Ilan (00:00)
Quick thing before we get started. If you're on a product team listening to this and your team is burning over 10 hours a week on manual status updates, data pulls, or repetitive workflow BS, we're taking three teams through a free workflow audit before the end of January. On the call, we'll show you exactly where you're bleeding time and how to automate it.

If you want us to build it for you after, we can talk. If not, you keep the audit and you can build it yourself. But here's the deal. We're only doing three because these are time consuming to prep for. That's all we have capacity if you want one, go to the link in our show notes right now and grab a spot before they're all gone.

David (00:39)
Hey, Ilan, that's a, that's a cool sweater you got on. Yeah. Yeah. You like it? You like what I got?

Ilan (00:42)
yeah? You too, David. Yeah? Yeah.

Available for purchase on devilwearsproduct.shop

David (00:50)
Yeah, come get your merch.

Hey everybody. This is Prompt and Circumstance. My name's David. And it is the end of 2025. So why don't we round up what are the top 25 AI powered products of the year?

Ilan (00:56)
and I'm Ilan.

David (01:19)
All right, so Ilan and I went through a very scientific and rigorous approach of vibe ranking 25 different AI powered products that ⁓ really were influential for this year and that we think you all ought to know about if you don't.

Ilan (01:36)
our ranking methodology here, the products that we picked, they're not ones that came out in 2025. They're also products that really gained traction or whose functionality was fundamentally changed thanks to AI in the last year.

David (01:54)
Alright, and without further ado, let's get into it.

Ilan (01:57)
Now is the perfect time for me to tell you about our first sponsor for this week, N8N.

n8n is a workflow builder that goes beyond simple automations and goes all the way to allowing you to build production grade agents for your

It allows teams to quickly collaborate on building a prototype, testing with users, and then adding error handling and evals You can try it today by visiting the link in our show notes, and you'll get a two week free trial.

Ilan (02:26)
Starting with number 25,

Outbond is a product we reviewed a few weeks ago and we were really impressed by this automated SDR.

It was able to generate a great list of prospects and create an outreach plan.

This is one of a few AI SDR products that have grown in the last year, including Apollo and Exa Websets, but Outbound really stood out to us for its complete flow and ease of use. So shout out to them.

David (02:57)
Yeah, it's going to be ⁓ very different with go-to-market motions now.

And at number 24, it's Lumo and Duck.ai. Now we haven't covered these ⁓ AI products, but I think it's important for folks to know that they exist. So Lumo is made by Proton, the makers of Proton Mail and Proton Calendar, based out of Switzerland.

And it's important to know about these products because they are very privacy preserving. In an era of AI powered products where your data is used to train who knows how many of these models, these two are ⁓ chat bots basically, know, same thing like with GPT and so forth. ⁓

They're ones that have privacy built into them either by architecture and design or by contract. So it's written out legally that they will not use your data.

So it's important for folks to know that these options are available. ⁓ if privacy is a thing that's really important to them.

Ilan (03:57)
Yeah, and you lose some of the memory functionality that you get out of ⁓ an AI, but the fact that your data is not being used to sell you ads later can be really powerful for a lot of folks.

David (04:13)
Absolutely.

Ilan (04:14)
Coming in at number 23, AI Context Flow.

And this is a product that we reviewed a few weeks ago that allows you to port your context between different AI tools. Context is everything in AI. You saw lots of people talking about how we should stop talking about prompt engineering and really talk about context engineering. And this product is a huge help.

for those of us who are switching between different tools for different use cases and having to re-explain our entire context each time or maybe save a Notion document with your pre-written context.

So what we love about this product is that it allows you to augment your context across different AI products. So you are not locked in to just Gemini or just Copilot, but you can port your context across all of them.

David (05:11)
Context sovereignty.

Ilan (05:13)
That's right.

David (05:14)
Alright.

Up next is SAM or the Segment Anything Model. So this is getting a little bit deeper into the actual technical side of AI. But I thought this was worth a call out because it's so first off, this is a model that's made by Meta and it's used primarily in like audio and video contexts. And what it does is it lets you easily segment out maybe a conversation in a very busy video

or you know the the the train sound that's in the background right I only want to hear

It's almost at the level where it's a kind of fictitious tool ⁓ that you would see on CSI, where they would point at something and magically you would be able to separate it out. So worth a shout out because I think this is going to be very transformative to those who work with this kind of ⁓ problem set, especially coming up in the next year.

