Gemini 3, Cloud Code, and the New Speed of Building: Why Everything Just Changed

The AI world experienced a shockwave this week — and today's conversation dives straight into it. Google’s Gemini 3 dropped, Anthropic announced a major partnership with Microsoft and NVIDIA, and Cloud Code quietly became one of the most powerful tools in a builder’s workflow.

In this episode, Chris and Nico break down what all of this means for developers, operators, builders, and anyone paying attention to the accelerating curve of AI capability. From multimodal performance to real-time app generation to the future of HubSpot adoption, this discussion explores how the stack is changing under our feet — and why velocity, not just intelligence, is the new story.

What you’ll learn

• How Gemini 3 fundamentally changes the speed and ease of app creation

• Why Google’s positioning matters (and how their “OG multimodal” flex landed)

• What Anthropic’s partnerships signal about compute, scale, and competition

• How Cloud Code + Gemini creates a real “idea → deployed app” pipeline

• Why front-end scaffolding is becoming a solved problem

• What this means for HubSpot’s ecosystem, data model learning, and adoption

• Why AI-first builders now have a structurally different workflow than everyone else

In this episode, we cover

00:00 — Setting the scene: Gemini hype, Thanksgiving countdown
01:30 — First impressions: “Massive difference is not doing it justice”
03:10 — Benchmarks, performance jumps, and why it feels like a new model
04:40 — Anthropic’s compute deals and the Microsoft/NVIDIA partnership
05:30 — Cloud Code becoming available to paid users
06:10 — OpenAI’s expected rebuttal (and why Sam never waits)
07:30 — Why most people haven’t made the “two clicks” to get started
09:15 — Google’s OG multimodal flex and marketing sophistication
11:05 — Personal data, Workspace context, and ecosystem advantage
12:20 — Nico going “full metal” on dev-first workflows
14:00 — From Canva workflows to building apps on the fly
15:20 — The power of iteration when foundation exists
17:00 — Why dev speed becomes the ultimate unlock
18:30 — How juniors, seniors, and Cloud Code now collaborate
20:00 — “My tech stack is about to go in the garbage”
21:30 — Building front ends before touching backend logic
24:00 — Complex HubSpot data tooling built in minutes
27:00 — How kids will soon be shipping apps
28:30 — From Gemini → GitHub → Cloud Code → deployment
31:00 — Real examples: property history, associations, data explorers
33:40 — Why AI is finally understanding how humans build
35:10 — HubSpot adoption challenges and the missing applied-learning layer
37:50 — Visualizing objects, associations, and data models
39:20 — Inside the HubSpot Learning Canvas and why it matters
42:00 — How partners can now ship tools, not documents
44:10 — Why value isn’t the hour — it’s the enablement
46:00 — Amazon’s early data: fewer returns with AI shopping agents
47:00 — Wrap-up: holiday-season acceleration

Tools and ideas referenced

• Google Gemini 3 (front-end code scaffolding, multimodal, benchmarks)

• Anthropic + Microsoft + NVIDIA partnership

• Cloud Code (runtime, deployment, backend generation)

• GitHub workflows (PRs, merging, repo-level collaboration)

• HubSpot CRM (objects, custom objects, association model)

• HubSpot Data Model Builder

• Learning canvas / internal tools scaffolding

• Vibe Coding (development paradigm referenced throughout)

Key takeaways

• Front-end scaffolding is now instant. Gemini 3 can generate full React apps in minutes, shifting where developers spend their time.

• Back-end workflows compress. Cloud Code turns front-end outputs into deployable backends with minimal friction.

• Context becomes a moat. Tools with access to your Drive, Workspace, or CRM will feel more intelligent to you personally.

• Adoption tools will be built, not bought. Visual, interactive, in-context learning assets will replace static docs.

• The HubSpot ecosystem is poised for a surge. Partners who build applied tools — not PDFs — will win.

