Viktor 00:00:00.000 November last year, I was in denial about AI. Complete, I mean, I knew that it's a thing. I knew that it is on a road to become useful, but I don't know whether I said it, but I'm hundred percent sure that I thought that there is no way that a year later, today, I would be basically using it all the time. Kind of everything, for everything.
Darin 00:01:30.444 Back at the beginning of the year, we had our predictions episode for 2025 today. Surprisingly enough, it is New Year's Eve 2025. We're heading into 2026 and we're going to see how good or bad we were with our predictions. Viktor, what do you think our prediction probability was?
Viktor 00:01:51.538 I mean, historically, uh uh, I missed most of predictions so far. I'm terribly wrong most of the time. I dunno why anybody listens to me.
Darin 00:02:00.660 Well, we didn't have that many last year. And if you wanted to go back and listen to the episode, you know what? I don't have the number in front of me right now. Just go back because it was, I think it was even January 1st last year. This is another one of those years where everything sort of falls on the same days it feels like. But again, date math is hard. Uh, we had four items that we had as sort of the big buckets. I am going to do them in the order that I think we did them in. 'cause yes, I did the research, but I didn't do that close of a research. At the end of 2024, there were a number of rug pulls, I think back to, HashiCorp with Terraform going to BSL. There were a number of other, and our prediction was there were gonna be a lot more rug pulls in 2025. I think we completely got that one wrong. I don't remember seeing anything of great importance this year.
Viktor 00:02:52.500 I mean, at least in my area, CNCF, I think that companies are struggling, but it's not yet. So that was wrong, kind of. It did not come to fruition yet. Uh, or no, fruition is actually wrong word, right? Kind. Uh, there, there were rock pools did not yet happen, but I, I stand, I'm going to use that in the next episode as my repeated prediction.
Darin 00:03:18.001 Really, you think there's gonna be more rug pulls even we won't get into the predictions for next year, but you really think that's gonna keep going on?
Viktor 00:03:24.153 Yeah, because a bunch of companies that are not massive, like AWS and companies that are not born in ai with ai, they're having a real trouble, um, raising funds. And you know, when you have travel raising funds, then you do desperate things.
Darin 00:03:44.549 And usually desperate things lead to even worse things.
Viktor 00:03:48.373 Yes.
Darin 00:03:49.164 So going back to which episode that we made the predictions on, it was actually, I'm gonna look at my notes again here. The prediction episode was actually on January 1st of this year. It was episode 2 9 6. Just in case you're wondering again, I should have kept all that. In my notes while I was preparing for this. Uh, so rug pulls? Yes. Not really, but yes. But maybe, but no. Right. Is that, is that what I'm hearing? I'm doing my notes at the same time. That's the reason why I'm looking away at this
Viktor 00:04:17.638 All of it at the same time. Yes.
Darin 00:04:20.019 Okay. I'm actually going to, the next one was ai, but I wanna hold on that one for just a minute.
Viktor 00:04:27.463 Mm-hmm. ' Darin: cause I think the other two we can go ahead and knock off the list. Uh, we talked about Wasm. Wasm was going to. still not be there. In fact, we'd, we'd actually named the episode the year of not yet. Yeah. Uh, I honestly, I forgot what I said, but if I said it, wasm is not gonna get far. Yeah, that's true. It's not happening and. Now it's even less likely to happen.
Darin 00:04:56.797 Yeah. Okay, so Asum, we were right or wrong, depending on how I listened to it. Again, I probably should have listened closer.
Viktor 00:05:04.026 I think that we were right if, if the prediction was, it's not going to happen. Yeah.
Darin 00:05:10.157 yeah, still not there. The next one was, there were gonna be a lot more small company purchases, like roll-ups, buying smaller companies. I think that was probably 50 50. I saw some, but it wasn't as many as I thought because of what we'll be talking about next.
