Episode Transcript
[00:00:00] This is the AI news you need to know for the week of March 15, 2026. I'm Craig Hewitt. Let's start with an overview of just the models that have been released in the last couple of weeks. We have from OpenAI GPT 5.4 from the anthropic team, Claude Opus 4.6, Gemini 3.1 Pro Llama 4 Maverick and Grok 420. I think the thing that gets me here is just the race and the competitiveness between the big players.
[00:00:31] Anthropic, OpenAI, Google and Gemini and Xai and Grok. Just the competitive nature of this is great for everyone because these frontier models are advancing so quickly and becoming so much better. It really benefits everyone. I think there is some question about defense and military and spending which we'll get into in a minute. But just so we know these companies are moving forward big time and the rate of change I think is just increasing.
[00:00:58] And speaking of the frontier model companies, Anthropic released code review right inside Claude Code this week will cost between 25 and $50 available on team and enterprise plans. And the big takeaway here for me is if you're relying on tools like Code Rabbit, Cubic or Baz or any of those, I think this is just another example of where the frontier model companies are just going to eat these wrapper companies. And I put that in italics because thus far there's been a lot of really good value from us creating businesses out of AI wrappers and I think there could be in the future. But the risk here of course is Anthropic just releases this or what they did with their legal, you know, product inside Claude Cowork a few weeks ago or their security product inside Claude cowork.
[00:01:46] It's just going to decimate thin moat companies. And I think the honest truth is these code review apps are pretty thin and, and so the Anthropic team releasing this code review tool for 25 bucks per use. If you compare that to a fully baked US based developer at 150 to 200 grand a year, it's a huge steal.
[00:02:07] Sticking with the frontier model companies, Google released a public CLI so command line interface for Google Workspace. So now your agents, your open clause can integrate with an official command line interface to do things like mail, calendar, docs, Google Drive, all these kinds of things all officially. So this is a huge step forward for me in terms of how I work with my openclaw. Her name's Janet. Because we do all of our work in Google Workspace and Google Docs. Previously it was kind of a homespun cli tool by Peter Steinberger, which worked really well. But now you can just imagine, hey, there's going to be a lot more infrastructure that's agentic around the Google Drive ecosystem. Google, of course, is my kind of dark horse to win the AI race long term because they already have the embedded nature in how most people work in Google Workspace.
[00:03:00] Okay, next is a big release from a second kind of tier company I think, which is Replit. So replit their agent 4. This is to me the easy way for you to get into advanced agentic work with AI. So if you're using ChatGPT or Claude right now, this might be a really great way to say like, hey, I want to build my first agent or hey, I want to go build this little web app for myself internally or I want to build a little thing for my kids soccer team where ChatGPT and regular Claude might not do this. Replit a really cool way to do this.
[00:03:33] A couple of the highlights from Replit4 is they have this concept of like a design canvas so you can click on elements within the canvas and comment on them and move them around. And they also have parallelization so you can be working on front end, back end infrastructure design all at once and it's not going to kind of confound or have merge conflicts or anything like that.