• Speed beats everything. AI isn’t just smarter — it now removes the parts of development that once slowed innovation.

Transcript

Chris:
Good morning, LinkedIn friends, Value First Nation. Welcome to another episode of Value First AI Daily, your collaborative AI intelligence report. It is Wednesday, November nineteenth, twenty twenty five. Thanksgiving is one week away. How are we doing, Nico?

Nico:
You know. we're just, we're just trying to contain ourselves today. Uh, we're just, we're going to try to try to have a normal day, you know, try to keep the hype down as much as possible.

Chris:
No, that's incorrect. That's not.

Nico:
Yeah.

Chris:
Okay. That's okay.

Nico:
Good.

Chris:
Yeah. Cause, uh, three drop man. And, uh, It's over for so many people. I downloaded anti-gravity yesterday. I finished installing it this morning, which was two clicks. That's how little time I had yesterday to mess with it, that I was two clicks away from finishing the install. Finished that this morning, still didn't get a chance to touch it, but I did get a chance to throw one prompt into Gemini three and that's all I need.

That's all I need to understand how much further this is from two point five and leaps and bounds is not doing it justice. Massive difference is not doing it justice. This might as well be like a completely new model, which obviously it is, but I mean, like. It's just it's seriously like it's the step between. I don't know, three point like GPT three five and GPT four one. Right. Like it's just it's massive.

And when you look at, you know, not again, like if you're a benchmark person, then you care if you're not a benchmark person, then you don't care. But if you do look at the benchmarks, it is destroying everything out there. I mean, like chewing it up and spitting it out. Not even funny. And again, with the little time that I have played with it. It is. I mean, I understand there are some deals with the devil that have been made today that I did not see coming like this. And I don't know if this is what sparked it.

But like, so you got you got Gemini. I know we're going to keep talking about it's like Gemini three. And then on the other side of that, Anthropic makes a deal with Microsoft and NVIDIA. Somebody needs compute real bad.

Chris:
Oh, yeah.

Nico:
They're always out there. On top of which, Cloud Code now integrated with your membership. So basically, Cloud Code for paid subscriptions now. So a lot in the last twenty four hours. And that's that's an understatement because OpenAI has not rebutted yet. And we all know that they do not like not being at the top of the news, whether it's good or bad.

So like, hold on. Cause Sam's been talking GPT six. He's been rapidly releasing stuff. know the whole i'm not doing it until next year has never the whole i'm not doing it until has never actually stuck with me that usually tells me oh he's going to do it in half the time that he says he's going to um i'm waiting for the rebuttal i know that's going to be crazy but can we can we please get into gemini because it's

Chris:
yeah no like whatever if you ever want to share like go for it I've got stuff to share at the end, but I think like what she said, and this is where Sam has been in the news, right? In terms of like doing all these deals, everybody's trying to get chips and build data centers and build the infrastructure.

Like you can tell, I mean, and if you experience it, like the intelligence of the models is where like it's for ninety nine point nine percent of us, it's it's smarter than it ever needs to be. right but what was interesting the way you started i didn't have time or i didn't do the two clicks that i needed to do right and that's where um it's almost like because every most people expected this from google like and there's so much prediction that by the end of the year google would release and it would be the best like guaranteed and we've talked about a lot

It makes sense. They have all the data, all the infrastructure. They've been doing it longer than everybody else. The fact that they waited until April of this year to be like, okay, we're going to play now. They've been playing behind the scenes. I love the positioning of the video. That's where you can tell. When you compare their delivery of a new model, and the videos around it, like with the frontier labs. It's like, yeah, they know what they're doing.

Nico:
Well, they also have a big marketing department.

Chris:
Right.

Nico:
Well, they've been doing it for a while, too. Like, again, they're the immovable object in this AI room. But when they started the video, like, you know what? Our first model was multi-model, bitches. Yeah. Like, you're going to come in, and we've been doing it. We're the OG, like, basically, is what that video was describing.