Viktor 00:05:27.462 Yeah, because kind of nobody wants to buy the companies, which I believe I had in mind a year ago. Right. And this is probably related to Rockpool, because. If this year, if you're buying companies, you're buying companies because of two letter word, right? you're not buying anything else. So it would be silly of you as a company to buy a different company for anything else. I'm not saying that that's right or wrong, just to be clear, but that's where all the budget goes.
Darin 00:06:03.249 And that two letter word of course is ai, which was the one I was skipping over. So we had stated AI will just be there. We won't need separate bespoke tools to do our work. The way that I phrased that right then, I do not believe that we're there yet.
Viktor 00:06:22.919 can you repeat that? We'll not need separate bespoke tools.
Darin 00:06:26.680 yes. To do our work.
Viktor 00:06:29.436 Ah, okay. Okay. Like, like it'll be engraved into the tools that we use. that's actually, I think that we are, I, I dunno how far, uh, whether we are fully there, but you can see that already. Right? So where do you code in vs. Code? Right? Where is AI in VS code? What do you do to edit images? Photoshop. Now, I don't use Photoshop much, but I have friends who do, and I understand that, uh, none of, and quite a few other models are now part of Photoshop, And you're going to use Photoshop. In combination with models, right? Some things you're going to do yourself, like topography typically kind of something that AI is bad at, but then you want to edit the image ai, right? It's so it's there Now, go going further, right? office 365 and basically Microsoft put AI into everything. Literally everything. Windows ai, uh, office 365, ai, and so on and so forth, right? what else? Uh, Google something, something Docs, uh, Gmail. What's not, uh, I'm looking at my Gmail right now. There is that, star shaped icon over there. Ai,
Darin 00:07:46.349 Yeah, there's a way you can turn that off by the way.
Viktor 00:07:48.858 uh, yeah, I, I don't use it, but how useful it is. I dunno whether that was prediction, but it is part of the tools and it, to me, it's pretty clear. That existing players are dominating that market, not necessarily mar market of the of models themselves. That's a separate story, but AI is definitely this year went into the tools that we are using.
Darin 00:08:16.326 I think. One of the things that we didn't see coming or maybe couldn't have seen coming was this explosion of agent-based ai.
Viktor 00:08:30.310 Okay.
Darin 00:08:31.026 Again, to rehash the story again, um, my wife got sick in February, so I basically went offline until basically June, and when I came back in June, that just, that March to June timeframe was like. What did I miss? What happened? And I cannot remember a time in my career that a tech came in that strong, that fast, and that pretty much good. I mean, sure we've had stuff come in fast, but it wasn't like, wait, this is actually usable. This isn't just a toy.
Viktor 00:09:12.646 exactly that, that that's Gen AI is essential in a way. A way how to allow AI to use the tools. And this is pretty much related to what we were talking before, right? In our case, agen AI is, hey, operate, uh, by executing commands in my terminal and, uh, writing code and uh, calling some APIs, right? If you're designer, those tools is Photoshop itself. now, it's not fully there, but it's definitely agen, right? And basically everything else, if it starts using tools, it, it becomes agen. That's, that's the major difference and we're definitely there. This, this was the year of agen ai. No doubt.
Darin 00:09:53.781 Yeah, but could we even have guessed that? So, you know, full transparency, we record these 2025, uh, actually the year in review and looking ahead, we record those at the end of November. Typically, normally it's Thanksgiving week is when we record it. 'cause that gives
Viktor 00:10:08.335 Yeah, back then you could hardly even hear the word agent.
Darin 00:10:12.441 Exactly. There wasn't, there just wasn't much going on. It was all chatbots at that point. Or I'm gonna say, excuse me, not chatbots, but chat. So it was chat, GPT and the other chat. One shot. Well, I consider one shot type deals. Right? Because chat effectively is a one shot, A threaded one shot.
Viktor 00:10:32.208 I'm not even sure that we even, or I, let's say, mentioned the word age agent, agent in when we talked about it a year, something ago,
Darin 00:10:42.139 we did. Yeah.
