akshaan

[horse dentist @equine__dentist]

don’t talk to me until i’ve pondered my corb (cold brew orb)

[akshaan @akshaan]

Another cold brew straight to the dome

[akshaan @akshaan]

the fridge cigs will continue until focus improves

[akshaan @akshaan]

Sometimes Sol speaks in tongues to deliver a prophetic vision before returning to business as usual

[akshaan @akshaan]

No babe I love your mind, the unicameral ones scare me

[atlas @creatine_cycle]

there is nothing in this life that artificial stimulants cannot solve and i think that's beautiful

[akshaan @akshaan]

the fridge cigs will continue until focus improves

[akshaan @akshaan]

Guys on twitter have solved Navier Stokes and you're bearish, anon?

[Joscha Bach @Plinz]

The artificial is natural, too

[akshaan @akshaan]

and you have to swim there to get in

[Alex Kehr @alexkehr]

they should turn Alcatraz into a Soho House

[ً @artfuIchaos]

to fuck around is human. to find out is divine.

[akshaan @akshaan]

Love how Apple watch chose to prompt for iCloud password verif when I was in the SF bay, fighting for my life between the currents and an overly curious baby seal

[David Pfau @pfau]

It's kind of wild that there are not one but two cults built around Bayesian inference and they are almost entirely disjoint.

[Jonathan Birch @birchlse]

AI may be smashing it at analytic number theory but continental number theory is so far untouched.

[Lewis Jones @thelcjones]

Transformers are cool but diffusion models are probably the most beautiful and elegant discoveries of the last decade

[sway @SwayStar123]

Diffusion models are schizophrenia machines

Adding my perspective on diffusion models, I believe that hallucination is actually the core of how diffusion models work!

https://t.co/YbqJZV3Nr6

[akshaan @akshaan]

Interesting that a flow matching model with linear conditional paths and a Gaussian prior can be sampled not only via its deterministic ODE but also via a family of diffusion SDEs with the same marginals.

In this setting, the velocity field, score field and posterior mean (denoiser) are linear reparametrizations of each other. Once we recover the score from the model��s learned velocity, we can choose a diffusion matrix and “correct” the drift so the resulting SDE has the same probability current, and therefore the same marginals, as the ODE.

https://t.co/HwuW7lXKBw

[Georgios Konstantopoulos @gakonst]

Open Sourcing Centaur: Multiplayer, self-hosted, secure agents for Slack.

Centaur has been transforming how @paradigm and @tempo invest, build and research.

Now you can run it yourself on infrastructure you control. Instructions below.

[hope hopes hoping @hopes_revenge]

PSA if you take enough stimulant medication you can actually understand Wittgenstein