1 Year In Consumer AI - AI Girlfriends, P**n, Dating, And The State Of The Market

About a year ago we launched Rizz.ai - for anyone born before 1997, Rizz means Cha"rizz"ma.

It was an app for going on dates with AI girls, and it was potentially the first realtime consumer voice app.

My brother had just built the tech used by OpenAI's realtime voice mode at LiveKit, so we built all the voice tech from scratch.

It was fairly successful, we had a couple viral moments, I engineered a large mostly-male-audience podcast to shout it out. We had made it.

The whole thing blew up like a week later due to a disagreement with a past investor who was still on the cap-table.

Okay. Now what?

We were mad. We had just poured ~100 hrs of nights&weekend time into Rizz, and now we were left with nothing.

Well we had one thing - our realtime voice tech.

Throughout this process, we had a ton of people ask how we built Rizz, so we started thinking "we didn't even know what Rizz meant 2 months ago, should we really be in gen-z consumer?"

This got us thinking about an infra play.

The only real-time voice solutions on the market at the time were Vapi & Retell.

Vapi/Retell were expensive and really difficult to integrate into apps (they've come a long way, still quite expensive).

If they focus(ed) on task-based business use-cases that replace workers, could we focus on task-less consumer use cases? There is a friend drought after all.

As we started running napkin math, we knew we could cut the market price way down. We also already had latency/quality on par with Vapi.

We toyed with building a language learning or career coaching app, but infrastructure kept calling.

By kept calling, I mean my brother was already building the platform and he kept talking about it.

He suggested the name Gabber, and I begrudgingly accepted. (Not because it's a bad name, but because this guy doesn't miss with naming things and it's bullshit).

Now it was time to find a couple customers.

The first one was easy.

Our last company was a metaverse company, and during that time we had partnered with an NFT community to build 🌈THE PORNOVERSE 🌈

We did build it, we were taking over the Metaverse-as-a-Service market, and then the Luna crash happened and all our NFT customers previously tight-knit "communities" evaporated as their NFTs went to 0.

But we had gotten to know the owner of the PORNOVERSE, and turns out he is kind of the man in that industry.

I’m coming back from the gym the day after the Rizz meltdown and I see a thread get revived. My brother had pinged him asking if they were exploring the AI girlfriend space.

😈Neil had gone rogue 👿

He was immediately interested.

I spent more time exploring the market, and turns out almost no one was building consumer AI apps.

Well, almost no one. There were a ton of AI GF apps.

Adult seems to be one of those industries that isn’t price sensitive. We can thank the male libido for that (only time I’ll do that, that has led me astray a time or two *cough cough 6 ft goth girl*)(no regrets, just scars)

My dreams of a 90s Ferrari and a mustache inched closer.

We would be the cheaper, easier to integrate infra platform for building realtime voice ai apps (that were NSFW-compatible).

I told my mom and my priest, and they both gave me their blessing. My dad laughed. I lied about the priest.

Let’s go.

I got about 10 apps building through a mix of cold dms, tweeting, and networking. We were set.

We had started a new C-corp, we’d just brought on another co-founder to own GTM while I was underwater re-becoming a fullstack engineer, and we felt unstoppable.

Given bandwidth constraints, we had to make a choice between working with an AI toy company or going deeper into the AI GF space. Given the momentum with the adult company and some uncertainty around the hardware angle being possible (costs were too high, integration was unclear), we chose the AI gf space.

We inked an enterprise deal, we had more inbound (even a handful outside of AI GF), and we felt like it was time to raise some money (we had raised a small round, but we wanted to accelerate).

We talked to maybe 6 firms.

The top firms loved it but said we needed to have revenue first. Totally reasonable.

A couple mid sized firms said “what if consumer AI ends up with just a handful of winners? Shouldn’t you just build the app?”

One even said “are you really just trying to be cheaper? AI prices always come down”

Hmm.

Okay, well VCs talk to more people than I do, and I’m clearly not doing a good job of convincing them that being the platform is the right move, so I had to stop and think.

They were all right. We were making 4 figures a month. Did we really believe we could turn that into 6-7 figures a month? Would there be a ton of winners? Is being cheaper really all we’re offering?

We started answering these questions one by one.

Revenue - For starters, we learned that voice was not as much of a slam dunk feature as we’d thought. In a call center use case, voice is obviously hugely important as it’s essentially the product. But for consumer apps, it’s more akin to a friend. It should text you, call you, send photos, remember you, text you first sometimes.

We talked to the teams building on us, and they wanted more breadth. They did still want voice, but their users didn’t want to be financially punished for using it, and if the AI did talk, it needed to feel GOOD.

Our intuition was correct. We were $4/hr. We had planned to work toward $1/hr (and had already had a handful of teams ask for that). We also knew we needed to unlock more than just lower latency (which one VC insisted was the key to all this), but better flowing conversation.

Big market - Would all of these roll-up to an app like Character AI?

This was a tricky one too. A few of our customers were getting acquired, which suggested that there is some validity here to a roll-up thesis, but I also started thinking about examples from the internet. A few came to mind:

• Websites

• E-commerce

• Social media

I believe AI personas will be an amalgam of the three of these.

Wix wouldn’t exist if there weren’t still huge demand for creating websites (even though most people go to the same 10 websites). Shopify wouldn’t exist if Amazon was the end all be all to e-commerce. And social media profiles wouldn’t exist if people just wanted to follow their favorite actor (they want to follow their friends too).

To avoid this sounding like cope, I strongly believe AI companions will draw from all of these.

People will likely have a cadre of AI they interact with every day to serve all of those roles—branding, commerce, assistants, friends…lovers 😏

Even in just the toy space it’s obvious that GI Joe and Barbie toys will be powered by different tech stacks. Maybe the building blocks are similar, but Mattel and Hasbro will want to be building at the infra level.

And finally, are we really just competing on price?

That’s an emphatic no, especially now.

Not only have we expanded far beyond voice (bolt-on long-term memory & tool-calling, BYOLLM, voice-cloning, proactive outreach, persona management, context-switching, telephony integration, cross-platform SDKs), but we’ve made it trivial to integrate with 0 backend required.

We work with any OpenAI-compatible LLM, we allow you to implement everything on your frontend, we make tracking the usage of your users trivial.

The goal for us is to make it so you can still differentiate your product on the tech side while not having to maintain virtually any infrastructure.

So that really brings us to today.

Consumer is still spooling up.

There isn’t an obvious “replacement” use case like in the call-center example, but obviously we’re starting to see AI toys, companions, GF apps that are starting to gain traction. Pricing is still a huge factor, as is quality, but both are moving in the right direction. When the market hits is probably pretty closely tied to those two things.

We have a lot of conviction that we’re building something useful to the next 1000 companies building consumer facing apps that incorporate an LLM. Whether it needs to call external functions/do things, build relationships, provide assistance, we have these very stack agnostic building blocks that are priced for consumer, incredibly easy to implement and maintain, and provide the flexibility to build anything without compromise. We have customers that love us to death and a handful more that will love us to death once they finish their integration, so now we continue building for a future we believe in, and we make it easy enough, cheap enough, good enough that that future comes sooner.