Viktor 00:10:43.488 but we did talk about that becoming part of the existing tool sets, which is in a way. Almost the same. We didn't catch the wording of it.
Darin 00:10:53.513 I think that was the one thing that we missed. I'm trying to think of other things that we missed there. I, as far as I remember, there weren't any major, this goes back to the buying up of small companies. I don't think there were a lot of buying of companies, period this year.
Viktor 00:11:06.772 Oh. Like, uh, the winder,
Darin 00:11:10.013 Okay. You're right.
Viktor 00:11:10.912 bought twice.
Darin 00:11:15.278 Go look that one up if you don't understand what Victor's talking about.
Viktor 00:11:18.812 Yeah. Yeah. Have to sell a company two times. Um, and a bunch of, uh, I think that two. We are at a slight disadvantage, at least me, that I don't really memorize. I know that a bunch of companies were bought, but if it's not directly related with software, I don't kind of, so sorry. With that being for software engineers to be more precise, it's always software. Uh, my brain does not record it, but a bunch of companies were bought. It's just that, what I didn't predict is that. Companies that are a year or less old are gonna be bought, not companies that exist longer.
Darin 00:12:02.426 Again, this year has been a, a blur to me in, in a number of respects, but I think the thing that we can say about our predictions were that, you know, again, we threw laws 'em out there just 'cause we have to, you know. It always shows up and people are like, oh, awesome. I have not seen people talk about it hardly anymore. companies, we know that money is going towards ai. Unless you have AI happening for most in your company, you're not getting funding. Even the big AI companies, the big, the, the AI startups are only getting single digit, seven digit, influx is like $3 million. Not a lot of money. I mean, that might pay their. Open AI bill for a month? Well, not even a month, A day. Let's, let's be real. Um, I don't know how we could have seen this year coming, 'cause other years. I mean, sure we had, it was hit and miss, but it wasn't like this big shift that happened
Viktor 00:12:57.676 No, I mean, at least I was in denial. So November last year, I was in denial about ai. Complete. I mean, I knew that it's a thing. I knew that it is on a road to become useful, but I don't know whether I said it, but I'm hundred percent sure that I thought that. there is no way that a year later today I would be basically using it all the time. Kind of everything for everything. and doesn't matter whether we are talking about writing code, whether we are talking about operating cluster, whether we are talking something in AWS, whether we are talking, writing a blog post, or, I dunno, like, uh, trying to figure out what to watch next to Netflix. Can everything, I, I had no idea. I honestly haven't, no.
Darin 00:13:55.071 Can you think of a a year, a calendar year where. In your career, things have happened this fast.
Viktor 00:14:03.825 On that scale, never there. were moments when they were happening. For me personally, like
Darin 00:14:09.082 Sure.
Viktor 00:14:09.956 when I discovered Docker, that was the full year. Kind of like, I did not want to think about anything else. I don't wanna do anything else. It's Docker. I'm, using just an example, right? But that's for, that was of great importance for me that year. It wasn't that great importance for the world, right? It did not change the whole industry or the whole planet or whatever. so nowhere near that, not even a tiny fraction. so no, never on that scale, never.
Darin 00:14:42.835 So I think we have to give ourselves a pretty much failing score, just like we do every year, or 50 50, because I, I just don't know. I don't think we could have seen it coming. Uh, well actually lemme take it back. We might have been able to see it coming, but because of what we were doing in our side jobs, day jobs, all the jobs last November of 2024, it looked like just another year upcoming to me.
Viktor 00:15:14.601 I I feel that basically it's all we should for, for this failure of ours. We have to blame agents, right. Simply. I understood that. Okay, so models are moving fast. I understood that. And, but kind of what do you do with them? Kind of like, how much time are we going to spend with GPT? Right. I don't think that the models are now much better than what I thought they would be. It's just that they're not the most important thing anymore. Agen ai, that's really what matters. And that this just happened.
