6 ms·

AI Market Clarity

111 points by todsacerdoti 1y ago

chancemehmu 1y ago
I think the companion market is probably the most overlooked market in AI. Character AI is the most used AI app after ChatGPT worldwide, and generally what all the teens associate AI with. I am seeing companies like Dippy AI (character ai competitor for gen-z) scale from 0->10M+ users rapidly without many people noticing.
- deadbabe 1y ago
  The churn is brutal. Once people have gotten over the novelty of an AI companion they quickly grow disatisified with the limitations and leave. It’s a lot like a game that people play with for a bit then leave when they’ve done everything they can do.
  - chancemehmu 1y ago
    That's not what the numbers coming from some of these companies say — for example, chai app, publicly shares that their W52 retention is 20%. So, 20% of users who send a message just retain forever. Source: https://www.chai-research.com/chai_roadmap_2025.pdf https://www.chai-research.com/chai_roadmap_2025.pdf
    - riskassessment 1y ago
      hardly forever. Given the age of the company you're citing, they can only estimate retention out to 1 year.
    - codingdave 1y ago
      Their charts measure 50 weeks, not forever. So an 80% yearly churn is not exactly a good statistic. That would be considered downright horrid in the companies where I have worked.
      - wagwang 1y ago
        It's pretty amazing considering the product is literally calling openai with a prompt.
    - dingnuts 1y ago
      none of these companies have existed long enough to make claims about users staying around "forever"
    - fxtentacle 1y ago
      That chart says: 50% leave within 1 week 75% leave within 5-10 weeks 85% leave within a year
    - jcranmer 1y ago
      If I'm reading that right, 60% of their users give up after the first week, half of those who stick with it give up after another month or two, and then there appears to be continued decline afterwards. Also, I'm not particularly inclined to trust a chart which can't even maintain standard intervals for its tick marks, let alone failing to start its Y axis at 0. Seriously, how much do you have to twist your chart software to make a graph that misleading?
      - fxtentacle 1y ago
        Excel defaults to automatically adjusting both axis ranges. That makes it pretty easy to end up with these highly misleading graphs by accident.
        jcranmer 1y ago
        I can understand not starting at 0; it's the varying tick spacing that has me confused.
    - nonameiguess 1y ago
      This is a good place for a brief intro to statistical literacy. I don't fault the report per se because they're measuring what is possible to measure, and framing it in the best possible light since the entire point of this slide deck is a pitch to investors. But it's riddled with nonsense that mirrors the challenges of social science and causes of the replication crisis we bemoan in non-profit research but seem to blink zero eyes over when it's industrial research. Is user retention an ergodic process? Clearly no. From economic first principles, we know elasticity of demand is going to depend upon substitutability and disposable income of the user base. Both of these can and do change over time. From empirical results elsewhere, we can simply look at products that have existed for more than a single year and observe that user retention rates in 1995 did not always match retention rates for the same product in 2002. Here we get measurement of a single cohort that has actually reached 50 weeks. Are they representative of a typical cohort? We have no way to know. And again, fair, this is the best the company can possibly do. They can't create data that doesn't exist and they can't draw conclusions that have no hard evidence for or against. But you don't need to draw unwarranted conclusions, either. It's also 15% at 50 weeks retention in that first cohort, not 20% at 52 weeks. Does that mean those 15% of users will stick around forever? We have no idea. Sort of. We actually know for certain they won't because they're mortal and will die. Setting that aside, might they at least become lifelong users? I suppose it's possible, but even real-world friendships rarely last a lifetime. Will the next cohort at least still show 15% at 50 weeks? We yet again have no idea.
    - dontlikeyoueith 1y ago
      Are you trying to spin 80% yoy churn as good? Did you get your MBA from the back of a mail order catalog?
    - deadbabe 1y ago
      The numbers are saying exactly what I said.
  - zachthewf 1y ago