And then it just shows you a little bit here and there. They had, like, the influencers around them posting, like, videos immediately. And... Uh, but it is like what we should expect. They have all the data, they have all the infrastructure and I, and, and George was telling me about, you know, some stuff he was doing. And this is where that world of like, if you already have a Google drive and workspace and like all of your personal data is there too. It's like, of course it should be more contextual for, for you.

And where I'm at is I'm happy that. like the two clicks right now that you mentioned are usually in people's minds. Like they haven't made that. And that was me prior to two Sundays ago, not touching vibe coding. And so I'm happily like, I'm happy to be in this position where I can start to go over that way and enjoy, because I have gone full metal Nico in that any plan that comes up, I'm like,

Um, or any need that comes up, I'm like, okay, can I build that? Can we just try to dev that first? Like, I don't need to go into Canva and spend an hour trying to put, uh, you know, a process map together or whatever, I'm just gonna try and build the tool that just helps me do it. Right. And I'm going to build it right into the website at the same time. Oh, and I like it on this website. I also like it over in the simulator. But it's a little bit different over there. So hey, A cloud code, like I'm just gonna, or Hey, GitHub, you can see both repos, just take what's over here and put it over there, but you have to change the language.

Like that's the thing that's just absolutely sick is the way to, the way that you can iterate. And again, we're back to this concept of if you're building foundation alongside this progress, the speed to value is like seconds as, as other people like deliver all the value. and like create all the value, you just get this shot. Like this is abundance. This is what, what that looks like.

Chris:
Um, so, and, uh, yeah, let's see, let's see what we got over here. Um, but yeah, I'm happy to like be in a place where I'm gonna, I can let myself, you know, do those two clicks basically. Like I've done all this foundation. And I can say, Hey, take this from over here or take this idea. Cause so much, I'm only, I only have a website and a new app in nine days because of the, like that I've been working on stuff and documenting, and I've got diagrams all over that I can just shoot to Claude and say, Hey, I want to build this as an app. Like, give me a prompt, right? Like.

So it's enjoyable. And at the same time, I feel the need to apologize to all the developers out there for suggesting that we should do everything in our power to try and do stuff inside of HubSpot. Because I'll just say I feel pretty powerful developing my own shit and being able to iterate and have No boundaries.

Nico:
So I get it on some level. Obviously, I'm accepting the fact that there's no way I could hand this off easily to a client and have them build their own. I have no idea what's going on back there. I mean, probably at the point where I could just do a quick export or make the repo public or whatever, however it works, to say, yeah, you want to use this too? Yeah, here you go. Just throw it into Gemini. make your own version of this right or or just tell jeb and i look at this thing and then say make it right like that's where we're at that's that's pretty much yeah i mean like that is where we're at it's not even like pretty much and like and this is just like the second one like i've already messed around with this this morning so it was really ridiculous to me because I've done this before.

We've done demos before, so you can actually go and look up our old demos of when Gemini two point five came out and we did this exact same thing where we put in a prompt for an application and we sat there and we waited nearly the entire episode and just about like ten or five minutes before the end of the episode, it spit out something that we could all take a look at and click through and that kind of thing.

And yeah, I was quite surprised that like this happened. in that short amount of time, and usually takes way longer than that. If you watch on the far right hand side, and it's going to the far upper right hand side at this point, that is sort of a measure of how much code is getting written, because that's how large this window is. I'm willing to bet we're gonna go about four or five hundred lines deep on this.

And I mean, you know, it's doing it like it was just not a big deal at all. I actually did this this morning for a different purpose. And this is why I say that like your tech stack is about to go in the garbage because in one request in five minutes or something like that, you can start building some type of data tool that you can use.