Darin 00:15:51.882 if you try to keep up with the models, does it even matter because you don't use a, a model directly? Let's exclude nano banana for just a minute because Okay. Use that one
Viktor 00:16:01.886 Yeah, yeah. Or you know, maybe change a PT or something like that. But I would say that if you are tropic user or Germany user. There is no reason to switch. I mean, there I can come up with reason. Just to be clear, I do believe that Tous, uh, is absolutely amazing. I've been using it for a couple of days. I can, it's, it's, it's, it's mind blowing. But those are not the differences that you say, okay, I'm going to switch to a different vendor because next week it'll be something else.
Darin 00:16:36.288 When the foundation models were just getting going, then the argument could be made. I think the foundation models now are, they're not parody, not by any stretch of the imagination. But they're all just so darn good that you know, just, okay, this week Opus four five is better than Gemini three was last week, and next week it'll be somebody else or whatever. It's right. Those are, but we're not seeing, what we're not seeing in those places are the hockey sticks that we were seeing early on. The hockey sticks are in the usage of those models, not the actual models themselves. And that's where, you know, again, shouting out to Anthropic, where Claude Code, Claude code comes into play and how they have gone from skills to commands to agents. And they all do the same thing, but they don't. It's like, what do we do with these things? I mean, it's like that, the tooling that that's being put into our hands as developers, whether you like it or not. is just what it, where it's at now, and I wish we could have seen that coming a year ago.
Viktor 00:17:49.764 I mean, just if you look at Claude code alone, I can safely say that every month there was a mind blowing update. Every month you come and say, okay, oh, MCP, oh, this changes everything, right? And then next month, subagents, oh, this changes everything, right? And so on and forth. It's just amazing.
Darin 00:18:17.537 all those things were happening, but now, I mean, we're still seeing some of those come along every once in a while, but I'm gonna put it back in your space. This is almost like what we had with Kubernetes, right? Hot, hot, hot, hot. Slow, slow, slow. A little peak gateway. API. Oh, but the other thing we didn't see coming this year, let's mention it while we're here. Uh, ingress, NGINX, getting tombstones. Uh, I didn't see that one coming.
Viktor 00:18:46.041 If you ask me specifically, I'm pretty sure I would have guessed that it's not, it's going to be in a bad state, whatever the, the actual state is, but kind of ionix was never on in my, on my radar, to be honest.
Darin 00:18:58.607 Yeah. But anyway, going back to that, we can see the parallels to what happened in the Kubernetes ecosystem right there, there they're still happening a little bit faster in the AI space. But again, I think that's going to eventually slow down much like what we saw in the Kubernetes
Viktor 00:19:15.051 Oh yeah. Because we are so early, so we, we are approximately give or take a year into agen ai. And this is the, the year of discovery and probably next year as well. Right. We are still discovering what works. I don't know, man. That, that's when we get ideas, and that's, that's very much very similar to what was happening at Kubernetes. Once we discover all the things that actually we need, then we don't focus anymore on, on continuing discovery, but making it work.
Darin 00:19:48.201 That might be a good place to go ahead and end This one, and you just gave us the title for what 2026 is gonna be is the year of discovery, whereas. 2025 was the year of, not yet. We're looking at 2026 being the year of discovery. That's, that's pretty good.
Viktor 00:20:01.900 Oh yeah.
Darin 00:20:02.931 I wish you were that. I wish you could remember come up with all these every year like that, but we'll have to wait and see what, uh, 2027 is. That'll be interesting. But anyway, so 2025, we did not get it right. But here's what we do know. Again, it's December 31st as you're listening to this, or at least when it's being released. It is New Year's Eve. So have a fun, safe New Year's Eve and remember, you can start doing your taxes tomorrow after you wake up. So, Viktor, anything else to say about 2025?
Viktor 00:20:36.250 I think that's it. Big failure on our part, extremely successful year. On any other aspect, it.
Darin 00:20:43.175 What do you think? Head over to the slack workspace. Go down to the podcast channel, look for the episode number 3, 3, 1, and leave your comments there.