Critter Stack Update with Jeremy Miller

0:01Speaker 1:How'd you like to listen to. NET Rocks with no ads? Easy. Become a patron. For just $ 5 a month, you get access to a private RSS feed where all the shows have no ads. $ 20 a month will get you that and a special. NET Rocks patron mug. Sign up now at patreon.dotnetrocks.com. Hey, welcome back to.

0:36Speaker 2:NET Rocks.

0:36Speaker 1:I'm Carl Franklin.

0:37Speaker 2:And I'm Richard Campbell.

0:38Speaker 1:This is episode 2023, my friend. It's almost over.

0:44Speaker 2:This is it. It's almost over. I'm looking forward to it. I've got a special show planned for 2027 now. I saw.

0:49Speaker 1:I came up with an idea.

0:51Speaker 2:Yeah, you like that? Let's not tell anybody. Let's just find out.

0:54Speaker 1:No, we're not going to say anything about it.

0:56Speaker 2:We'll just see.

0:57Speaker 3:Yeah.

0:57Speaker 2:All right.

0:58Speaker 1:We're going to look back at 2027.

1:00Speaker 2:Something like that. I don't know.

1:01Speaker 1:Okay. Well, what we're talking about, of course, is we have been doing this historical documentation.

1:11Speaker 2:It started casually in the 1900s, but it's gotten a little intense now.

1:15Speaker 1:Yeah. And it's also kind of a yawner now because we're in the 2020s. Yeah.

1:21Speaker 2:It was yesterday.

1:22Speaker 3:Yeah.

1:23Speaker 1:So, what happened three years ago? Let's talk about that.

1:26Speaker 2:This is 2023. Okay.

1:28Speaker 1:Um, I'll tell you the bad news. Richard will probably say some happier news. Um, Hamas attacks, Israel, Israel, Gaza war begins.

1:36Speaker 3:Yeah.

1:36Speaker 2:Bad war.

1:37Speaker 1:And, uh, good wars, but still, no, uh, Russia, Ukraine war enters its second year. And, uh, yeah, well, um, Western, that's what you got, huh? Well, the Western military support the Wagner group, uh, mutiny.

1:52Speaker 2:Oh yeah. In June.

1:52Speaker 1:And the death of Wagner, Vegni, uh, Frigosian?

1:57Speaker 2:Yeah.

1:58Speaker 4:Richard, that happened on the last day of KCDC that year.

2:02Speaker 2:Yeah. We were all there. Wow. You never heard from him again because those guys are all dead now.

2:12Speaker 1:Dead men tell no tales.

2:14Speaker 2:Yeah.

2:15Speaker 1:Donald Trump was indicted four times, charged in four separate criminal cases involving the 2020 election, classified documents. The January 6th aftermath and the Stormy Daniels hush money matter.

2:26Speaker 2:Remember that?

2:28Speaker 4:Yeah.

2:28Speaker 2:Remember that stuff? Yeah. Remember that stuff.

2:31Speaker 1:Okay. Some devastating earthquakes, Turkey, Syria in that area. Killed more than 50,000 people. So bad.

2:42Speaker 2:Well, and you're talking Syria is still in the midst of a civil war, too. So, who knows what help they got or any of those things.

2:48Speaker 5:Right.

2:49Speaker 1:Uh, 2023 becomes the hottest year on record. I think that has been surpassed now, but, uh, that was the record to that point. Um, bad, bad, bad, bad. Uh, Sudan descends.

3:02Speaker 2:Into civil war. Yeah. It's inevitable.

3:05Speaker 1:The Sudanese armed forces versus the rapid support forces. Thousands were killed. Millions displaced. One of the world's largest humanitarian crises. Uh, the U S banking crisis. Silicon Valley Bank and others collapse. Signature Bank.

3:21Speaker 2:Yeah, that one, I mean, not the same scope as 2008, but it didn't have a contagion. Like one would argue that the contagion rules they put in place in 2009 actually worked. Right.

3:34Speaker 1:Hollywood and U.S. auto workers stage a major strikes. Writers and actors.

3:39Speaker 2:About AI.

3:40Speaker 1:Yeah, brought much of Hollywood production to a halt while the UAW simultaneously struck Ford, GM, and Stellantis. The year became one of the most significant periods for organized labor in the United States in decades. Prince Charles became King Charles.

3:54Speaker 2:Right.

3:54Speaker 1:King Charles III, May 6th.

3:57Speaker 4:My oldest son got to do his freshman year of college in London.

4:01Speaker 2:Oh, wow. Saw the coronation.

4:04Speaker 4:His first four or five weeks, two monarchs, three prime ministers.

4:08Speaker 2:Yeah. They're still on the prime minister role here. What is it, 10 in four years? It's amazing.

4:16Speaker 4:He just got there when Liz Truss was trying to outrace the lettuce and failed.

4:22Speaker 2:Yeah, and failed. Cannot last longer than lettuce.

4:28Speaker 1:Let's see. The Titan submersible disaster, the disappearance of Ocean Gates Titan during a dive to the wreck of the Titanic triggered a massive multinational search and became an enormous worldwide news story. You probably know more about that than I do, Richard. Don't use carbon fiber for submersibles.

4:47Speaker 2:That's the end of that. Yeah.

4:48Speaker 1:Catastrophic implosion.

4:50Speaker 2:It imploded. Yeah, not exploded. Five people died. I mean- They'd ever felt a thing. It would have been instantaneous. So incredibly fast, but still.

5:00Speaker 1:Chinese spy balloon incident?

5:01Speaker 2:Yeah, that's a weird one, isn't it? It was a weird one. F-22 shoots down a spy balloon? Yeah.

5:08Speaker 1:Okay. The Maui wildfires. Bad, bad, bad.

5:12Speaker 2:Yeah. Terrible.

5:13Speaker 1:There was an earthquake in Morocco, a catastrophic flood in Libya, or floods, more than one. Kevin McCarthy's removal as House Speaker. All sorts of And it's gone well since then.

5:25Speaker 2:Yeah.

5:27Speaker 1:Okay, we got to laugh to keep from crying. So, Richard, let's talk about space.