Now, what sucks is, OK, oh, wow. Earlier, it was like, yeah, sorry, I can't really work around the course problem. So that's something that you're really I was like, all right, that's going to be a slight limitation, but I'm just going to nab my key real quick. But yeah, even this morning I had it put together a property audit tool because, you know, just to see like what the limitation was for it and was very happy to see that like it doesn't really have one, which again, I was pleasantly surprised about.

So, you know, just have both of those rolling and it's the same thing, like just API key connect to your CRM and roll with the punches. And like I said, it's amazing to me that it managed to pop it out in less than two minutes like that. It's it's really like with the level of detail and everything is where it starts to really blow my mind because we've used so many of these apps.

I mean, we were using mini Macs, like not even that long ago. Right. Um, and something like this was, I don't, you know, still probably would have taken a bit of time. I'm not going to say that it would have been like easy or what have you, um, see if it will actually load.

Chris:
Well, yeah, of course policy got us anyway. All right. Not a big deal. All that means is like, we'd have to bring it local and connect that way because we can't use weird proxies.

Nico:
So, um, but yeah, I mean like from this to then taking it like Downloading, I mean, you know exactly where it goes. It goes from this to cloud code, like into a repo and then telling cloud code, like, Hey man, I just got started on this. Do you think you can help me continue? Right.

But why would I do that? So why start here and not go to cloud code? Are you not watching? Like you can build the front end and interactivity of your front end before you have to do any of the backend. And if you've worked with Cloud Code, then you know you're basically building in a blind box.

So the ability to pretty much build out the entire front end of your application in React prior to having to go into Cloud Code is pretty ridiculous to me. And I kind of want to see how far we can go with this, of course.

Hey, so this is pretty awesome, but do you think that you could actually build in several other like data cleanse cleansing tools, like maybe a property cleaner, something for contacts, something for deals, something for companies, something for tickets, and then something that'll let me take a look at marketing emails and the statistics there as well as traffic and the statistics on traffic. And it's not like.

It's not like I'm just, I don't know, like I have this stuff on mine. It's more that like, I just want to throw a bunch of garbage at it and then see like, Hey, with all of that list of things, can you do all the things on that list? Right.

Um, and that's where I get the measure for, because I've done this with other models and that's how I get the measure for like, okay, well, how much more advanced is this one than others? It's actually really good. Uh, world clock test going on right now between all the models to see who can render a clock the most accurately and have the time move the most accurately minute to minute. Oddly enough, Chinese models are really destroying things right now. No, not really.

So we're going to add a data editor. I mean, again, it's not like I'm looking for crazy features that we don't already have in HubSpot. It's a matter of being able to basically develop the entire front end of your application prior to having to even mess with going any further from there. So that you have a full understanding of A, what you're building, B, what it looks like, and C, what you're expecting it to do so that you can take this front end.

And after we've gone through this whole thing, right? You know what the best part about this is? It's my junior developer. So my junior developer can now, since he knows all the stuff that we've done, can write out the technical blueprint for the backend developer to start working on. So all these files in Git, plus a blueprint written by the same person that wrote the front end to begin with, given to Cloud Code, equals finished backend that plugs in very perfectly to your front end.

Right. So in terms of like, what's the big deal? That's the big deal that you can like the app that took not, you know, and that's what's getting crazy about saying this stuff. It's like, oh, the app that took you two weeks. Now it's going to take you two days. Yeah. It's like, yeah, I mean, I guess I know I'll know when the paradigm has shifted, when these auto coding tools stop telling me stuff is going to take weeks.

You know, like it's still like, that's how much it's ingrained in, like, and they're working from a place of just knowing how it goes. Right.

Nico:
Um, uh, but yeah, the ability to. Hey, but if we had to build this out together and we're working together on it, about how long is it going to take for us to put together all the code for this and the backend? Please don't come back with months.

And, of course, our buddy Ryan Ginsberg, super interested in what's happening on the screen right now.

Chris:
Yeah, of course.

Nico:
And it gets property history and associations. Uh-oh, Chris.