[00:03:53] Relet and their Agent four also does things like slides and animation and media, which is a big place to look forward to, of course. And you know, my business Kastos in the podcasting space, always looking at, hey, how is Generative AI going to affect us into the future? I think the answer is we're early, but something to pay attention to. And speaking of Castos, this podcast is hosted and sponsored by Castos. If you're looking to get started with a podcast here in 2026 to increase kind of trust and thought leadership for yourself and your brand, check out kastos.com, we'll have a link in the description below and a coupon code for you get a discount off your first few months diving into politics. You know, we had a statement from Dario Amadei, CEO of Anthropic, a couple of weeks ago and and really this hasn't simmered down much since then. So Anthropic classified as kind of a key supply chain risk to the United States military and they call it Department of War, you could call it Department of Defense in the Pentagon they were ostensibly booted. Sense has filed an injunction or a lawsuit against the Defense Department and the Pentagon. OpenAI came in and kind of became the vendor of choice. And the thing I look at here is just like, hey, how is this going to go from a commercial enterprise and these AI tools to becoming a part of the state, you know, and like core absolutely necessary infrastructure for governments to operate. And I think that's where these go from really nice to really powerful to absolutely essential companies. And how does that change the commercial viability of them? How does that change how we use them? You know, there's the QUIT GPT movement going on. You know, how does all this politicizing of AI and AI tooling affect us as users? I don't know, but I just think it's something that we should be aware of. Okay, next one is Perplexity Computer and this one's really interesting. You know, this one really mimics Manus to me and how you might think about using like Claude Cowork on your computer. And so Perplexity in their computer product, really just like giving Perplexity as a model router, really arms. And so you can see here like it's using Cloudsonnet 4.6 and it can like create documents, you know, all sorts of like predefined templates here. It can create documents, slideshows, charts, calculators, tools, all those kinds of things. So, like, a lot of us would think about Perplexity as a good search tool. And that's how I use it. I actually use Perplexity every day just for the news, because I find it's the best aggregator of news for me.
[00:06:27] But increasingly I think we're getting this convergence of all these tools into doing about the same things. And so the next one here is payments and how AI agents are going to use payments going forward. And so we had news from both Stripe and Ramp this week on frameworks and methodologies for AI agents to use frameworks both with conventional payments. So Visa and MasterCard, American Express and crypto. So I think the big thing to pay attention to here is like, hey, agents have to be able to pay for things and transact either with other agents or with conventional, you know, fiat money in kind of old school commerce. How are they going to do that? My guess here is this is going to be the use case for crypto that we've all been looking for, right? So crypto has long been a solution without a problem. I think we just found the problem. So this is something really to pay attention to. Of course, Stripe is going to be the leader Here just because they're already embedded in the payment space so much.
[00:07:24] When they roll out and make this, this kind of ecosystem more robust for agenda commerce, I think they'll be the leader along with Coinbase. I expect them to do some things together. Todd Saunders had an interesting post the other day about Stripe. Now monitoring LLM calls that are being made to their platform and being able to kind of suggest and route model weighting and types for the kind of call that you're making. I think this is really interesting. This is getting into kind of the open router space to me. Not sure if Stripe is going to veer into being an open router type of product. I think they'll stick in payments, but I think the concept here is interesting. Okay, and lastly, some maybe not good news, but just something I think we all need to be aware of as business people is layoffs and unemployment. You know, if you look at the unemployment rate here over the last, what is this, six or seventy years, it's, it's gone up and down, right? And as a baseline, like, I think we're actually pretty good here, kind of four and a half percent right now.
[00:08:23] The thing that I think about here is I think this is vastly underreported, especially in the hiring aspect. Like, I think we hear the news of layoffs all the time. Oh, you know, Atlassian laying all these people off, Amazon laying all these people off. You know, these big huge companies laying tons and tons, like tens of thousands of people off. That's absolutely going to affect the economy, especially in that a lot of these folks are getting laid off are high earners. So operations folks, developers getting laid off and not getting rehired. I think that's the real thing. So it's not that the reported unemployment rate is high, it's that there's not new jobs being created. And I think that's the thing we really have to look out for as we're looking at existential threats to our business. I don't think that the risk to our business is AI and someone vibe coding our product. I think it's just that there's going to be, I think 20% unemployment. And that's historically not like unprecedented. Right? We saw 15% during COVID We've seen 10% a few times in the last 30 years.
[00:09:25] I think that AI is such a fundamental change to economy and commerce in general that we could see that. And if we do see that, then the government has to step in and do something like universal basic or universal high employment and give us all money just for kind of existing, because there's going to be so much surplus and abundance in the world. And I do think that's where we'll get to. I do think there'll be this kind of painful period between now and when that happens. And that'll probably last, say, two to five years from right now. So this is a new podcast. This is a new format. I hope you're enjoying this. If you are, please subscribe and share with someone who you think would enjoy it as well.