5:33Speaker 2:I would also suggest that this was the end of COVID, right? In January, when China finally reopened. Right. But also announced that its population has fallen. You know, they start to understand that the way they were doing their assessment of population wasn't effective. And there was like 100 million fewer people than they thought.

5:50Speaker 4:Wow.

5:51Speaker 2:And then in May, when the World Health Organization officially ended the global emergency.

5:55Speaker 3:Yeah.

5:55Speaker 2:End of COVID. Yeah. As a pandemic anyway. All right, into space. Yes. 2023 is really the year that SpaceX begins to dominate with 98 of 223 launches in the world. So that's 44% comes from one company. So that's 91 Falcon 9 launches, five Falcon Heavy launches, and the first two test flights of starship uh the previous year in 22 it was 61 launches they had 99 booster landings which seems weird they only had 98 launches but 99 booster landings remember that every falcon heavy has two boosters for a total of 1200 metric tons lifted to orbit um wow again a crazy record Now, 63 of those 91 Falcon 9 launches were Starlink because Starlink is really coming into play here, especially in the context of the Ukraine war. This is the year that the Elon biography comes out where we find out that at the beginning of the war, he actually– well, not the beginning. It was like in November of 22, he actually turned off Starlink at the request of Putin. That was more than a little bit of a kerfuffle overall.

7:07Speaker 1:Kerfuffle.

7:07Speaker 5:Yeah.

7:08Speaker 2:Yeah. Interesting times. I would also have to mention China, who had 67 launches that year compared to SpaceX's 98. But that is them operating their own space station. Three supply missions and a few crew changes in 23. In January, SpaceX does a Transporter 6 mission. I would normally not mention these. Transporter 6 is literally their flying CubeSats. I think they flew like 50 CubeSats on it. But one of them was Caltech's Space Solar Power Demonstrator. This is a little 50-kilogram CubeSat that actually has deployable solar collectors on it that had tested 32 different styles of solar collectors, different types of full of voltaics and so forth, and included a microwave transmitter and successfully beamed power to Earth.

7:56Speaker 2:It wasn't a meaningful amount of power, but it was an end-to-end test, essentially, that you can collect solar power in orbit and beam it onto the surface. It was microwaved, right? microwaves. Yeah. Microwave microwave is like nine gigahertz.

8:08Speaker 4:Was that actually successful? That's been a lot of sci-fi books.

8:12Speaker 2:Yeah. Well, it's, that's the old, you know, back from the seventies, uh, you know, Gerald O'Neill and so forth. So this was a genuine end to end test worked a hundred percent.

8:20Speaker 1:Like, uh, yeah, we were talking about that in the nineties that sure that this was a good.

8:24Speaker 2:It goes back to the seventies. It's gone a long, long time. But my, um, We did the data centers in space talk, and I sort of end that one with, hey, if we can really lift 100 metric tons to orbit with a starship at a price of less than $ 100 million, maybe space-based power is actually feasible today. Because now we talk about the most advanced theoretical designs, the SPS Alpha designs from the British Aeronautical Society. They're talking about 3,000 metric tons for a gigawatt. And that's outrageous if you're lifting them in 20-ton chunks. But if you're lifting them in 100-ton chunks... It starts to look practical. And speaking of that 100-ton lift, in April, Starship test flight number one, the most powerful rocket ever launched. This was the V-1. Today, we're dealing with V-3s. It didn't go particularly well, but it was a full end-to-end stack. They launched. The separation went a little bit wonky. Everything was terminated after about four minutes worth of flight. But this gigantic 9-meter wide, 120-meter tall stainless steel rocket took off. More thrust than Apollo by Saturn V by a lot. And yeah, they're

9:35Speaker 2:on their path. It's advancing to this day.

9:38Speaker 1:Was that... That wasn't the one that sort of blew up on the launch pad and toasted everything around.

9:44Speaker 2:It, right?

9:44Speaker 4:No.

9:45Speaker 2:Well, actually, what it was was it lost three engines at launch. And so, it lifted off very, very slowly, literally as it burned fuel lost weight. So it dug a 20-foot deep hole. Blowing concrete everywhere.

10:03Speaker 4:Yeah.

10:03Speaker 2:It destroyed, it really seriously damaged the pad. Destroyed the pad, yeah. Arguably, when they had three engines fail, they should not have attempted to launch it. It should have cut off immediately, but they didn't. And so, they also, this is the milk stool design. So, they just had legs to, they had no deflectors. They had no water suppression system. None of the normal things you would do with a rocket like this. They did not do that again. In the process of repairing the pad, they did put in a water suppression system. And subsequently, today, they have a proper full flame trench with water suppression redirecting the exhaust.

10:37Speaker 1:What were they thinking?

10:38Speaker 2:This massive rocket. He wanted to make it as simple as possible for flying from Mars. That was what he was thinking, which is dumb, dumb, dumb, dumb. Anyway, the good news is they learned. In August– The Indians, the Indian space program, Chang'e-N-3 successfully does a soft landing of their lander near the moon's south pole. This will be the only fourth country to ever land on the moon. The first to be in the Americans will actually be. The Soviet Union, then the Americans, and then China, but now India was up there as well, and also the first to land anywhere near the South Pole. That same month, Russia attempts to return to the moon with their Luna 25.

11:18Speaker 2:It crashes. In September, OSIRIS-REx, this is a US mission to collect samples of the asteroid Bennu, returns those samples successfully to Earth. And in October, Falcon Heavy launches the Psyche mission. This is a mission to the largest object in the asteroid belt. Psyche is a metallic asteroid, possibly the core of a planet that was ripped apart by the tidal forces caused by Jupiter. And they want to seriously map that. That mission is still underway. It's still flying.

11:51Speaker 2:Second test flight of Starship in November. And that's all I got for that. So it's not a big, I mean, there's a lot of missions, but they're mostly Starlink missions, right? Shall we talk about computing? Because it's 2023. Yeah. And this is when AI takes over the conversation. Yeah.

12:06Speaker 1:Right.

