17]:
So thank you for tuning in. I hope to see you back tomorrow and every day for more everyday AI. Thanks, y'all. And
Midroll [00:46:43]:
you next time.