"Follow the gradient" That's great. We are learning so much from open source, shipping, screwing up and figuring it out. Plans are good, reality is better
"Follow the gradient" That's great. We are learning so much from open source, shipping, screwing up and figuring it out. Plans are good, reality is better
@jimkxa Nvidia doesn't do long term strategic planning. Ship and iterate.
@jimkxa "Feedback beats planning" "knowledge only comes from empirical observation" "Follow the gradient" post is full of bangers
@jimkxa How do you know the gradient doesn't trap you in a local minimum?
@jimkxa How do you apply this to silicon design? I assume atomic semi can help there?
@jimkxa Is Tenstorrent planning to start Cloud Notebooks (like Google Colab) anytime soon so that users can get their hands on the hardware and you guys get the feedback?
@jimkxa Have a better strategy. Only rely on planning if past predictions are correct. Otherwise send it and iterate fast.
@jimkxa "Everybody has a plan until they get punched to face." - maybe for sw its until they ship something out.
@jimkxa @__tinygrad__ Has also His interviews with @lexfridman are very informative Their work on self driving using already available technology within existing vehicle stock if implemented would save a lot of minor to major injuries around the 🌎 Could be shipped & installed locally
@jimkxa It's why agile SDLC can be superior to waterfall. It's difficult to spend time up front on BRDs and detailed designs and do better than reality checks along the way through an MVP and beyond.
@jimkxa Y. Feedback = a jagged hypotenuse to the right thing. By Pythagoras this is much shorter than water-falling to the wrong thing that no one wants, then traveling the long distance to what they do (at best).
@jimkxa these folks are too close to the brain stem for that philosophy. a company like tenstorrent is best fit for this optimization in the short term. but meta will use deep neural nets to gradient-descend on neurotransmitter release, for maximum integrated attention.