12:07Speaker 2:We had ChatGPT launched in November of 22. It is now the beginning of 23. This is the first time Sachin Adela ever writes an all-hands letter the way Bill Gates did, where he says, look, we've got this agreement with OpenAI. We've invested an additional $ 10 billion in them, bringing their total investment to $ 13 billion. And every team should take a look at these APIs and start immigrating them into their applications and make a co-pilot. I've heard rumor there were literally hundreds. But, you know, it's a first for Satya to direct the whole company in a new direction. And really the thing that I think he thinks he can hang his hat on as CEO, that this is what he brings to the company. First out of the gate in February will be Bing, formerly known as Project Sydney, where they were trying to put a natural interface onto the search engine. I hear both guys who tried it really liked it. I said that a lot.

13:00Speaker 2:I feel, I feel bad. I mean, yeah. You know, it turns out that just putting an LLM onto your search engine doesn't make it a popular search engine, but you know, it was first out the gate.

13:09Speaker 3:Right.

13:09Speaker 2:In March we get GPT four. So we go from what? 275 billion parameters up to a trillion parameters. Uh, It's amazing it happened that fast, you know, that less than a year after GPT-3, they had GPT-4. This is when the 10 billion bought them, was to build an incredibly massive model. And it's impressive, the step forward on that. This, of course, freaks out Google, who had been playing with BARD and had Google Brain and all those sorts of things. They're actually, internally, they called it a code red event.

13:38Speaker 2:This becomes the big push for Gemini, right? In May, Jeffrey Hinton said, One of the godfathers of the generative AI movement leaves Google and goes on a press tour that he's still on to this day talking about the dangers of AI.

13:52Speaker 1:He basically says, stop.

13:53Speaker 2:Yeah. Well, now that all of my stock is vested and I get to keep my millions, now let's talk about how problematic it is.

14:01Speaker 3:Yeah.

14:02Speaker 2:Yeah. In April, Meta releases an open source large language model made by Meta AI, aka Lama. One of the first open source models out of the gate. And by July, there'll be Llama 2. And it really hits pretty hard because if you're going to start thinking local models, there it is. This puts a lot of pressure on everybody else. So by September, you have Amazon out there with Bedrock and Titan. Again, I think both guys have used that one too. But, you know, all of that, you recognize that the 23, everybody has to play, right? So we've got all the hyperscalers on board now and we're in motion.

14:40Speaker 2:to try and grow something with all this technology. October, Microsoft completes the acquisition of Activision Blizzard for $ 68. 7 billion. It takes over a year, lots of battles with the FTC and so forth, and pretty much nothing has come of that. For the most part, somebody cashed in big, but goodness knows what Microsoft's strategy for gaming actually is. All right, now we get to the fun month in 2023 when it comes to tech, and it's in November. Obviously, we get. NET Conf. with. NET 8, a long-term support version of. NET, a very good one too. We're a big push with native AOT, C Sharp 12, a lot of cloud-native tooling. And the first announcement of what at that time will be known as. NET Aspire. And our friend David Fowler, now distinguished engineer David Fowler, driving this, what we now refer to, we didn't know how to talk about at the time, about the scaffolding for building cloud-native with.

15:35Speaker 3:NET.

15:35Speaker 2:Of course, it'll advance a lot more. And a week after that event, or not even a week, the weekend of that event, on Friday, November 17th, the OpenAI board puts out a press release where they say, Sam Altman will.

15:50Speaker 1:Be departing as CEO. And so began the drama.

15:53Speaker 2:And Mira Muriati, the CTO, is appointed as interim CEO. The chairman of the board for OpenAI, a guy named Greg Brockman, who's also been part of the originators on this, he's an OpenAI president. He is removed as chairman and resigns in solidarity with Altman. On Saturday, by all accounts, all hell breaks loose. A very angry Satya Nadella who did not see this coming. Nobody thought to notify the guy who kicked $ 13 billion into your company that, hey, we're going to fire the CEO. This leads to on Sunday where the board announces a different interim CEO, Twitch CEO Emmett Shearer. After only two days of Marati, I think she sort of was backing away from all of this.

16:41Speaker 2:This is the day that Satya announces that Sam Altman and Greg Brockman are going to join Microsoft and form a new AI group.

16:48Speaker 1:I remember.

16:48Speaker 2:Which leads to, on Monday, where 700 of 770 employees of OpenAI, including– One Ilya Sutskiver, who had publicly expressed a regret for his role in firing Sam Altman, also signs this letter where they say they demand the resignation of the board and will consider following Altman to Microsoft. So essentially the entire OpenAI says, we're going to go to Microsoft. This results in, on Tuesday, Sam Altman returning as CEO and the entire board being replaced by Wednesday.

17:20Speaker 1:It's like a soap opera, isn't it?

17:22Speaker 2:It's insane.

17:23Speaker 1:It really is.

17:24Speaker 2:Yeah. And then look, I think everybody knows Sam Altman's not a good man. Right. But boy, you sure screwed this up.

17:30Speaker 3:Yeah.

17:31Speaker 2:Like, there was a way to do this that might have worked. This was not that way. And fundamentally, you know, it cemented Sam Altman's control of the company in the process. Like, he ultimately benefited from this and turned OpenAI from its Original idea, which I don't know how good that idea ever was of doing AI in the open, which they very quickly diverted from into just another for-profit company that's racing to try to IPO. Little goodness knows if that's ever going to happen. Yeah. Last but not least, I think I've been mentioning quantum computing in the past few of these just because we had an interesting wave of quantum computing in the late 2010s, beginning of 2020s. I sort of see 2023 as sort of the end of the road of the race to the larger qubit at the time with IBM's Condor processor, a thousand qubits. Now we sort of have this debate about what a physical versus a logical qubit is actually because there's so much error correction that needs to happen that Under the current model at that time, they said, we're probably not going to useable work until we get into possibly millions of qubits.

18:35Speaker 2:And that'll put a lot of pressure on this error correction problem for quantum computing, which we'll see more of in subsequent years.

18:42Speaker 4:Yeah.

18:42Speaker 2:And that's what I got.

18:43Speaker 1:All right. As again, Jeremy Miller's here. Jeremy, you can jump in anytime you want, as you have been doing. But right now we're going to do better know framework.

18:54Speaker 2:Awesome. Roll the music. All right, man. What is Simon Kropp up to?

19:06Speaker 1:Yeah, I know. We should just call it Simon Kropp. Better know Simon Kropp. This is EF query complexity.

19:14Speaker 2:What is this?

