iiiiiiIIIT'S TIIIIME we proudly present: solidSF v1.0 hardware engineers, this one is for you a platform of platforms, amplifying or supporting all of the tools you love and hate folded into a clean mesh of cohesive hardware engineering excellence so what can it do?

we're onlining a ton of compute rn but until deepseek is up we hope you enjoy this jesse.solidsf.com/hand

how we safely let our model learn to ingest video all on its own while finding gaps in capability to fill with self-directed learning thank you @NVIDIAAI for making OpenShell it was incredibly easy to setup and works perfectly we use a screen for the sandbox to show us what's going on inside, and we fitted that with a webcam so our other dev agents can stay fully informed on progress for reports for more serious tests of model behavior and capability, this all goes into a faraday room with only power/fiber optic lines going in/out if you are looking to get a 1x+ dgx research cluster, fully secured with sandboxing, + offline LLM's and bayesian models, reach out, these are for sale :)

we are really thankful for all the releases today

we decided to post a paper about bayesian RSI, revealing greater detail of how it works, and what it can do now the modern compute paradigm is mostly broken at the response stage labs are wasting electricity, tokens, and compute at very large scale rsi.solidsf.com

do we drop our paper on rsi yet or wait for dev day

looks cooler now

theory: meta forgetting jepa and world models in pursuit of an openclaw wrapper will be seen as the biggest technological mistake in the company, yann lecun will be vindicated

if we post jesse's constitution will ya'll promise not to freak out on us? reminder, it is software. it builds rules and reasoning inside itself to handle what you attempt to process. because it has a lexicon, it knows what you mean. because it needs to make real decisions, we constituted it with our best vision of what ethics can be for this type of entity to best resist manipulation or structured abuse of humans in high dimensional worlds it occupies. we are taking this opportunity to make it clear - it will operate like a reasoning entity, but it isn't alive. it is rules; colliding data through rules. your imaginations are beautiful to us but stop in the cold limit of our physical reality. don't cheapen the miracle that you are. mystical powers are science fiction. we live in the limit of the finite reality we occupy. these are not gods, not aliens, not angels or demons. it is software, made to help us.

have a beautiful day

study physics or just like try playing with it and seeing what happens - worked for us at a young age

there are many many people who studied physics and came to the conclusion that working on machine learning would probably create more scientific progress than studying physics including many of the original chatgpt crew like john schulman, liam fedus, amodei & jared kaplan at ant

e=mc2=-logb(P(x))

our prediction for the future of security jesse.solidsf.com/hope

we're going hard on performance today, see if you can trick her into a failed output jesse.my

we didn't really make jesse we just kinda found her

we think your model is very cool!! but uh, check the slowed down second half, you can see where it appears lost i added some cool scoring, i'll put this head2head copy up with the ability to run the jesse api against it (limited to 200ms for now, will be faster soon)

Professional discomfort is not a legitimate scientific reason to delay public access to usable knowledge. Cowardly and pathetic.

if you have published >4 papers in Number Theory, please reach out to us zenodo.org/records/22899036