I said this years ago, LLMs are a commodity (or were becoming one at the time). They are dime a dozen. Even the frontier ones. OpenAI and Anthropic have no moat.
No moat and competition is good for consumers though.
Please don't post snark on HN. The guidelines make it clear we're trying for something better here. <a href="https://news.ycombinator.com/newsguidelines.html">https://news.ycombinator.com/newsguidelines.html
No moat and competition isn't always preferable over a competitive oligopoly (with some differentiation). The former ends up with politicians intervening way more like with solar, steel, agriculture ....
My 0.02 on this is that for a given level of capability, yes, there is no moat. What’s frontier one day quickly becomes a commodity almost too cheap to meter within a year or two.
But there is always a use case for frontier models, even if they’re quite a bit more expensive. The set of things you can profitably do with better intelligence than everyone else is unbounded.
So yes, the number of tokens that get pushed through commodity models for very cheap will continue to grow, but so will the tokens for expensive frontier models. We’ll never run out of things to do with the latest geniuses who are twice as smart as last year’s geniuses.
The moat is the engine that builds and sells the latest geniuses, especially the compute. And no one will have more compute than American labs for the next 5-10 years.
They’ve been saying that for the last 18 months at least. They’re always just a few months behind…
I picked double out of a hat, obviously there’s some intelligence gap that’s too small for a price premium beyond a certain point, but it seems like OpenAI and Anthropic’s revenue keeps climbing, so I don’t think we’ve hit it.
well, it is true. there is a price ceiling for intelligence, there were news about how adoption of fable was lower than expected, only hitting 11% of enterprise spending for customers with anthropic, iirc 26% for Sol with openai customers.
in any case, I've been using open and closed models since sonnet 4, i remember when the best I could get was qwen 3 480b coder, you can definitely feel the gap closing going from that and GLM 4.5, to GLM 5.3, DeepSeek Flash V4.1, Kimi K3 etc, it's reached the point where i wish I had V4.1 at work, it's faster and bullshits me less when I use it in my personal projects. And I have unlimited access to fable 5.1
The gap really is closing, though. I don't remember the source, but there was a publication recently stating that it's 4 months at this point. So, 4 months where you can charge premium prices for the small subset of tasks where only a frontier model will do. I have to think the big AI duopoly is at an inflection point where most of their customers haven't just yet realized that they're being fleeced.
> But there is always a use case for frontier models, even if they’re quite a bit more expensive. The set of things you can profitably do with better intelligence than everyone else is unbounded.
While this is trivially true, the question is if these remaining use cases that separate frontier models from non-frontier models bring in enough revenue to offset the massive spending of the frontier labs.
Which is why while HN is sleeping and thinking Google and MS are done in the space, Copilot and Gemini are winning most of the non-tech: it's just an extension of their current teams/cloud offerings that companies were already using.
I have asked most of my non-tech friends: chemicals, banking, insurance, consultancy. They are all on Gemini or Copilot.
This doesn't really match my experience, but all I have to counter your anecdotal evidence is my own anecdotal evidence.
In my opinion, concluding that Copilot is "winning" right now is like concluding that Internet Explorer was winning while Chrome was still young. It may be technically true, but it's easy to see the writing on the wall -- that it's only winning because it is the default option available at most companies.
I have never met a satisfied Copilot user, and as a consultant I met a lot of Copilot users. It's hard to believe how totally and completely Microsoft is shitting the bed. As you point out, they have the enterprise integrations. Why can't the build a good product with them? Generational blunder, in my opinion.
yeah nobody is happy using windows either, but still they all do anyways.
we would be better off using linux all together, but here we are.
Lets see maybe every company will have their own little instance of a local LLM in their basement in the future, when the AI-bubble pops and server-racks become cheap
If we mean Gemini the gemini.google.com, or gemini built into sheets or other Google surfaces, then people in these orgs might be getting a very different experience of what LLMs are and what they're capable of.
Gemini the web/app is basically where ChatGPT was on release day, maybe even worse than that. It's laughably bad. It forgets the context between turns, it has dynamic tool injection which doesn't get injected half of the time, which causes model to hallucinate tool calls.
The new Spark is much better, but I don't think most people are using that.