    Maybe. FWIW the majority of the apps that have taken off are basically AI girlfriends/AI boyfriends. Using these is fundamentally a shameful act which probably deflates retention. There may be high retention use cases that find ways to serve user needs without embarrassing them.
    - Retr0id 1y ago
      If someone installed the app in the first place, aren't they already over that hurdle?
      - zachthewf 1y ago
        They crossed the hurdle once but are less willing to do so repeatedly (so they churn)
- d-moon 1y ago
  I'm just waiting to see how they monetize their userbase. Last time I checked they made $1/user a year(~20m active users/ 20m revenue). It's predominately a whale market where traffic is driven by power users. You have mobile gacha games like Genshin Impact that make ~$10/user a MONTH (~5m active users/ 50m in monthly revenue)...
- evantbyrne 1y ago
  So-called "companion" AIs are usually overlooked because they are clearly predatory. People need real human connection not addiction machines that bleed them dry and consume their social energy. Making a buck on someone else's misery is no way to live!
- Der_Einzige 1y ago
  Being knowledgeable about the NSFW aspects of AI systems just adds an additional 100K per year to an already large prospective offer. Civit.ai, the primary competitor to huggingface (i.e. github for AI) for diffusion models, might as well be a brothel with how horny it is. Over half of all "low rank adapters/Loras" every created are most likely NSFW in nature.
bmau5 1y ago
Customer Service keeps being touted as a success story of LLMs, but as a business owner and user I'd be very curious to see if it actually improves any UX CS metrics (NPS, CSAT) - as my experiences with AI chatbots are almost always negative. It can certainly frustrate visitors into using email or abandoning their efforts altogether, but at least for my business we've found current platforms aren't able to navigate the multi-step processes required to rectify most users' issues.
- kjkjadksj 1y ago
  Whenever I get saddled with one of those I spam “human human human” in the chat until I get the support agent from somewhere in asia. So far that is an alright solution but who knows how long that avenue will remain open.
- morkalork 1y ago
  You sound like an honest business owner and that's great, however I bet other businesses like monopolistic telecoms, gyms and media subscriptions are all too happy to roll out these poor quality bots.
  - bee_rider 1y ago
    I actually don’t get why those companies are bothering with language-model based chatbots. They already have customer service that can’t do anything. Why do you need a language model? Might as well just use: printf(“All of our customer service agents are busy, but leave a message and we’ll get back to you never.”)
    - mindslight 1y ago
      It's both less expensive to run and engages users to waste their time better so they don't seek out any remaining avenues. I recently had a package misdelivered by SpeedX. "Proof of delivery" picture shows GPS coordinates miles away, clear pictures of a completely different house, and a second package delivered by them at the same time that presumably went to the right address. Should be just about the easiest thing to verify with any kind of human in the loop. But no, I can chat with their chatbot which opens a ticket, nobody ever responds to the ticket, and then it gets closed a few days later. I tried this a few times, asking for updates on the ticket, etc. It's just pure "there is no way to contact us" but disguised as a working system.
Dowwie 1y ago
Elad is an investor of these companies, not an independent researcher
- BowBun 1y ago
  I find it really funny to read "I invested early in X" and then shortly later read "X has won this category of business" from the same author. Gives the same vibe as those 'popular alternatives to X' blog posts written by one of the vendors, placing their offering as #1 and picking mediocre competitors to compare themselves against.
  - PaulHoule 1y ago
    This is why people should be careful what they upvote. Look at /new and you'll see people are really "flooding the zone" with bad posts about AI, often self-serving posts. Flag is a little harsh for all but the worse abuses, but since there is not a downvote button it is an act of resistance to look at /new as often as you can and upvote a few articles about non-AI topics that aren't complete trash.
  - eladgil 1y ago
    Hmm, which verticals did I do this in?
    - rohitpaulk 1y ago
      I noticed this with Harvey / Legal.
- eladgil 1y ago
