@cryptopunk7213

the chinese labs really are relentless with accelerating open source ai. deepseek is doing everything it can (and succeeding) in maintaining parity with top models, large american companies are shifting to chinese models (coinbase, microsoft) deepseek is newly funded with $8B warchest. zhipu is basically state funded, stock is up 15X off the back of GLM’s success the irony of china leading open source while america throttles the release of gpt 5.6 and fable ofc china’s goal isn’t to open source intelligence it’s to kill the american lab valuations and win on cheapest cost to access models. china has the most compute if they can solve the chip problem

few things that’ll become increasingly more valuable over the next 6 months: - which ai models you use for what tasks - how you route across multiple models per task (harness) - caching to reduce token spend - ai cloud: efficiently managing memory and data bases the bottleneck is shifting from LLMs to aggregation

i’ve never seen anything like the gravitational pull openai and anthropic are currently exerting on the talent market - it’s like everyone’s thrown the towel in and accepted 2 companies will run the world going forwards: in the last 10 days key leaders at google deepmind, meta, apple, leading universities have quit and joined them. and it’s not just researchers - economists, scientists, psychologists, priests (even the pope!) are strategically aligning themselves with these 2 labs. this is going to have major implications on societal hierarchies, policies/law and government going forwards - not to mentions the obvious wealth distribution absolutely insane times

are we really ending the week on a depressing cliffhanger? fable 5 almost certainly won’t return soon and when it does it’ll likely be US only. gpt-5.6 getting the same treatment - no public access and yet the massive irony of apple lobbying trump to buy chinese memory chips and microsoft serving ai models to chinese companies via their cloud service technology’s magic comes in the ability to scale *universally*. that’s how you end up with an immigrant building a decacorn on american soil. nationality mandates is only going to end up dystopian imo

i wonder who overturns the government ban first, openai or anthropic the “best model” will be one with safeguards the gov actually likes, switches up the competition completely to a politics game both companies are in chats with trump, but sam is arguably doing a better job atm (not as hated)- will be curious to see if gpt 5.6 speedruns the approval.