Ilan (06:39)
Yeah, I mean, I haven't played around with SAM too much myself, but really excited for the capabilities that this is bringing.

David (06:48)
Very powerful.

Ilan (06:48)
And up next we have Gong

Now Gong is a product that's existed for a long time, but in the last year, they've really integrated AI in a thoughtful way where they can now provide AI summaries across multiple different clients and generate a single context of.

what all of your customers and prospects are telling you about your product or their problems.

Just a couple of weeks ago, they rolled out a new feature which will tell you all of your customers' pain points in a multi-page summary with exact quotes, links to the conversation, where it came from. And so this is something that I'm sure lots of us have been looking for. There are folks out there selling tools that just do that.

out of all of your different customer contacts and if all the calls are recorded in Gong anyway then you might as well go to the source. great job from Gong this year.

David (07:53)
Yeah, very powerful for product discovery, ⁓ not only for product managers, but also even for ⁓ sales folks, right? For sales teams to see what kind of objections are trending and they can then, you know, build their enablement material differently to account for that.

Ilan (08:12)
Absolutely.

David (08:12)
Next up at number 20 is Suno. So Suno makes music and it does a pretty darn good job. In fact, our intro song was generated in Suno.

This is going to be something that anybody in the music and entertainment industry are going to want to know about. ⁓ because it is just that powerful. ⁓ and the fact of the matter is that even with a, with a free trial, you're able to generate a good amount of decent music.

So given its impact, I think it well deserves its position here as number 20.

Ilan (08:52)
Absolutely.

Alright, up next we have Gamma. Gamma is the PowerPoint killer. It's an application that generates presentations from notes or a quick summary. I use it myself and it's super powerful in creating the, let's say, first draft of your presentation with compelling visuals according to the style that you want.

They've actually crossed a hundred million in ARR with only 50 people on staff. So that's two million per person ⁓ on staff Incredibly powerful product for getting you to you know, 60 to 80 percent of a final presentation

David (09:34)
Yeah, nobody likes to live in PowerPoint.

Ilan (09:36)
Absolutely not.

My timelines for creating compelling presentations have dropped significantly thanks to my adoption of Gamma.

David (09:44)
And next up is Linear. So Linear is the Jira killer possibly.

It's something that a lot of product teams use to track work and conversations around what they're doing to their products and perhaps even implementation projects.

Now Linear has a lot of outstanding AI features that say like summarized tickets and you know, really helped push the zero bugs culture. And what's remarkable about Linear is that in 2024, they only had 57 people, right? And just earlier this year in 2025, they raised their series C with a $1.25 billion valuation. they are unicorn with

with under 100 people. So that's really something remarkable.

Ilan (10:32)
Yeah, Linear is one of several products on this list that really embraced the AI native user flows in 2025, something that is definitely going to be expanding in the future.

All right, up next we have Adobe Express and Canva. And both of these tools are also older tools. They've existed for quite a long time, but they've democratized design and have also embraced AI

A regular person who doesn't have a design background, UX background, really leverage these tools with the AI features that they've built in to create compelling workflows, mock-ups, etc. for their products. And there's nothing better than the AI background remover. Saves you a lot of time in any given day.

David (11:28)
Yeah, absolutely. know, gone are the days where you needed to use the lasso tool in Photoshop, know, excruciating ⁓ thing. And even then it was kind of janky. ⁓ So yeah, removing background and also removing object, right? So if something is in the foreground, you can't just clip it out. You got to generate something to fill out where it was. ⁓ And both of these tools have that capability and that's very powerful. ⁓

Ilan (11:34)
Mm-hmm

David (11:58)
in so many different cases.

Ilan (12:00)
Absolutely.

David (12:01)
Next up are a trio of voice dictation tools, Wispr Flow Aqua Voice, and Superwhisper. So these three, I think, are tools that anybody who needs to type ought to consider.