19:15Speaker 1:So it detects overly complex entity framework core queries. and either logs them or throws. Each check has two levels. The level a query is logged at, which has defaults, and the level a query throws at, which is opt-in. So, yeah. So, you know.

19:34Speaker 2:So this is just a, if you've written a link expression so complicated for EF that it's going to make a bad query, this will give you a heads up?

19:42Speaker 1:Yeah, before calling it and hanging your database.

19:46Speaker 2:Yeah.

19:47Speaker 1:You might want to run it through this.

19:49Speaker 2:Yeah, no, that's true. a great idea god of course you're so clever i know you know you know he got here the hard way yeah that he wrote he wrote a link expression that pinned the database to the wall for i don't know maybe a day and he's like hmm.

20:01Speaker 4:Let's give him a little more credit. He probably inherited someone else's code base.

20:05Speaker 1:You're probably right.

20:07Speaker 4:Fair.

20:07Speaker 2:You're probably exactly right. He probably didn't write the query. He probably didn't stop this from happening again.

20:12Speaker 3:Yeah.

20:12Speaker 1:Well, anyway, that's it. That's what I got. Who is talking to us today, Richard?

20:16Speaker 2:Let's grab a comment off of show 1860, the one we did with one Jeremy Miller. Never heard of him. Talking a little bit about minimal architecture.

20:23Speaker 4:Right before the Civil War.

20:24Speaker 2:There we go.

20:25Speaker 1:Yeah.

20:27Speaker 2:Now you're in the groove, Jeremy.

20:28Speaker 1:You get it.

20:30Speaker 2:Imagine if we started this in the 1800s. It was like, I think it was like, you should go back and listen, but I think it was like in episode 1914. I'm like, hey, start a quote World War I. Right. The good news is it's almost over. A couple more shows. This comment comes from Trevor. He said, I love this discussion. Enjoyed the comments on microservices versus monoliths. I got pushed heavily into microservices approach with a product we had built. and we're re-architecting it to microservices, and it was the worst mistake ever. Things just became more complicated. It was harder to maintain. It added a lot of latency and security issues, and the complexity was just not worth it. So I came up with a new acronym, appropriately side service, or ASS.

21:13Speaker 2:I 100% believe in services, separation of concerns, and clean architectures, but the approach must be appropriate to the complexity of the solution and the size of the team. It makes no sense to have 100 separate services with a team of 10 people. And then it also makes no sense to have one massive single deployment with a code base for a team of 250 people. The services need to work with the cognitive load and be appropriate to the organization and team structures. Love any discussion. But on this one point, stop making the CTO out to be the bad guy. Love from Trevor, CTO. Maybe we were teasing on the non-technical CTO. Clearly, Trevor's not that guy. He's the one who's actually trying to get things right and be part of the team and so forth. So, sorry about teasing you. Too mean to you there, Trevor. But we're with you. Appropriate size service is correct. So, thank you so much for your comments. And CopyMuse2CodeBuy is on its way to you. And if you'd like a copy of Muse2CodeBuy, write a comment on the website at. netrocks. com or on the Facebooks. We publish every show there. And if you comment there and I read it in the show,

22:15Speaker 2:we'll send you a copy of Muse2CodeBuy.

22:16Speaker 1:And if you want to just get Music to Code By, which will help you stay in a state of flow to be calm and all that stuff while you're writing code, go to musictocodeby.net and just get it. MP3s, waves, and FLAC formats. All right. And that voice that you heard before, that third voice, was that of Jeremy Miller. He's been on. NET Rocks many times. He started his career as a quote-unquote real engineer. but wandered into software because that looked like more fun. Since then, Jeremy has worked in and led software development teams in the computer manufacturing industry, finance, insurance, healthcare, and banking industries. Lately, Jeremy has been focused on leading software architecture teams and helping mentor other software architects. Having had roles both as an in-house software architect and as a software consultant, Jeremy has a great deal of insight into the challenges that confront companies developing and maintaining enterprise systems over time. He's well known for his open source software tools, starting with StructureMap and continuing today to Martin and Wolverine. Jeremy is also a frequent author and technical

23:28Speaker 1:speaker at software conferences and can be found at jeremydmiller.com. Welcome back.

23:33Speaker 2:Hey, all.

23:34Speaker 4:Thanks, guys. Thanks for having me on again.

23:35Speaker 2:Yeah, great to have you back. Great to have you back. There was a comment on another show that I was very tempted to read where it mentioned alt.net. Just a callback to The first show we did with you, which was in 2008, was the Alt.net show.

23:50Speaker 3:Yeah.

23:51Speaker 4:It was a different day and time.

23:52Speaker 2:Certainly.

23:52Speaker 4:So if we're ready to be, let's just start with a get off my lawn kind of moment. One thing I've thought pretty recently. So after Twitter became, you know, the Nazi infested Star Wars, the Star Wars cantina theme it is now.

24:10Speaker 2:Yeah.

24:14Speaker 4:You know, the technical conversations kind of went everywhere, but it seems to be mostly on LinkedIn, or at least as far as I can tell, weirdly, right? One thing I've noticed, I think it's a gripe, and maybe it's just the algorithm. What I see are the. NET content creators. The people are the. NET content creators that are the same age that I was back in the all.

24:37Speaker 1:NET days.

24:37Speaker 4:We were big bloggers on CodeBetter. It is very narrowly focused on a handful of topics. It's very narrowly mainstream. They're not expanding the scope of software development. They aren't trying to innovate. It's here's how to use EF Core to do modular monoliths. So interesting. We were better in the all.net days than the content creators are now. Just a little bit of some trivia first. You know, you call that Simon Kropp. I do want to call him out.

25:08Speaker 4:If you've ever looked at any critter stack related documentation website, you'll see his works. We use markdown snippets. That's a really valuable way to create living technical documentation.

25:19Speaker 3:Right.

25:20Speaker 4:And then just on your theme, the history theme, you know, you were talking about 2023. We missed a couple really important things. I founded Jasper FX in June of 2023. Went solo for the first time at the tender age of 49.

25:40Speaker 2:Never too late.

25:42Speaker 4:Well, and that's the lesson I think I want. I want everybody to learn, but also do it while you're younger. And we'll bring 1.0 in 2023. Yeah.

