Bookmarks

We're in the middle of the second renaissance where proof of work replaces the resume, an audience replaces the employer, and taste replaces credentials.

These curves turn linear algebra into intelligence.

Without activation functions a deep network is just one big matrix multiply.

σ(x)=1/(1+e⁻ˣ) squeezes to (0,1); tanh(x) centers at zero; ReLU=max(0,x) zeros the negatives; Leaky ReLU, ELU and Maxout keep a gentle slope so gradients never die.
They are the reason neural nets recognize faces, write code and drive cars.

The simplest nonlinearities often unlock the deepest learning.
Which of these do you reach for first when a new model refuses to train?

Now that Unitree has IPO'd, I'll drop what will likely be my last blog sometime next week. I'm not quite done writing just yet 🙃 It's ~3 months (~200h) of research condensed into ~70 minutes. Exploit is wormable, so if one G1 gets popped, it RCEs all other G1s in proximity

good strategy, creating your own experience is permissionless

thanks for all the love on this, i'll publish for you to play soon

in the meantime here's a demo of some new mechanics:

• bending light around a black hole
• splitting with a prism
• focusing with a convex lens

more physics - puzzles - levels coming soon :]

@signulll Zero latency personalization is magic. Once you stop paying per-token on the micro-copy touches, the entire product surface opens up.

we’re using local apple intelligence models a lot these days in our product for a lot of little things & they work incredibly well, especially for personalization & copy.

you can cache these so you create a pretty lovely little experience in subtle places where the app feels tailored than just one to many (all for free!).

it’ll feel cool as hell when they get even smarter.

@dwr how do you tell if something is ai written?

best cold email i've seen in my life

Be a reply guy.

If you’re under 1000 followers, it’s tough to get traction on your posts.

You gotta grind it out until you get there, then switch your focus to posts.

Annoying but everyone has to do it.