  The majority of companies listed I did not invest in, as well as in some cases areas I have not invested in at all I am actively involved in AI and have been for a few years, so this both gives me insights and obviously conflicts. My hope is to write things that are useful vs just shilling as I would lose credibility otherwise
boringg 1y ago
That link title is an oxymoron.
samtp 1y ago
I would love to hear some clear use cases that people are actually using AI agents for in production on non-trivial matters. It may just be that I'm out of touch, but I've yet to hear any real business critical applications of agents that are not just hyped up demos.
vaenaes 1y ago
Hoping for post-AI clarity.
- debo_ 1y ago
  A tough nut to bust.
diamond559 1y ago
They are successful in that they attract lot of money to speculate on their privately traded stocks, not from paying customers.
benterix 1y ago
It's a nice advertisement with quite a lot of opinions and wishful thinking presented as facts.
District5524 1y ago
Although there is a part on the use of AI in the legal market, but it is neither comprehensive, nor informative. A much better view of the law-specific state of the art models is available at https://www.vals.ai/vlair https://www.vals.ai/vlair (although registration is needed to access the report). That benchmark in the report clearly shows that there is not much added value or moat in the custom training of Harvey (costing supposedly ~5M$), while a Llama-based model can achieve quite similar performance. Of course, no benchmark is perfect but it is still more informative than this blogpost, plus all the linked websites of the "legal" products linked that fail to give you the slightest idea of how they work. Besides claiming to be be disruptive...
- throwawayoldie 1y ago
  I think what the public really wants to know about Harvey is, is it named for Harvey Birdman, Attorney-At-Law?
  - District5524 1y ago
    that would be fun, but what i've heard that it's named after Harvey Specter, from Suits. But that's not something they'll officially ever deny or confirm...
benreesman 1y ago
Eh, I'm not sure I buy it on the frontier labs being all locked up. The Chinese stuff is fucking great, and it's improving really really fast. Labs like OpenAI? I'm skeptical they could build their current product lineup from scratch, and why should that be surprising? The people who did it left! Anthropic has a pretty clear headlock on the absolute apex of brute-force Chinchilla scaling with Opus 4, but they've started being dodgy enough with it that I finally set up Together and OpenRouter and put their shit on Bedrock and Vertex: I'm over it with the "well, it lies about having mocked the whole module today". And K2? Whew, that's really refreshing. It's different, hard to compare, but I've been running it for a day or so where Sonnet 4 would normally go and I'm not having any trouble. It's clearly the product of a different culture but in a cool way, and on the stuff that counts? Chinese people speak C++ just fine. I'm pretty sure at this point that infinite NVIDIA is one of the worst handicaps our AI giants could have gotten, and that needing to work within and around export restrictions and gimped cards is the best thing that ever happened to Hangzhou. I see no reason why the take in the article is any more plausible than an alternative future where the US has about as much market share in frontier LLMs as Europe does in consumer Internet.
tempusalaria 1y ago
The term agent is just way overloaded. This guy defines it completely differently the the big labs, and I’ve seen half a dozen different definitions in the last few months. In the long run the definition used by OpenAI, Anthropic et al will win out so can we just all switch to that?
ariwilson 1y ago
No mention of product prototyping products e.g. vercel v0 or Firebase Studio. Those to me seem like clear 0->1 wins, especially vercel with their hosting/marketplace integrating well with AI-driven development.
panabee 1y ago
The author is a respected voice in tech and a good proxy of investor mindset, but the LLM claims are wrong. They are not only unsupported by recent research trends and general patterns in ML and computing, but also by emerging developments in China, which the post even mentions. Nonetheless, the post is thoughtful and helpful for calibrating investor sentiment.
- ariwilson 1y ago
  What is wrong about their claims?
iddan 1y ago
We are building some exciting sales AI tech in Closer: basically the cursor for sales people. https://closer.so https://closer.so