Right.

so gone are the days where you'd be typing out extensively, some long document. these are AI powered, voice dictation tools where you just talk at your computer and it'll type it out for you. And this is really powerful for a couple of reasons. One is that transcription has never been better. Right. mean, sure. Voice transcription has existed for a long time, but now that it's powered by AI models, it's a lot better.

And by being a lot better, not only is it more accurate, it's also ⁓ great at corrections that you would need to make. So if you say one thing and then you go back and say, actually, I meant this, it's not going to transcribe all of the little mistakes that you've done and or the filler words like ums and ahs. It's going to cut those out. And additionally, you can give it context. So you can say, hey, look, whenever I say this word,

it means this and here's how you type it out. You can give it that context so that you don't got to go back and correct some funny name. Like for example Wispr without an E. You can tell it that this is how it's spelled whenever I say this.

Ilan (13:24)
Yeah, and we've shared this stat before, but speaking instead of typing is about three times faster. So you're gonna triple your productivity if you can really adopt these tools in your workflows. Thankfully for me, I have a job where I don't have to type.

David (13:45)
You dictate though, regardless.

Ilan (13:47)
Ha ha ha.

With that, next up is Granola. And David just talked about the power of transcription with AI in the background. And Granola to me is the product this year that really took this to the next level.

We've seen a number of other tools adopt the standard that Granola has set and really use that as a benchmark for what great AI note-taking looks

like.

David (14:14)
Yeah, absolutely. ⁓ you know, everybody should be using AI transcriptions in their meetings these days because that is what builds the context. And we talked about how important context was earlier.

All right, next up are agentic browsers. So I'm talking by Perplexity, Dia by the browser company and Atlas by OpenAI. So these are browsers that you can use like any other. And additionally, you have an AI agent.

That's baked into that browser. So that agent can do stuff on your behalf Whenever you're looking at a page or across multiple tabs so you can summarize stuff across multiple tabs You can tell it to extract data from the current website in a certain form You can tell the summarize the current web page that you're looking at very powerful also

still very new. And so, you know, be a little bit careful with some of the security concerns that, you know, we've talked about this in the past. heads up on that. But I think this is going to be the beginning of a new era of how people will browse on the Internet.

Ilan (15:24)
And you can check out our YouTube short that we created on Comet and how it could help you with QA.

Next up we have Deepseek R1. And we're calling out this model in particular because of how disruptive it was when it was launched early in 2025.

It actually caused a mini market correction by itself it proved that you didn't need the resources of an OpenAI or Anthropic in order to create a cutting edge model.

DeepSeek actually bested the leading models of the time and they've released a number of open papers on their training methodologies. So anyone can go and really understand how these deep reasoning models are created.

David (16:09)
Yep. And also because ⁓ DeepSeek is open, So that's also another thing about democratizing access to these frontier models too.

All right. Up next are video generating models. ⁓ Notably this trio of Sora 2.

Veo 3 and WAN 2.2. right, so ⁓ Sora also showed up as an app.

made by OpenAI. And this really goes to show that, power of these models, at least this generation of models, generating video of how realistic they can be. ⁓ And I think this is really a step change in terms of ⁓ making this now something that is a lot more feasible to be used in order to generate video for ads or, you know, just comedy relief, know, whatever, whatever

might

And the third one, WAN 2.2, is notable. Now this was made by Alibaba, ⁓ and this model in particular is something of note because...

It is one that does video to video very well. And so if any of you have seen example videos where somebody is talking on one side and then on the other side, it shows the resulting video where that is exactly the same as how they're posing and so forth. That is very likely Wan

Ilan (18:01)
for product teams, think about the use case where you want to announce a new product, maybe even internally and generate some hype for an idea that you have.

These tools really give you the ability to have, really picture that show how your product is going to be used out in the market.

All right, next up we have n8n.

n8n is one of many products that have adopted agentic capabilities within workflow builders. But we call it out specifically because its completeness in creating production-ready

flows and agents. It has all of the components needed to handle errors, to handle your evals, and so its power really eclipses the rest. And the fact that it can be self-hosted means that it can truly be part of your ⁓ tech stack for your product.