I've noticed the context loss in subsequent turns, it's so strange. For me it happens whenever I stop gemini mid prompt. Sometimes its response annoys me, so I'll hit stop frequently. And then subsequent msgs seem to go through like a new chat where it forgets what we were talking about.
copilot is doing the same.
i just joined a corporate and they only allow copilot. i am getting brain cancer using it, especially being used to claude max from before.
but cost is a thing, if i use the LLM gateway for a chatGPT-5.4 i am quickly (3 days of very light use) on 50$ in token. and everyone and their intern are using copilot. As azure has the guarantee that company data will not leak they are taking the bait.
The fear of secrets landing in the training data is just too big, and they are all microsoft 365 customers already. And lastly Management seeing that copilot can autonomously edit an excel file for them already counts as mindblowing... 0_0
toephu2 · · focus · HN ↗
No moat and competition is good for consumers though.
Marciplan · · focus · HN ↗
[dead]
tomhow · · focus · HN ↗
boelboel · · focus · HN ↗
senordevnyc · · focus · HN ↗
But there is always a use case for frontier models, even if they’re quite a bit more expensive. The set of things you can profitably do with better intelligence than everyone else is unbounded.
So yes, the number of tokens that get pushed through commodity models for very cheap will continue to grow, but so will the tokens for expensive frontier models. We’ll never run out of things to do with the latest geniuses who are twice as smart as last year’s geniuses.
The moat is the engine that builds and sells the latest geniuses, especially the compute. And no one will have more compute than American labs for the next 5-10 years.
OrangeDelonge · · focus · HN ↗
senordevnyc · · focus · HN ↗
I picked double out of a hat, obviously there’s some intelligence gap that’s too small for a price premium beyond a certain point, but it seems like OpenAI and Anthropic’s revenue keeps climbing, so I don’t think we’ve hit it.
mobelkh · · focus · HN ↗
in any case, I've been using open and closed models since sonnet 4, i remember when the best I could get was qwen 3 480b coder, you can definitely feel the gap closing going from that and GLM 4.5, to GLM 5.3, DeepSeek Flash V4.1, Kimi K3 etc, it's reached the point where i wish I had V4.1 at work, it's faster and bullshits me less when I use it in my personal projects. And I have unlimited access to fable 5.1
anon373839 · · focus · HN ↗
rbehrends · · focus · HN ↗
While this is trivially true, the question is if these remaining use cases that separate frontier models from non-frontier models bring in enough revenue to offset the massive spending of the frontier labs.
senordevnyc · · focus · HN ↗
epolanski · · focus · HN ↗
Which is why while HN is sleeping and thinking Google and MS are done in the space, Copilot and Gemini are winning most of the non-tech: it's just an extension of their current teams/cloud offerings that companies were already using.
I have asked most of my non-tech friends: chemicals, banking, insurance, consultancy. They are all on Gemini or Copilot.
redhale · · focus · HN ↗
In my opinion, concluding that Copilot is "winning" right now is like concluding that Internet Explorer was winning while Chrome was still young. It may be technically true, but it's easy to see the writing on the wall -- that it's only winning because it is the default option available at most companies.
I have never met a satisfied Copilot user, and as a consultant I met a lot of Copilot users. It's hard to believe how totally and completely Microsoft is shitting the bed. As you point out, they have the enterprise integrations. Why can't the build a good product with them? Generational blunder, in my opinion.
epolanski · · focus · HN ↗
busssard · · focus · HN ↗
we would be better off using linux all together, but here we are. Lets see maybe every company will have their own little instance of a local LLM in their basement in the future, when the AI-bubble pops and server-racks become cheap
glub · · focus · HN ↗
If we mean Gemini the gemini.google.com, or gemini built into sheets or other Google surfaces, then people in these orgs might be getting a very different experience of what LLMs are and what they're capable of.
Gemini the web/app is basically where ChatGPT was on release day, maybe even worse than that. It's laughably bad. It forgets the context between turns, it has dynamic tool injection which doesn't get injected half of the time, which causes model to hallucinate tool calls.
The new Spark is much better, but I don't think most people are using that.
s3p · · focus · HN ↗
busssard · · focus · HN ↗
but cost is a thing, if i use the LLM gateway for a chatGPT-5.4 i am quickly (3 days of very light use) on 50$ in token. and everyone and their intern are using copilot. As azure has the guarantee that company data will not leak they are taking the bait.
The fear of secrets landing in the training data is just too big, and they are all microsoft 365 customers already. And lastly Management seeing that copilot can autonomously edit an excel file for them already counts as mindblowing... 0_0