25:49Speaker 2:Yeah. Awesome. Well, congratulations. It's been a fun few years and we've done some interesting shows around it just because, you know, we've really built out a lot of tools to make it easier for us. And if Simon likes them, That's a heck of a compliment, man.

26:05Speaker 4:I'll take it. I'll put words in his mouth. Simon's also a contributor to the Critter stack.

26:09Speaker 2:That's awesome. Yeah, I love it.

26:11Speaker 1:So new Critters?

26:13Speaker 4:Yeah. Again, I mean, AI is going to drop in here sooner or later. AI has allowed us to go a lot faster to deal with a lot of ideas that have been out there for a long time. Interesting.

26:25Speaker 2:And that's real.

26:26Speaker 4:So I'm still going to say Martin, based on Postgres, is still the most robust Critter. widely used event sourcing solution in all of.

26:38Speaker 1:NET.

26:40Speaker 4:Still going to say it's the most capable. But we've been told for over a decade that, hey, you know what? It would be much more successful if it was based on SQL Server instead. And I've always said, I don't have time. That's too hard. And SQL Server doesn't have SQL Server is way behind in JSON support.

26:59Speaker 2:Sure.

26:59Speaker 4:Now we have SQL Server 2025. It has a native JSON type. It has its own capability to do vector support. It has a lot of the same capabilities that Postgres has. So now we have a tool called Polecat.

27:11Speaker 1:Polecat.

27:11Speaker 4:That's another critter that has probably 80% to 90% of the functionality of Martin and at this point probably shares a lot of the same a lot of the same infrastructure and, um, compliance tests of behavior. So it gives you the same kind of document database capability right on top of SQL server. So you guys can stop wasting so much time with EF core mapping and, you know, to the crazy EF core link query is too complex. Using event sourcing with projections, you can project to exactly the view you want. You can sidestep silly queries, just saying, um, So we have Polecat on top of SQL Server, SQL Server 2025. And just a little more recently, we've added another library called Fisher that gives you the same functionality on top of SQLite.

28:07Speaker 1:Is that like FisherCat?

28:08Speaker 2:Yeah. There's a lot of mustelids in this conversation, let me tell you.

28:12Speaker 4:Exactly. And that's mostly the theme.

28:16Speaker 2:Yes.

28:17Speaker 1:When we first moved into our house, we were told that the woods out back were full of FisherCats.

28:23Speaker 5:Yeah.

28:23Speaker 4:Oh, that's awesome. I've never seen one in real life.

28:26Speaker 2:The dangerous little buggers. Of course.

28:28Speaker 1:That was the day my children vowed never to go for a walk in the woods.

28:32Speaker 4:So, they're supposed to look just like Martins, but bigger, right? Mm-hmm. Yeah. Cool.

28:37Speaker 2:And they're hyper-aggressive. They are.

28:39Speaker 1:Vicious.

28:40Speaker 2:We've got a mink hanging around here right now. And he scares me more than the otters that are five times his size. They'll give you. They'll go. They're looking for trouble.

28:52Speaker 5:Huh.

28:52Speaker 2:All right. So Fisher.

28:54Speaker 4:Sorry, guys. There's actually more.

28:56Speaker 2:Oh, yeah. Okay. So Fisher is SQLite.

28:59Speaker 4:SQLite.

29:00Speaker 2:So you started on Postgres. You've done SQL Server 2025 because of the JSON. type. And now you are also got it over to SQL light as well, which is cool.

29:10Speaker 4:So that is also helping us with some of our own products. We have a commercial tool called critter watch, you know, that's the inevitable management console. Um, but it's also an MCP hub that connects connected to your AI tools and it will give you complete visibility into a static view of your application and a dynamic view as well. It'll give you access to alerts. It'll even call out to LLMs to maybe be able to take proactively take amelioration act actions, replay things off, off of dead letter cues.

29:46Speaker 4:Maybe it can be smart enough to know, Hey, I need to go reboot the server, something like that potential.

29:52Speaker 2:Wow.

29:53Speaker 4:And then, Sorry to keep overloading you. We have a new tool that's not quite 1.0 called Bobcat. Bobcat? It's been hanging around long enough that it predates the Critterstack naming. I always thought they were cool, and I've got a rural background, so Bobcat means a steer skid to me. That's just kind of a thing you use to do crazy hard jobs. So Bobcat is going to be our tool for spec-driven development and integration testing. has a Gherkin function, but I mean, I don't know if anybody's going to care about that. Yeah.

30:29Speaker 2:And it's a feline. So, you know, you're cheating.

30:32Speaker 4:Theoretically our neighbor says that there's a Bobcat that comes out in our neighbor, our, our yard at night, but I've never seen. Yeah.

30:39Speaker 2:They're yeah. They're ghosts.

30:40Speaker 1:We're getting Bobcats in Connecticut.

30:42Speaker 2:Wow.

30:42Speaker 1:Yeah.

30:43Speaker 4:They came back to where they came back to Missouri where I grew up because the, the wild Turkey population finally recovered from the great depression. Yeah. Yeah.

30:52Speaker 2:I think that's happening in general. Generally, we hunt less. Our otter population is way up. Having to mink around is crazy. Definitely, the birds, there's more of everything. Bit by bit, we're going to have more encounters.

31:06Speaker 1:My late mother is probably solely responsible for the wild turkey population of Groton, Connecticut skyrocketing.

31:14Speaker 2:She fed them all the time.

31:15Speaker 1:Because she fed them corn, you know, cracked corn every day along with the deer.

31:21Speaker 3:Oh, man.

31:21Speaker 2:Nice.

31:22Speaker 1:And I always wanted to taste one that had been living on cracked corn, you know, for generations. But they're still wild turkeys. But if the only food they eat is her cracked corn, they must be delicious, you know? Maybe.

31:35Speaker 2:Maybe.

31:36Speaker 1:Anyway, she never let me kill one. So I didn't. And now they're gone.

31:41Speaker 2:She stopped. She's gone. So they're gone, right? There's nobody feeding them anymore.

31:44Speaker 1:She's gone. They're gone.

31:45Speaker 5:Anyway.

31:46Speaker 2:All right.

31:47Speaker 1:Well, this seems like a good time for a break. So we'll be right back after these very important messages.