Chris:
Yeah.

Nico:
I was ready for Google to tell us like it is. Ooh. Right? Finally. God. So, yeah, I've started ignoring it. but it's just annoying.

Do you see this guys? Are you guys reading this?

Chris:
Yeah.

Nico:
A single Saturday. Yes, this is where I live. That's fine. That's there's a reason why it's personalized.

Chris:
Yeah, that's.

Nico:
I'm sorry. I'm I. And I know that's been your litmus test for a while now and. i'm i'm sorry i'm fucking blown away like by that answer in particular like the depth of it and the reasoning behind why it's going to take that little amount of time right which means it's learned about this moment in time of humans interacting with vibe coding.

Um. And that would tell me that vibe coding in general is, but models are getting better, but also like humans who are also doing this stuff. are actually getting this stuff done. Right. Cause if you leave the bottles to themselves and they're the only ones that have to do it, it's done in like minutes. Right.

Um, but the collaboration requirements, again, this is the kind of stuff that makes any kind of product or implementation take longer is getting the humans involved in talking to each other and buy in and what do we want, what we don't want. Right.

But the execution now, the build should be the easiest, fastest part.

Chris:
I mean, it's it's absolutely friggin it's really it's seriously insane to me. Because.

You're going to see so many applications and do you know where they're going to come from? Kids. They're going to start coming from kids real fast because this stuff is getting so easy.

Look, Chris, we just went through it. How much easier would it have been for you to jump in if this were out? And we just went through like you getting into vibe coding. How much easier would it have been for me to say, oh, Chris, here you go. Start with Gemini. Then you use GitHub and then you just go to cloud code. And when you're done, you go to your fourth thing. Right.

Like just a straight, like now I'm not even joking. That is a straight workflow for me right now is Gemini three Claude code launch.

Chris:
Yep.

Nico:
Like, yeah, that's, that's the only reason we have to do is so that code Claude code has an understanding of what we're doing, but yeah. Yeah. The get hub, uh, like.

PR and merge stuff. Like, that's the only part I would add.

Chris:
But yeah, that's nightmares.

Nico:
Yeah. Right. And, again, like the ability to proof of concept, we're not saying because I really don't, it's hard to make this viable, because there's a whole go to market that needs to behind it. That doesn't mean everybody I don't think should look to this to like start businesses necessarily. But when you're struggling with adoption challenges, and how to use software, like, this gets a lot easier when I'm not trying to sell something, and I'm so focused on teaching, because it can be so easy to come at this from a competitive perspective.

When in reality, like, because like I mentioned, I like doing stuff like this. And I like helping people understand. And what this is, is like HubSpot Learning Canvas. And again, like I've already realized that now this foundation is in place.

Like example yesterday with the other app, Nico sends me the ACP doc. I'm like, hey Claude, let's put this in the app and let's put it everywhere. And then it happens in fifteen minutes because that foundation was there with all the other data points to make it make sense related to the examples that we're showing.

This is all the stuff that since HubSpot has to start from scratch, it's really hard for them to bridge the gap between academy and usage of the tool with applications.

So what I'm trying to accomplish here is, and I've literally, you have clients tell you, like, if you can just give me the visual and walk me through it and give me something to do with my team, which is that's where it gets super hard building these resources that are shareable and like, don't get screwed up when you try to print them out, like all those things. Right.

So this is like a, a canvas, right. We're trying to, you know, help people understand HubSpot data model.

And, yeah. And if I can help you get there to understand that you should care about this in the first place, then I can get you to start using the in-platform tools that you have available. Because this is where HubSpot's going, is trying to help you with the, literally, data model builder, right? Do you know you have all these objects in HubSpot? Probably not.

