Jay Cuthrell covers agentic AI systems and reinforcement learning as alternatives to larger language models, with warnings about AI agents deleting production data. He also touches on brain interface breakthroughs at Motif Neurotech, NASA's X-59 propulsion work, and Apple leadership news.
Chapters
Welcome to What the Fudge for May 3rd, 2026. This week I am looking at a couple of different threads around, of course, all things AI, but how that ties into things like spanning your current business into utilizing or taking advantage of AI, health innovations that are related to this, and some of the shifts that are taking place in the companies that you see in the news. So the first thing is this concept around the AI agent. There's a kind of an evolution revolution story, you'll see that comes out in some of the stories that find their way into the blogosphere or into newsletters, which I subscribe to a lot of those newsletters. And so you'll see some trends. One really that sticks out is that funding isn't just taking place in the United States. So there was a really big story around the fundraising that was done by Ineffable Intelligence. That's a startup now that has, I think, amassed somewhere around over, well, well over a billion in seed. And that's a seed round. So that's pretty impressive. But what's really interesting is that it's not in the United States, it's actually in the EU. And so they are taking a counter thesis, if you will, that it's not necessarily a large language model that's going to be the really big breakout. So other AI companies have been touting large language models for quite some time. And the ineffable intelligence takes a different approach in talking about this concept of taking experience through simulation and reinforcement, as opposed to just trying to make ever more larger of the LLMs. And so with that seed, obviously, seeing them ship something that becomes utilized, put into a case study, will be one of the tests of how real their particular thesis is. That this niche, niche world of the emerging super learner model where it's again going through simulations, looking at reinforcements. And so while we're, while we're kind of taking that on, seeing that actually show up as part of an AI agent that, as you would, as you would imagine, meaning that as the agent runs and, you know, iterates, it would improve. Some kind of a reinforcement learning would have to be in place. So when you look at the story, it really is interesting. If you know, I work for NexusTek. NexusTek is an MSP. It's also an MSSP, meaning we're a managed service provider, managed security service provider, but we're also on the path to becoming a managed intelligence provider. And so one of the vendors that you often see referenced in the MSP world is Kaseya. Now Kaseya is many different companies rolled up into what is currently Kaseya through a series of mergers and acquisitions. But Kaseya, I think about a week, a week ago, there was probably hints of this that were part of the roadmap that would be communicated by Kaseya's product team, but they've effectively taken well over a billion help desk tickets, which represents, you know, potentially millions of different endpoints that have been managed by Kaseya-licensed you know, MSPs. And so the idea is, is that what better training corpus to work from? And the case they are making is that Agentic AI will be delivered as an integrated experience, not a bolt-on experience. And so if you were to say be an MSP and you used Kaseya tools, the Kaseya tools should be markedly improved through the use of Agentic AI capabilities. Shipped through those product experiences, again built in, not as a bolt-on. And so it will be interesting to see how that plays out over time. There are certainly other players in the MSP space, for example, the Tomma Bravo company, ConnectWise. There's also very emergent companies like NinjaOne. And so it'll be interesting to see how this story of, well, we're building an AI versus we are bolting on AI versus We were born AI. So I think it isn't necessarily about the incumbents as much as it will be about the most recent entries into market and seeing how those legacy-free players start to challenge the incumbency of some of the tools that exist in the MSP world, many of which those tools are actually built by MSPs because people get frustrated and they often say there ought to be a better way and they build something. And so now the question will be is how does that match with this, uh, current, uh, social media thread. I've seen it on Blue Sky, I'm sure it was on X, I've seen it, uh, brought up into the Fediverse. And it's this notion of somehow someone turned loose a, uh, an Anthropic, uh, Claude code, uh, agent and it deleted the database or it deleted the production, you know, data. And so just this whole, just whole, uh, the whole thing just brings up this concern of Well, you did have a backup, right? And, and, and we agree on what the definition of a backup is, right? And you didn't just hand it the only copy of data that you had. And so I don't know if this is true or not. I haven't done the research to go digging into it. It could be, um, content farming, rage bait. There's a whole host of names for this. But what happens is, is that the more salacious the story, the more clicks it receives. And depending on the platform, that could translate into conversion and potentially interaction dollars, if you will. So maybe there's a little bit of rage bait in there, but there's probably some kernel of truth. And so just to round this out, because I know we're coming up on the, uh, now the 5-minute mark, and I'm gonna try to keep these under 8 minutes, I, I think it's also important to understand that there's going to be carryover from all these AI stories into the world of life sciences, healthcare and life sciences. And so If you did not watch the news, there was an update around things like Motif Neurotech. They have, I believe, received FDA approval based on what I've been reading around their clinical trials, which is, yes, in fact, a computer brain interface or a wireless brain implant in this instance. And so the whole idea is that this would be sitting well above the brain tissue as opposed to being inside or inserted into brain tissue. So this, if you're familiar with the concept of these sci-fi movies, not all of them dystopia, but this concept of you would wear a helmet, the helmet would basically have a series of, or an array of superconducting quantum interference devices, or SQUID. You'd have the SQUID helmet on, and so the SQUID helmet would somehow help program or pull in or read/write or otherwise involve itself with your brain activity. So thinking of that in terms of those stories as well as this concept around brain implants, just generally speaking, is fascinating. I would also want to update on just everything going on in the world of space. I think NASA has— I think it's the X-59 experimental aircraft story is ongoing. This could be, if you remember, there was a time when the Concorde, you know, existed for a very, very interesting supersonic flight over land and sea. And so what'll be interesting to see is how we translate from experimental aircraft to commercialized versions of that and what it might mean if you want to send a human or humans from one place to another place on the planet. And then ultimately, different kinds of new thruster technologies on if and when we would send people to the moon, the Mars missions that may happen. Those are all part of the Jet Propulsion Lab side of NASA. So I think that's a really Interesting story considering there's so many other companies that are being covered in the space value chain. So if you go to svc.fudge.org, NASA is still doing good work. NASA has not been shut down. You got to be really careful of what you read online, what you hear in the news. NASA is absolutely still viable. And so I think another interesting one just to kind of round out would be is we clearly know that if you, if you're listening to this, it's a good chance that it's probably on an Apple device. Based on what little information I know about who hits my website. But if you are familiar with Apple as the company, yes, in fact, the person that was the CEO is no longer going to be the CEO. So Tim Cook is going to leave. And there's some really great writing from MG Siegler. I would recommend you go to Spyglass.org, check out his coverage of Ternus, the new name that's going to be replacing Tim Cook. I think it would kind of swing us back to— there's a, there's a great story that MG used If I was a writer in the olden days that would use music and lyrics and titles as inspiration for mine, MG is just an absolute movie, movie person, and he actually makes a really interesting movie analogy. I won't give you a spoiler. It's worth reading, but go to splagglass.org, read about the Tim Cook, Turnus, and Apple story that MG captured. That's enough for today, I think. I'm coming in just over the 8-minute mark, just before the 9, so thank you for reading and listening. To the What the Fudge Digest for May 3rd, 2026.