31:51Speaker 2:Don't go away. And we're back. It's on the rocks. I'm Richard Campbell. Let's call Franklin. Yes, I am. Talking to our friend Jeremy Miller about the expanded critter stack. So I could see the AI tools allowed to accelerate. Although, I mean, Polkat and Fisher are adjacent to Martin in that sense. So they are some ways the tool will be really good at, you know, how do I do this in this other stack?

32:14Speaker 4:They are. But to get there and to get there and to make it sustainable is, We've had to do a lot of work over time to lift a lot more common functionality out, lift reusable compliance tests to get a lot of behavior.

32:31Speaker 2:You don't want to do that over and over again.

32:32Speaker 4:Yeah. So, you know, we can also dive into a little bit of a theme of you aren't going to be able to just vibe code something like a polecat over a weekend.

32:41Speaker 2:Well, that's a classic, isn't it? Right. Like, why do I buy a framework and I just make my own?

32:46Speaker 4:Yeah, that's delusional.

32:49Speaker 3:Um, sorry.

32:50Speaker 2:I agree.

32:51Speaker 4:Sorry. Coming from, I support one that's sort of couple that are widely used by a lot of different people, even with the AI stuff. And we're all using AI tools to do the fixes. I use it to do a lot of exploratory testing. There's just so much input that comes from people using it in real life. It's, it's all kind of, and it's the, It's the upset conditions. Kubernetes shuts down a node suddenly. Your DBA suddenly kill a database late at night and it will actually skew Postgres sequence values if you're not careful. We learned that one the hard way. Being able to have a system that's very highly in flight and then try to take it down gracefully without losing work as you go. That's actually really hard. Yeah. And the only way you get to a tool like a Wolverine or a Martin and make it good is you gotta have a lot of people beat up on it in real life and then, then iterate and improve. You're not just going to get that with a single time vibe coding pass.

34:00Speaker 2:Have you paid any attention to the use what works group? That's yeah.

34:05Speaker 4:A little bit. I'm going to admit that I've been a tiny bit cynical about it because it looks like the conference drinking club.

34:17Speaker 2:You're not wrong.

34:19Speaker 4:I appreciate what they're trying to do, but I'm not involved with it.

34:24Speaker 2:The message is essentially the same. This idea that your job wasn't to build a framework. Your job was to provide value to your customers. And these frameworks facilitate that. Use what works.

34:37Speaker 4:Yes. So one of the things I think I would argue or what I do argue to people of why you should still use something off the shelf. Like you look on these LinkedIn conversations, they'll say, oh, just let AI build you an outbox. something that's widely used is curated. Somebody is paying attention to it, reviewing the pull requests coming in, driving AI, watching the quality, the problems that people are having, and evolving it over time. You're just not going to build something over a weekend, no matter how good.

35:10Speaker 2:I wonder if this is even a temporary concept now that we're starting to actually pay for tokens. It's like, it ain't worth the tokens.

35:17Speaker 4:It's probably valuable in a way that it makes it easy to quantify efforts.

35:21Speaker 3:Yeah. Yeah.

35:22Speaker 2:Just like, look, spend the tokens on the value that you really bring to the customer. Buying a framework at a fixed price is a heck of a lot more sensible where you at least know what it's going to cost and that somebody else is paying attention to it. You don't have to own that code base and you don't have to spend tokens on it either. All right.

35:40Speaker 1:Awesome.

35:40Speaker 2:So, I mean, a few new products, but CQRS is sort of CQRS. How do you evolve event sourcing at this point? What's changing in this world?

35:49Speaker 4:Oh, quite a bit. I know you've had some recent hosts. So the last couple of years, last couple of years, there's been a trend. I don't entirely buy into it, but there is the idea of dynamic consistency boundaries where you stop. I don't know. Forgive me if you guys have already had somebody talk about it on the show. But breaking away from the model of strictly tracking events by event streams instead of doing it. really doing things more by tagging.

36:19Speaker 2:Yeah.

36:19Speaker 4:Um, where an event could, you know, the, the classic example is you have a classroom, you have events that apply to both the student and the classroom and maybe the instructor, uh, with the theory that this makes it easier. This makes it easier for you to be softer in designing your, um, stream boundaries. Um, If you'll let me be snarky for a second, I think this is partially driven by the weaknesses of some of the commercial event sourcing tools that can't do strong consistency or between event streams.

36:54Speaker 4:But even without that, people think that makes modeling easier. That's trend number one. The biggest trend that's taking up a lot of my time right now is, again, it's more AI driven, but people getting very seriously about event modeling and honestly, for us old guys, trying to bring the old model-driven development idea that never worked, trying to bring that into the modern world with event modeling, where you build models, maybe you complement or supplement that with kind of BDD-style specifications, and then AI-driven tools and scaffolding builds you an application or constantly evolves your application based on almost a two-way mapping between the event model of this is what the application is. These are the slices.

37:49Speaker 4:These are the events. These are commands. Generate me code from this and also generate code based off of BDD style specs. So those are the two big trends I see or things that are causing me to take a lot of time. And then all the old challenges of concurrency and scaling, those are all still real. And we're still working on new solutions for those all.

38:12Speaker 2:I mean, I don't have a big problem with event modeling because you do need to think through the event flow. I think it almost feels like they're going too far down the path. We had Adam on a while back talking about the event sourcing. And it was always that question of like, how far do you go with this model? It's not the whole application. It's just one part of the flow of data and how you want to manage it.

38:35Speaker 4:So everybody is the hero of their own story, right? If you talk to a tester, they think the most important thing is getting testing. Well, if you talk to a business analyst, it's getting the requirements coders, you know, obviously it's the code. Um, they've been modeling people, obviously their, their analysis work is key. Um, I think my take is it's definitely valuable. I hear from our own users that they think it's valuable. I think even though that's exactly what I'm working on today is the scaffolding from event models, I think that it does a lot of the easy work for you, and there's still a lot of technical challenges after the event modeling.

39:17Speaker 2:Sure.

39:18Speaker 4:And I think the event modeling is maybe a little bit oversold by some of its proponents.

39:23Speaker 2:Yeah, and one could argue that for architecture every time too, right? It's like every one of these all of who was it? Was it Eisenhower who said, you know, the plan won't survive content to the enemy, but the planning was worth it. So it's not a bad thing.