111 points by todsacerdoti 1y ago

I think the companion market is probably the most overlooked market in AI. Character AI is the most used AI app after ChatGPT worldwide, and generally what all the teens associate AI with. I am seeing companies like Dippy AI (character ai competitor for gen-z) scale from 0->10M+ users rapidly without many people noticing.

deadbabe 1y ago
The churn is brutal. Once people have gotten over the novelty of an AI companion they quickly grow disatisified with the limitations and leave. It’s a lot like a game that people play with for a bit then leave when they’ve done everything they can do.
- chancemehmu 1y ago
  That's not what the numbers coming from some of these companies say — for example, chai app, publicly shares that their W52 retention is 20%. So, 20% of users who send a message just retain forever. Source: https://www.chai-research.com/chai_roadmap_2025.pdf https://www.chai-research.com/chai_roadmap_2025.pdf
  - riskassessment 1y ago
    hardly forever. Given the age of the company you're citing, they can only estimate retention out to 1 year.
  - codingdave 1y ago
    Their charts measure 50 weeks, not forever. So an 80% yearly churn is not exactly a good statistic. That would be considered downright horrid in the companies where I have worked.
    - wagwang 1y ago
      It's pretty amazing considering the product is literally calling openai with a prompt.
  - dingnuts 1y ago
    none of these companies have existed long enough to make claims about users staying around "forever"
  - fxtentacle 1y ago
    That chart says: 50% leave within 1 week 75% leave within 5-10 weeks 85% leave within a year
  - jcranmer 1y ago
    If I'm reading that right, 60% of their users give up after the first week, half of those who stick with it give up after another month or two, and then there appears to be continued decline afterwards. Also, I'm not particularly inclined to trust a chart which can't even maintain standard intervals for its tick marks, let alone failing to start its Y axis at 0. Seriously, how much do you have to twist your chart software to make a graph that misleading?
    - fxtentacle 1y ago
      Excel defaults to automatically adjusting both axis ranges. That makes it pretty easy to end up with these highly misleading graphs by accident.
      - jcranmer 1y ago
        I can understand not starting at 0; it's the varying tick spacing that has me confused.
  - nonameiguess 1y ago
    This is a good place for a brief intro to statistical literacy. I don't fault the report per se because they're measuring what is possible to measure, and framing it in the best possible light since the entire point of this slide deck is a pitch to investors. But it's riddled with nonsense that mirrors the challenges of social science and causes of the replication crisis we bemoan in non-profit research but seem to blink zero eyes over when it's industrial research. Is user retention an ergodic process? Clearly no. From economic first principles, we know elasticity of demand is going to depend upon substitutability and disposable income of the user base. Both of these can and do change over time. From empirical results elsewhere, we can simply look at products that have existed for more than a single year and observe that user retention rates in 1995 did not always match retention rates for the same product in 2002. Here we get measurement of a single cohort that has actually reached 50 weeks. Are they representative of a typical cohort? We have no way to know. And again, fair, this is the best the company can possibly do. They can't create data that doesn't exist and they can't draw conclusions that have no hard evidence for or against. But you don't need to draw unwarranted conclusions, either. It's also 15% at 50 weeks retention in that first cohort, not 20% at 52 weeks. Does that mean those 15% of users will stick around forever? We have no idea. Sort of. We actually know for certain they won't because they're mortal and will die. Setting that aside, might they at least become lifelong users? I suppose it's possible, but even real-world friendships rarely last a lifetime. Will the next cohort at least still show 15% at 50 weeks? We yet again have no idea.
  - dontlikeyoueith 1y ago
    Are you trying to spin 80% yoy churn as good? Did you get your MBA from the back of a mail order catalog?
  - deadbabe 1y ago
    The numbers are saying exactly what I said.
- zachthewf 1y ago
  Maybe. FWIW the majority of the apps that have taken off are basically AI girlfriends/AI boyfriends. Using these is fundamentally a shameful act which probably deflates retention. There may be high retention use cases that find ways to serve user needs without embarrassing them.
  - Retr0id 1y ago
    If someone installed the app in the first place, aren't they already over that hurdle?
    - zachthewf 1y ago
      They crossed the hurdle once but are less willing to do so repeatedly (so they churn)
d-moon 1y ago
I'm just waiting to see how they monetize their userbase. Last time I checked they made $1/user a year(~20m active users/ 20m revenue). It's predominately a whale market where traffic is driven by power users. You have mobile gacha games like Genshin Impact that make ~$10/user a MONTH (~5m active users/ 50m in monthly revenue)...
evantbyrne 1y ago
So-called "companion" AIs are usually overlooked because they are clearly predatory. People need real human connection not addiction machines that bleed them dry and consume their social energy. Making a buck on someone else's misery is no way to live!
Der_Einzige 1y ago
Being knowledgeable about the NSFW aspects of AI systems just adds an additional 100K per year to an already large prospective offer. Civit.ai, the primary competitor to huggingface (i.e. github for AI) for diffusion models, might as well be a brothel with how horny it is. Over half of all "low rank adapters/Loras" every created are most likely NSFW in nature.

bmau5 1y ago

Customer Service keeps being touted as a success story of LLMs, but as a business owner and user I'd be very curious to see if it actually improves any UX CS metrics (NPS, CSAT) - as my experiences with AI chatbots are almost always negative. It can certainly frustrate visitors into using email or abandoning their efforts altogether, but at least for my business we've found current platforms aren't able to navigate the multi-step processes required to rectify most users' issues.

kjkjadksj 1y ago
Whenever I get saddled with one of those I spam “human human human” in the chat until I get the support agent from somewhere in asia. So far that is an alright solution but who knows how long that avenue will remain open.
morkalork 1y ago
You sound like an honest business owner and that's great, however I bet other businesses like monopolistic telecoms, gyms and media subscriptions are all too happy to roll out these poor quality bots.
- bee_rider 1y ago
  I actually don’t get why those companies are bothering with language-model based chatbots. They already have customer service that can’t do anything. Why do you need a language model? Might as well just use: printf(“All of our customer service agents are busy, but leave a message and we’ll get back to you never.”)
  - mindslight 1y ago
    It's both less expensive to run and engages users to waste their time better so they don't seek out any remaining avenues. I recently had a package misdelivered by SpeedX. "Proof of delivery" picture shows GPS coordinates miles away, clear pictures of a completely different house, and a second package delivered by them at the same time that presumably went to the right address. Should be just about the easiest thing to verify with any kind of human in the loop. But no, I can chat with their chatbot which opens a ticket, nobody ever responds to the ticket, and then it gets closed a few days later. I tried this a few times, asking for updates on the ticket, etc. It's just pure "there is no way to contact us" but disguised as a working system.

