
rw ./
7.1K posts

rw ./
@gradientintern
push past limits | real niu for @Gradient_HQ | part time troll | full time janitor | professional edging specialist





construction spending still going up for data centers, but its growth rates has been slowing. likely due to constraint on resources and public pushback. more spaces for clankers compared to humans continues


US construction is now spending 9% more on Data Centers ($50B) vs Office Space ($46B) Orchestration & efficiency of resource usage increases in importance, I bet agentic SaaS and platforms will be top beneficiaries Fleet optimization will be defaulted to asynchronous operations


looks like price forecast and hikes is hitting DeepSeek previously did a permanent 75% discount, now hiking prices during peak hours by 2x likely also signals a significant increase in user growth. our beloved whale also lets us run locally on our own hardware. awesome folks







the convenience of AI will likely hit consumer electronics prices even harder then it already has. memory prices going higher from widening supply gap. and we all know how companies are with price hikes, it stays elevated after hike.


building ideas is the easier than ever before. but that also means low usage on alot of garbage ideas. execution phase of the process is now more important then ever before. in this lens you can have near infinite tries going local and cost efficient tries on open labs API. do you agree?



with $100,000 you can get 21x more tries on Kimi K2.6 compared to Opus 4.7 or 42x more tries compared to Fable 5 performance close enough with multiples more in tries. with corporations cutting back on token spend, this could be the rise of more efficient models.


A self-evolving agent + a 428B model + 3 Macs = ? Your own AI lab. We ran @MiniMax_AI M3 locally with @tryParallax, right on our desk. Then @GA_agent_ai took over to create a 5-stock portfolio and write it to disk. No cloud. No API bills. Nothing left the machine. Wild to see a ~3K-line agent drive all this with a 400B+ model on local hardware. Thanks to the GenericAgent and MiniMax teams for making local AI feel real.




