Systems engineer Yacine argues heavily funded robotics labs are on the wrong path by relying on massive compute
Story Overview
Systems engineer Yacine positions his lighter-footprint methods for robot control and simulation-to-real transfer as a direct alternative to the heavy compute budgets at places like Generalist and Pi, framing the difference as a strategic misstep rather than a scaling advantage while he builds an unfunded hardware startup.
Exchange turns personal fast
Chris Paxton’s reply flags Yacine’s active job search as evidence against the critique, turning the thread into a back-and-forth about credibility instead of code or hardware specs.
Claims stay simulation-bound for now
Yacine’s cited wins center on cart-pole style tasks with no released metrics, datasets, or real-robot comparisons against the labs he names, leaving the performance gap untested in public.
Positive users praised the builder's low-compute robotics approach for emphasizing intelligence density and challenging funded labs, while negative users dismissed it as unoriginal or questioned the claims.
Most Activity
@yacineMTB If they were wrong why are you looking for a job
I'm trying to bring robots to life with incredibly little compute, relative to the big silicon valley super funded labs like generalist & pi. I think that they're wrong
@chris_j_paxton i-i need funding
@chris_j_paxton happy to consult!
@LonnyLot All robots
@chris_j_paxton seriously though; if you have a problem that you are struggling with at your work i am happy to help solve it for money
@yacineMTB Is there a specific type of robot you are trying to build?
@MancerAI_ mingru
@yacineMTB I keep bothering you but have you looked into ReCoN (request confirmation networks)? I think they'll be perfect for robotics. https://ceur-ws.org/Vol-1583/CoCoNIPS_2015_paper_6.pdf
@yacineMTB I'm working to bring bio-bots into the main stream to help with solving the fertility crisis......
biobots.......women
@yacineMTB Same but with my brain (it also has very little compute)
@yacineMTB Get it bro, literally these people are throwing LLMs at problems that require 4 bits
@yacineMTB you’re such a nerd i love it
@yacineMTB but can it do my dishes or laundry. literally all i want for now from a robot
FOR NOW
@yacineMTB If it can't fit in my pocket it's not asi
@yacineMTB Probably right, Taalas HC1 is insane for what it does. And if you switch to analog processing instead the amount of power required would be laughable. Orders of magnitude down at the bottom.
@yacineMTB You can put a brain in a petri dish and it can play doom. A guy in Montreal did it who is not you and way more smart and talented. He also made a shirt from shrimp
@yacineMTB 👀 ahhh a fellow chad
@yacineMTB Intelligence density is the way
@yacineMTB the ETH way, this is very good for specific well defined problems, but remember that the big labs are trying to solve general intelligence for robotics and not single focus tasks
@yacineMTB I believe in you pal if I were VC & sh*t I would have backed you with good money but moi poor so you carry on with my moral support