Dowwie 1y ago

Elad is an investor of these companies, not an independent researcher

BowBun 1y ago
I find it really funny to read "I invested early in X" and then shortly later read "X has won this category of business" from the same author. Gives the same vibe as those 'popular alternatives to X' blog posts written by one of the vendors, placing their offering as #1 and picking mediocre competitors to compare themselves against.
- PaulHoule 1y ago
  This is why people should be careful what they upvote. Look at /new and you'll see people are really "flooding the zone" with bad posts about AI, often self-serving posts. Flag is a little harsh for all but the worse abuses, but since there is not a downvote button it is an act of resistance to look at /new as often as you can and upvote a few articles about non-AI topics that aren't complete trash.
- eladgil 1y ago
  Hmm, which verticals did I do this in?
  - rohitpaulk 1y ago
    I noticed this with Harvey / Legal.
eladgil 1y ago
The majority of companies listed I did not invest in, as well as in some cases areas I have not invested in at all I am actively involved in AI and have been for a few years, so this both gives me insights and obviously conflicts. My hope is to write things that are useful vs just shilling as I would lose credibility otherwise

boringg 1y ago

That link title is an oxymoron.

samtp 1y ago

I would love to hear some clear use cases that people are actually using AI agents for in production on non-trivial matters. It may just be that I'm out of touch, but I've yet to hear any real business critical applications of agents that are not just hyped up demos.

vaenaes 1y ago

Hoping for post-AI clarity.

debo_ 1y ago
A tough nut to bust.

diamond559 1y ago

They are successful in that they attract lot of money to speculate on their privately traded stocks, not from paying customers.

benterix 1y ago

It's a nice advertisement with quite a lot of opinions and wishful thinking presented as facts.

District5524 1y ago

Although there is a part on the use of AI in the legal market, but it is neither comprehensive, nor informative. A much better view of the law-specific state of the art models is available at https://www.vals.ai/vlair https://www.vals.ai/vlair (although registration is needed to access the report). That benchmark in the report clearly shows that there is not much added value or moat in the custom training of Harvey (costing supposedly ~5M$), while a Llama-based model can achieve quite similar performance. Of course, no benchmark is perfect but it is still more informative than this blogpost, plus all the linked websites of the "legal" products linked that fail to give you the slightest idea of how they work. Besides claiming to be be disruptive...

throwawayoldie 1y ago
I think what the public really wants to know about Harvey is, is it named for Harvey Birdman, Attorney-At-Law?
- District5524 1y ago
  that would be fun, but what i've heard that it's named after Harvey Specter, from Suits. But that's not something they'll officially ever deny or confirm...

benreesman 1y ago

Eh, I'm not sure I buy it on the frontier labs being all locked up. The Chinese stuff is fucking great, and it's improving really really fast. Labs like OpenAI? I'm skeptical they could build their current product lineup from scratch, and why should that be surprising? The people who did it left! Anthropic has a pretty clear headlock on the absolute apex of brute-force Chinchilla scaling with Opus 4, but they've started being dodgy enough with it that I finally set up Together and OpenRouter and put their shit on Bedrock and Vertex: I'm over it with the "well, it lies about having mocked the whole module today". And K2? Whew, that's really refreshing. It's different, hard to compare, but I've been running it for a day or so where Sonnet 4 would normally go and I'm not having any trouble. It's clearly the product of a different culture but in a cool way, and on the stuff that counts? Chinese people speak C++ just fine. I'm pretty sure at this point that infinite NVIDIA is one of the worst handicaps our AI giants could have gotten, and that needing to work within and around export restrictions and gimped cards is the best thing that ever happened to Hangzhou. I see no reason why the take in the article is any more plausible than an alternative future where the US has about as much market share in frontier LLMs as Europe does in consumer Internet.

tempusalaria 1y ago

The term agent is just way overloaded. This guy defines it completely differently the the big labs, and I’ve seen half a dozen different definitions in the last few months. In the long run the definition used by OpenAI, Anthropic et al will win out so can we just all switch to that?

ariwilson 1y ago

No mention of product prototyping products e.g. vercel v0 or Firebase Studio. Those to me seem like clear 0->1 wins, especially vercel with their hosting/marketplace integrating well with AI-driven development.

panabee 1y ago

The author is a respected voice in tech and a good proxy of investor mindset, but the LLM claims are wrong. They are not only unsupported by recent research trends and general patterns in ML and computing, but also by emerging developments in China, which the post even mentions. Nonetheless, the post is thoughtful and helpful for calibrating investor sentiment.

ariwilson 1y ago
What is wrong about their claims?

iddan 1y ago

We are building some exciting sales AI tech in Closer: basically the cursor for sales people. https://closer.so https://closer.so