David (18:59)
Another shout out to open source, right? Like that's really cool that these things are open.

Ilan (19:05)
Absolutely.

Ilan (19:05)
Before we get into our top 10, we'd like to tell you about our other sponsor for this week, Querio Have you ever gotten blocked on answering product questions because you had to wait for data teams to do their analysis first?

Querio's AI agent sits on top of your data stack and allows you to ask natural language questions and get direct answers and beautiful visuals to tell you what's going on in your product. If you want to try it today, you can go to their site, querio.ai, and tell them David and Ilan sent you for two months free of their Explore product. That's a thousand dollar value for our listeners.

David (19:42)
All right, we're entering the top 10. So at number 10 is ElevenLabs

For those who don't know it, this is a company that has a product that will generate very good sounding audio in terms of voices. Right? So you can have ⁓ somebody announcing something that will sound like a sports announcer, for example, or maybe like a movie announcer, trailer announcer. You can also use it to ⁓ duplicate a voice. Right? And so if you provide enough of a recording of a voice, probably your own, you can then use

that to generate subsequent things to be said with your voice.

So this is really powerful. You know, I, wasn't that long ago that I think it was a mission impossible that Tom Cruise had somebody say something quickly and then suddenly he was able to talk with that person's voice. I mean, this is the future. And so ElevenLabs, like they existed before 2025, but I think in 2025, like they really made a leap forward in terms of the quality of the audio that they generate.

Ilan (20:43)
You

Not only that, but I think they showed the frontier labs what could be possible with their models and forced them to try to catch up. And we see some of this functionality from Google now ⁓ in AI Studio, but ElevenLabs is really the forerunner of those.

David (21:20)
Absolutely.

Ilan (21:21)
All coming in at number nine, we have Nano Banana. And this is also known as Gemini Flash 2.5 image. But.

The reason that this specific model is called out here is that it completely changed what our expectations were for image generation, especially with maintaining aspects of the image while changing the rest of it. So Banana, if you wanted and my face the same, but change our background and you went to chat GPT,

It would also change our faces a little bit.

But with NanoBanana, you can keep the identity of the subject while changing everything around them, or keep the scene and change the identity of the subject,

David (22:11)
Shoutout to whoever at Google decided to call it Nano Banana because that is much more memorable and accessible to the public than Gemini 2.5 flash image.

Ilan (22:15)
Hahaha

David (22:22)
All right, number eight, Perplexity All right, so we talked about a PowerPoint killer. We talked about a Jira killer. This one is a Google search killer.

So Perplexity, their original product, which again, you know, it was launched before 2025, but it really made a big push in 2025 for changing how people would search online.

this is actually something that, I use a lot personally to replace what would have been Google searches. And it's so much faster for me to get to the answer that I'm looking for. So shout out to Perplexity on that.

Ilan (22:55)
A funny story from this year. Early this year, Perplexity offered to buy Chrome from Google for $34.5 billion. So they're a little tongue in cheek as well on top of everything else that they do.

David (23:09)
That was a great marketing ploy.

Ilan (23:10)
Coming in at number seven we have the trio of Claude code Codex and Antigravity and we're calling these the agentic coding Trinity these are three models that are foundational for coding Claude code really led the way early this year but

OpenAI followed quickly with Codex and recently Google followed with Antigravity. And the great thing about these coding models is that they can be brought into different IDEs that software developers might be using.

On the product side, we're seeing product managers really extend the capabilities of these tools and use, for example, Claude Code as a generator for PRDs or other research materials that you may need to share within your team

David (24:04)
Wait, so you said Claude code does that, but did you mean Claude? Okay. All right. Okay. Just leave that in.

Ilan (24:08)
No, Claude Code

Let's just leave that in.

David (24:16)
All right. At number six, we have a duo. We've got Cursor and Windsurf. Right. So these are the IDE as an agent.

platforms that are really transforming how software engineers do their thing. Right? So ⁓ being able to use natural language in order to ⁓ either make a code change or propose a code change or to maybe even generate documentation. Right? So this is a big game changer for anybody in the software industry, but especially for those who are a bit more technical.