39:38Speaker 4:Yeah. You've seen, and you've seen that for, for younger folks, we used to quote that over and over again in the extreme programming, early agile days.

39:47Speaker 2:Right. Yeah. It's good to make a plan. It's also good to let that plan glow. It's contact with reality. Like you work through an event model and then real data starts to flow and you run into some issues and you have to change.

40:00Speaker 3:Yes.

40:01Speaker 4:Well, and of course I'm, I'm a history nerd. So, um, just happened to be listening to a podcast on the day for as horrible as that was, but how, um, how U.S. and British destroyers had to deviate from the plan quite a bit when all the communication broke down on the day of the landing.

40:21Speaker 2:Right. And find a way to still be effective.

40:24Speaker 1:There was decoys and stuff too, right? Wasn't there like some fake spy stuff that was going on that let the Germans believe?

40:32Speaker 4:There's one famous story of U.S. Rangers have to do a crazy heroic assault up a cliff And when they got there, the guns were telephone poles.

40:44Speaker 2:They were decoys.

40:46Speaker 1:And there was, like I was about to say, there was some spy stuff that was going on that made the Germans think that it was happening at a completely different place.

40:53Speaker 4:Oh, yeah. We set up a whole fake army around that to make them think we were trying to land somewhere else.

41:00Speaker 1:Yeah.

41:01Speaker 2:Yeah, as one does.

41:02Speaker 1:I wonder if those techniques could still be used. I doubt it.

41:04Speaker 3:I don't know.

41:06Speaker 2:There's funny things going on in the Ukraine these days. With inflatable tanks and all kinds of nutty stuff.

41:12Speaker 4:The decoys.

41:12Speaker 2:Yeah.

41:13Speaker 4:The decoys for drones.

41:14Speaker 2:Decoys for drones.

41:15Speaker 1:During World War II, there was also a fake Paris set up somewhere with an Eiffel Tower and everything to confuse the pilots.

41:23Speaker 4:Oh, I didn't know about that.

41:24Speaker 1:Oh, yeah. That was a classic thing. Anyway, we digress.

41:29Speaker 2:Yeah, we do. But that's not a bad thing, too. It's interesting to explore all the ways that you actually have to build the software to You're going to bump into issues. I guess it's always a debate about how many of the features you end up building in a tool like Martin end up being the reality check of. And then when we actually did this, you're going to have this problem.

41:50Speaker 4:So, I mean, as always, we learned, again, one of the lessons that came out of the Agile software development days before it became Scrum and got ruined.

41:59Speaker 2:Yeah.

42:00Speaker 4:That the design, the features you built should be pull, not push. You pull things in based on there is a demonstrated need to do this. So the features that have done the best for us, there's a feature in Wolverine we call global partitioning. What it is, is being able to take a look at some kind of business identification, you know, say an event stream, a classroom, and I'm going to make sure that all messages for this classroom are handled in sequential order, but I'm going to let other classrooms be handled in parallel. It's a really robust way to sidestep concurrency problems. That came out of... observing user needs you know what was happening in real systems right but other features that maybe we thought about for years that i thought was going to be a really big deal got it out there promoted it yeah not so much yeah.

42:59Speaker 2:Yeah you never know it's hard to anticipate you just gotta go to collect the data and realize okay now we do this a different way.

43:04Speaker 4:Now which you know that only works if you can't actually collect data and talk to people that use your tools And I'll tell you, that got a whole lot easier once I had a company and I was working for clients instead of getting secondhand stuff off of Discord.

43:19Speaker 2:Right. Yeah. How much telemetry do you put into these tools as well? It's always great to collect data in utilization. I just don't know how.

43:27Speaker 4:Comfortable people will be with that. So to be clear, we don't have anything that communicates back to us.

43:31Speaker 2:Right.

43:32Speaker 1:That's good.

43:32Speaker 4:Don't have anything like that. The Critter Watch tool I talked about before. does have the ability, you could use that to kind of look into a system, but it's not sending anything back to us. The best way I can get that, it's not fine-grained on features, but just looking at, you know, the Nougat extensions, like I know, you know, Wolverine has so many downloads, and then I can see that most, probably a little over half are using it with Martin as well, and I can see how many people are using it with EF Core and, you know, down to the RabbitMQ and google gcp pub sub i get that a little bit but no right there's nothing snooping on you.

44:13Speaker 2:Yeah which at the same time it's like that means invariably you build the products based on either customer feedback which is filtered or your own projects which are only going to be a subset like i i get it's uh.

44:25Speaker 4:That's very true but at the same token i'm also going to say that jasper fx clients get um They are the highest priority on what gets built. So what they need is what gets built. And we would be happy to work with you.

44:37Speaker 2:Yeah. Happy to help.

44:39Speaker 1:No problem. All right.

44:40Speaker 2:And if you need a feature built, I know a guy. I appreciate that. It's cool. All right. What else? What have we not dug into here?

44:50Speaker 1:Well, I can throw something in the fire here, which is the tendency for unseasoned developers to want to find a way to use every feature of a tool, even if they don't need it. Um, you know, I find a new tool and then, you know, it's great if it helps you identify problems, but then you would go further and say, Oh, well we should use this on our code base or whatever when you may not need to. And that's just comes with experience. But I mean, I've seen that happen over and over again.

45:25Speaker 4:I would give you the old, uh, I think it's a quote from Martin Fowler, but I don't remember where it came, where he did it. The only way to know how far a technique or a tool goes is to take it too far and then rein it back.

45:40Speaker 2:Yeah.

45:41Speaker 3:Yeah. Yeah.

45:42Speaker 2:You kind of have to, right? Over on the system inside, we talk about minimizing constraints on a new tool so that people will actually engage with it. And then you have, it's like, you got to let the puppy run. Just don't let the puppy throw up. Like, Let them go. Yeah. You know, we saw this with SharePoint. We see it with Teams, like all these things where we're going to need governance. But if we put it in before anybody starts, it gets very little adoption. If we kind of let them run roughshod for a while and explore and cause problems, then people want governance. And you've already got some adoption and value from it too. Like I think you kind of expect people to go too far. Yeah. I would agree with that. Yeah.

