Systems engineer Yacine flags how readily available frontier models like Anthropic's Fable 5 and OpenAI's GPT-5.6 Sol hand rival teams the same high-performance starting point, shrinking the proprietary edge that currently separates leading labs from everyone else.
I don't know how gpt sol and fable level models get vended out to end users without it immediately accelerating the development of every other AI / engineering lab out there such that it equalizes. Fable is really good. If sol is as fast as they say it is and also as good..
Fable 5 reached broad API and subscription availability after a brief export-control pause, while Sol remains limited to a small set of vetted partners, creating an uneven window for teams to experiment and close performance gaps.
Once advanced capabilities circulate openly, the original developers lose the exclusive runway that previously let them iterate ahead of copycat efforts, turning distribution choices into direct competitive trade-offs.
Negative users report Fable suffering errors and instability during tasks while criticizing US export controls as counterproductive against advancing Chinese open-source models.
No Digg Deeper questions have been answered for this story yet.
When I say AI or engineering lab, I mean specifically the Chinese ones. Which, by my guess, will probably continue open source great models. At which point, what is the remaining business edge?
I don't know how gpt sol and fable level models get vended out to end users without it immediately accelerating the development of every other AI / engineering lab out there such that it equalizes. Fable is really good. If sol is as fast as they say it is and also as good..
My prediction is that likely, these AI labs become machines that produce custom artifacts, custom intelligences for specific tasks -- the researchers and the research talent is provided as some service where you enable some client. Nothing else makes sense to me
When I say AI or engineering lab, I mean specifically the Chinese ones. Which, by my guess, will probably continue open source great models. At which point, what is the remaining business edge?
I wonder how well this echos the computing / banking revolution for example. From my conversations with the ones that have gone through it, most of the money was made from the deployment of the technology to the banks
My prediction is that likely, these AI labs become machines that produce custom artifacts, custom intelligences for specific tasks -- the researchers and the research talent is provided as some service where you enable some client. Nothing else makes sense to me
There are some real wild cards out there. Labs that seem to me, impressive, different enough, talent diversity. Like, the sakana-type AI labs.
It's really hard to predict what is going to happen to all of these companies. How do they stop what is coming?
So really, in consumer land, affordability and friction is what matters the most, and I think google wins that by default which is kind of sad? OpenAI is kind of sticky with normies, maybe some enterprisey things here and there. They also have a lot of compute
So really, in consumer land, affordability and friction is what matters the most, and I think google wins that by default which is kind of sad? OpenAI is kind of sticky with normies, maybe some enterprisey things here and there. They also have a lot of compute
@yacineMTB the trick is to bet on what’s coming
There are some real wild cards out there. Labs that seem to me, impressive, different enough, talent diversity. Like, the sakana-type AI labs.
It's really hard to predict what is going to happen to all of these companies. How do they stop what is coming?
@yacineMTB embrace open source models + tooling, build the capability factory
@yacineMTB the trick is to bet on what’s coming
@willccbb Wow. You're right. But I have a problem. My company is valued at two trillion dollars based on the premise that it WONT happen
@yacineMTB embrace open source models + tooling, build the capability factory
@whatdafuqkyle Actually, Jensen was right about export controlling the GPUs imo. The right move was to compete directly because anything else would cause too much damage long term. We are going to see it play out as people produce custom compute for training in xna
wtf happens when open source china models get better, and Americans move to deepseek and others??
seems pretty fucking dumb to be export control Nazis for h100-800’s, while leaving backdoors to all our models, for the Chinese to copy.
so they advance in lithography and come up with their own shit. we could pull a Huawei and backdoor their entire systems.
@JeffKazzee Corporation too corporate
@yacineMTB End users likely pay 10x? 100x? Who can actually run Fable / GPT sol 24/7 for 6 months straight?(other than the provider?)
@yacineMTB I'm not here to defend any single provider, but Google had so much potential. Why are they so far behind and acting like Gemini is good enough? Possibly the normies? Though i am a huge fan of of their new nano banana lite, it's decent and fast.
@yacineMTB have you considered pivoting to management consulting and or regulatory capture
@yacineMTB I suppose. but taunting them when they will likely take 20+ years to catch ASML seems like a bad route when you consider Taiwan and tsmc.
@yacineMTB to think these models are going to be the average baseline for engineering in each model in 6 months is insane
@yacineMTB The delay is probably so they are 1 generation ahead internally
@yacineMTB I get noon so orange text from Fable , saying it can't do it and downgraded.
Are you not seeing that? It flipped its lid just working on a Signal IR app I was developing.
@yacineMTB So, a consulting firm?
@PyZooKo One of the ways these things self improve is they become cheaper