Ilan (24:52)
Yeah, absolutely, David. These are incredible tools and we rank them above the agents from the foundation labs because they really thought about an AI native experience for a developer and what is a developer looking for in an IDE and then created that product.

and fun story from earlier this year, Windsurf went through a reverse acquihire by Google. This was a really incredible series of events. So Windsurf entered into an agreement with OpenAI to be acquired for $3 billion. Then the exclusivity on that agreement expired because I guess the negotiations took too long. And the next day, Google hired the two founders and signed a perpetual license agreement

with windsurf. But the story doesn't end there because the following week Cognition, which is the company that makes Devon AI, then acquired windsurf for under a billion dollars. So they got all of the IP from windsurf while Google only has the license to use it.

David (25:58)
That is quite the drama. You know, I wonder whether there's going to be a Netflix documentary about that.

Ilan (26:03)
One day I can't wait.

Coming in at number five on the list are Google's Gemini and Microsoft's Copilot. And these are specifically called out because of their integration into the workplace software that these companies create. So Gemini integrated into Google Workspace.

and Copilot integrated into the Microsoft Office suite. And that means that these products are brought to the hundreds of millions of users across both of these systems. In fact, Google estimates that they have up to three billion users of Google Workspace. We might quibble with that number, but... ⁓

David (26:44)
Might be a lot of spammers on Gmail, yeah?

Ilan (26:47)
But in general, it does show that you can bring AI tools to the people who need them, but aren't aware of their capabilities by displaying them directly where those people are doing their work.

David (27:00)
Yeah. I mean, you know, any product manager would be thinking about distribution, about, you know, meeting their market where the market is. Right. And so the fact of the matter is that being able to instantly reach those hundreds of millions of users, is remarkably powerful for, ⁓ Gemini and Copilot.

All right, number four is Claude and GPT. So these two foundation models, they had so many updates this year.

So for example, the most recent update to Claude by Anthropic was the Opus model up to 4.5 and then from OpenAI's GPT going up to 5.2. So just nonstop progress on these models. And we have this at number four because of just how foundational they are. Well, they're called foundation models, but you can interact with them sure as chatbots.

⁓ but then you can also use them in other products and there's lots of other products that would use these models. And we say cloud and GPT because it doesn't have to be Opus or 5.2. It could be Sonnet 4.5, for example, or it could be, ⁓ GPT 4.0 mini even, right? So these two models, ⁓ you know, being just so foundational to so much of the AI impact, ⁓ definitely is a top five contender.

Ilan (28:28)
Yeah, the only reason they might have been bumped down the list a little is because they've been around. These are the models that launched us into understanding what is ⁓ an LLM. So they've been around for a while, but the continued innovation all through this year is really impactful. And many, in fact, I would say almost all of the previous products on the list are only possible because of these foundation models.

David (28:55)
Exactly.

Ilan (28:56)
All right, we're up to the top three. So drum roll.

David (29:01)
hahahaha

Ilan (29:04)
Coming in at number three are the foursome of Lovable, Bolt, Replit and V0. These are the original four accessible Vibe Code tools, most of which launched at the very end of 2024 or beginning of 2025 and really gained steam in early 2025.

These really democratized coding, it changed the expectation for product teams everywhere from, hey, I'm gonna write a six page document explaining my product to I'm gonna write a one pager and show you a prototype of what's in my head so that you can see what the user would be doing.

and these have unlocked founders across multiple different industries to create their software products. So the impact that these products have had in how we all work is really hard to overstate.

David (30:05)
Yeah, I mean, the, the, the fact of the matter is that, uh, you know, by transforming how software could get built and by, you know, getting built, mean, the discovery and design aspect of it. Um, you know, that is going to have just such a wide radius of

And most notably it was because of the launch of these vibe coding products that this podcast got created.

Ilan (30:34)
Yeah, I mean, it was just so exciting for us that this was possible, that we had to try them out and bring them to the world.

David (30:41)
Yeah, absolutely. So thank you to ⁓ those, to the, the foursome.

Ilan (30:46)
And shameless pitch, you can go back in our logs and see our reviews of the four tools and our track meet where we pit them all against each other and crowned a winner.