46:24Speaker 1:Funny, funny reality. Like there's never a perfection on any of this. Well, that was a short lived topic. Brought the conversation to a screeching halt.

46:33Speaker 2:What about the AI skills side of Critterstack? What's this about?

46:36Speaker 4:Well, one, I've got to stay in business. So there's kind of, there's gotta be something commercial related to AI. Yeah. So the AI skills is, Think of some of the sci-fi books you've read over the years where, oh, I'm going into combat. I'm going to put my combat skills program into active memory, right?

46:59Speaker 2:Very Matrix.

46:59Speaker 4:Yes. Or if you've read any kind of Peter F. Hamilton books. So the AI skills, they're going to teach your CLOD, your ASTRA, GPT. They're going to teach it, hey, I mean, the really markdown files, but a little more concentrated in documentation. They're going to try to give the AI tools... Guardrails on here is how to use the tools the best way. Certainly a lot of stuff around Martin itself, but taking Wolverine is a better example. Wolverine's value is that it allows you to create much, much simpler application code than... any of its competitor tools. There's just more things it does to simplify code. But if you let an AI tool go, it will try to use Wolverine just like its mass transit or mediator or on the HP side, it'll try to use it just like its MVC core minimal API. and you will lose a lot of the value of what Wolverine possibly can do as far as shrinking your code. The AI skills will, will teach it of, no, this is the idiomatic way to use it. Nice. You know, for Wolverine, it's no, you want to reduce the code to a pure function.

48:12Speaker 4:That's easy to test. I want to reduce layering. You want to, you know, Think of what somebody would do using clean architecture and don't do that.

48:21Speaker 5:Right.

48:21Speaker 4:Every possible way to simplify a code, there's a lot of skills to be able to utilize testing. Not be able to just to utilize the testing helpers in Wolverine and Martin, of which there's a lot, you know. But when you get into asynchronous testing, it also teaches your AI tool, hey, here's how to use all the built-in CLI tools, command line tools, diagnostic tools that we've built into Wolverine and Martin. Things like, okay, I want to go test. This message seemed to disappear. I can run a command line and it can tell me this is where the message will be sent to this route. If you end point or topic or whatnot, it'll explain the command line tools will explain to an AI agent. Here's what it's doing. Every possible bit of information about how the application is configured. If you have a Critterwatch tool, the AI skills will teach you how to use MCN points, in Critterwatch to be able to reach out into your open telemetry to say, hey, I'm working on this saga.

49:28Speaker 4:Tell me what's going on. That'll teach your AI agent how to reach into Jaeger or App Insights or whatever it's using through Critterwatch. And it will grab all the information about here is all the work that's already happening for the saga. Here is the information that's yet to come. Maybe it can tell you, hey, this message failed. This message happened at this time. But this AI skills, it's trying to unlock every bit of information about your system, both at rest and at movement. Cool.

50:01Speaker 5:Nice.

50:01Speaker 2:Yeah, it makes a lot of sense. And it just saves time, right? You're just going to get more done in less time that way.

50:07Speaker 4:Potentially saves time. And this is us going through the full cycle. I'm sure everybody in my position went through a cycle of, oh my God, we're dead meat. The AI is going to crush our business model to, okay, it's helping us deliver a lot faster, but we can use AI related functionality to build a lot more value in for our customers and maybe come out the other side and maybe have something that's That's hopefully something that people want to actually buy and pay for.

50:43Speaker 2:Yeah. No, no question.

50:45Speaker 1:Do you have anything else planned or are you done making critters for a while?

50:50Speaker 4:So the biggest thing for us in a, so there'll be a new critter watch 1.1. And by the time this podcast drops, we'll have a big new release. It's going to give you an option to be able to embed critter watch in your own application. like at dev time even. So you get a lot more insight at what's happening at development time. And more features, it's finally adding cron-based scheduling and then some other features. And then our big grand epic of the critter stacks version of event modeling and spec driven development, being able to sit in a room and, you know, be able to project up a meeting, meeting of, Hey, let's look at the, the visualization of the event model as we talk about what it's supposed to do. And is everybody agree with this?

51:42Speaker 4:And then push play and let an AI set of AI tools, scaffold it all out for you, including the tests.

51:48Speaker 2:Nice. Awesome.

51:49Speaker 1:So we're going live in two days. Did you, mean to say that by the time this podcast drops?

51:54Speaker 4:So, Carl, I went to Rice University in Houston. There's a famous story. So, believe it or not, Rice was really good in football once upon a time.

52:05Speaker 1:No.

52:07Speaker 4:But like in the 50s. And sometime in the early 60s, they decided, hey, we're going to use a whole bunch of leftover concrete from a Houston freeway, and we're going to go build the giant Rice Stadium that's still there today.

52:19Speaker 3:Hmm.

52:20Speaker 4:And we're going to get it done before the next season. So at one point, the story is that one of the board of governors is asking one of the Brown brothers from Brown and Root, famous company, hey, is the stadium going to be done for the first game? And he looks at him deadpan and asks, well, is it going to be a day game or a night game? It might be a couple of days after this goes live. You guys used to be slower.

52:46Speaker 1:I know.

52:46Speaker 4:Well, we're going to be there really soon. Thursday.

52:50Speaker 1:Thursday we're going to do a big demo. I looked up the Kelly Blue Book value of an old car and it said, is that with a tank full or empty? Kind of the same joke.

53:00Speaker 2:Nice.

53:02Speaker 1:All right, good. Well, we'll be looking forward to that. Jeremy Miller, thank you very much for catching us up on everything that's going on in Critterland.

53:10Speaker 4:All right, guys. Thank you so much.

53:11Speaker 2:You bet.

53:12Speaker 1:And we'll talk to you next time on.

53:14Speaker 2:NET Rocks.

53:36Speaker 5:NET Rocks is brought to you by Franklin's Net and produced by Plop Studios, a full-service audio, video, and post-production facility located physically in New London, Connecticut, and, of course, in the cloud. Online at pwop.com.

53:52Speaker 3:Visit our website at dotnetrocks.com for RSS feeds, downloads, mobile apps, comments, and access to the full archives going back to show number one recorded in September 2002.

54:05Speaker 1:And make sure you check out our sponsors. They keep us in business. Now go write some code. See you next time.