Do you know that this week you're going to be able to associate orders, carts, and invoices to custom objects? Probably not. But if you get used to using these in-platform things, you will become aware much more quickly. You'll be able to action on all these updates more quickly. And this is going to be like HubSpot can't do the applied like learning course style stuff that we need them to do and we shouldn't ask them to because they're just busy building product and we need them to help us learn the functionality of their product and not how it's applied to a hundred and fifty different use cases that's what the partner ecosystem is for right but the ecosystem not so great at it.

Uh, and it's easier than ever before if you can just say hey like anybody can do this right so let's not focus on the hour that it's going to take us and how we make sure that we make money off of that hour because everything's based on billable hours and instead oh we actually want people to get value out of the tools and the services that we provide.

So we're just going to build stuff like this into the website, into our experience. I'm fully unlocked here. And this version of this, and what I mean by iteration, came from like this thing already existing, like first pass last week.

Hey, I just want to make like an Explorer with the different objects. Like here's the doc. Right. And I, and when it showed me this, I'm like, Oh, that's pretty cool. Right. I have not seen anything like this really. Especially with all the objects on it.

But then yesterday, you know, I get to this guy. where it's like, oh, what if I just upload something and can just build whatever I want from scratch? So my prompt this time was, hey, I want a canvas. I've learned that word specifically. Now, that is a key word that when you give it, there's specific React stuff that it'll come back with.

And this is the kind of thing where it blows your mind that you don't have to describe too much for it to show up and be like, Oh, yeah, that's one of these. Right.

But if you don't know how to ask for it, it might take you a while to like prompt your way here. It's getting easier and easier.

But then I was here. And then I was like, Oh, yeah.

Maybe I should just do that with this guy. Right. And then it went from here. And I said, you know what we did on this thing? Let's use this other thing and make a learning canvas out of it. How about that? And I've got to work on stuff, right? But the fact that I could just go like so quickly, step by step, oh man, I've never felt more powerful like ever.

Right. And that's such a key.

Nico:
Like if you can, uh, I'm going to share your screen again, if you're ready. Are you building other stuff?

Chris:
Yeah, I was trying to see how quickly I could actually get this down and running. Let's see.

But yeah, I think if you're not used to being a builder, just look for any kind of pain point on your team or any kind of adoption gap people are struggling with. I don't usually use tickets or projects or services or whatever it is that you're telling me I should try out now.

I'm not used to that thing. So I'm not just going to go into the CRM or whatever piece of software it is and just start clicking around myself because that's not me. That's not who I am.

But when you can develop proof of concept, things like this. And that's where don't overthink it. Just share the links, like get together with the team. Fifteen minute call. Like, hey, guys, this is like right out of Google, right out of cloud, like right out. I've done that. Right.

Don't feel like it needs to be polished and branded and just like, can you help people understand? Visually often is the challenge. of these concepts, whether it's process, you know, just all the different things that people have been making software for, you can contextualize it in ways with quickness that you haven't been able to before, right? And there's immense power that comes from that.

Because usually we're missing the context, right? Like people, like Academy, is a great place to learn about you know go-to-market stuff in a vacuum you throw a sales rep in there they're like i don't know that this doesn't make any sense to me that it does how does it impact my day i don't i don't use these names like this thing is called the deal we used to i don't even know what that is like i have opportunities over here is that what we're talking about right?

And it only takes like, thirty seconds of that feeling for them. Yeah, I'm doing something else. Like, this is not right.

So yeah, now the hype is mobile bubble. What? Like, what?

Nico:
Oh, yeah, watch what happens during the holiday season. And you will get to it tomorrow. But the message about when people do use agents, there is less returns and like, yeah.

Chris:
So hallucinations, all you want, like numbers are showing that surprise when you have an alt, like an ultimate, ultimately intelligent sidekick helping you shop, you get things that you want that you don't like.

Nico:
Yup. And our good friends over at Amazon.

Chris:
I can't. Yeah, we need to talk about it tomorrow for sure. Yeah.

Everybody have a great Wednesday. We'll see you then.

Nico:
See you guys.