David (30:57)
Exciting times.

All right, number two. At number two, we've got Notebook LM. why is it at number two? It's here because it is going to be impactful for anybody who wants to learn anything, right? So you don't have to be in the software industry. You could be in a completely unrelated industry and you could be a student. You could do, you could...

This is anything for anybody who wants to learn a thing. So the power of this is that you can give it any document ⁓ and you can even tell it to go find sources online And with that data,

it is going to be able to, of course, let you chat with that information, but also you would be able to generate an audio overview podcast. So if you wanted to learn about ⁓ quantum computing 101, right, you dump all of that into there and now you have a personalized podcast because you can tune it to exactly highlight the things that you want to learn. And then you can just listen to it during your commute. What a great way for people to learn stuff, right?

The other things that it can generate would be like videos and presentations and so forth lots of power there and So this is something that we think is going to be very transformative across

the entire globe and so therefore number two.

Ilan (32:16)
I'll give an example of where I've seen this impacting in my own work where I introduced this tool to the company and one of our teams is just creating training materials for other internal teams to learn how to do their jobs as we bring on temps or these kinds of things. And this cut down the time.

that they needed to create training materials because they just grab all the context from wherever it exists across the company, dump it into Notebook LM, and then they can create the videos that they needed to create. They create the audio overviews, which is not something they had before, and they can create detailed summaries that they can then share with people that are tuned to exactly those people's needs.

David (33:02)
Yeah. So absolutely. Shoutout to Raiza Martin, who was the senior product manager at Google, who led the creation of this whole thing, which by the way was actually launched in 2024, but it gained a lot of the traction in 25.

Ilan (33:20)
Well, this has been a really exciting list.

The top 24 products of 2025 something game-changing about them. But this brings us to number one on the list. And number a

Claude 4. Now, why did we pick this specific model?

when we were already calling out ⁓ Claude in a previous ranking. It's because Claude 4 stepped up the level of what we thought was possible from a model, especially around coding,

When Claude 4 came out, it transformed from, hey, AI is like a cute helper for ⁓ agentic coding. You know, it'll finish lines for you. It can even generate code for you, but it'll probably have like a lot of mistakes and, you know, somebody will need to clean up behind it. To, ⁓ AI can be your coder. It can actually be your junior developer. And now your job is to work on the architecture.

David (34:26)
Mm-hmm.

Ilan (34:30)
or understand how to connect these pieces. There's a funny stat from Lovable's founder, Anton Osika, that when Lovable incorporated Claude 4 into their product, they saw errors go down by 25 % and speed go up by 40%. So it was faster and less error prone than the models that came before.

David (34:53)
Huge win.

Ilan (34:54)
It's completely changed how we think about writing code. None of the coding products that we talked about before would have existed with the same power that they have today without Claude 4 directly, or at least Claude 4 showing the way on what is possible.

David (35:13)
Absolutely. mean, and think about the compounding effect of that, right? So being the, pivotal change that, ⁓ would result in all of these other software companies, like all these unicorns in the future, make, you know, being the one that makes that possible. ⁓ so definitely worthy of first place in our countdown.

Ilan (35:36)
And Anthropic on the back of this and their focus on enterprise capabilities, they actually went from less than a billion in revenue to over $7 billion in revenue just this year. So, you we talked about amazing growth from some of the companies earlier on this list, but Anthropic, which has been around for quite a few more years, they also had incredible meteoric rise in 2025.

David (36:00)
So congratulations to Anthropic for making the product that landed as number one. We were going to have Dario Amadei on here, but we just ran out of time, so maybe next time.

Ilan (36:13)
That's right, Dario, sorry to have to shaft you on this one.

David (36:16)
All right, so that is our top 25 of 2025. Leave us a like or a comment and we're happy to explore more AI products. So there's any ones that you want us to have a look at, by all means, let us know.

Ilan (36:33)
That's right. I'm sure we made mistakes. I'm sure we left out your favorite product. So please @ us. Let us know. What did we forget? Send us a message and we'll definitely look at it. Thank you everyone.

David (36:43)
All right,

take care.