Derive it, draw it, compare it.

Understand how the ideas fit together.

Embodied AI and robot learning, explained from first principles.

Course catalog

Example: the ELBO, one justified step at a time

1 $\log p(x) = \log \mathbb{E}_{q(z \mid x)}\!\left[\dfrac{p(x, z)}{q(z \mid x)}\right]$ Rewrite as an expectation under q = q(z | x)
2 $\ge \mathbb{E}_{q}\!\left[\log p(x, z) - \log q(z \mid x)\right]$ Jensen’s inequality: log is concave
3 $= \mathbb{E}_{q}\!\left[\log p(x \mid z)\right] - \mathrm{KL}\!\left(q \,\|\, p(z)\right)$ Factor p(x, z) and collect the KL term

Derivation cards: Every key equation, from the ELBO to flow matching, derived one justified step at a time.
Comparison charts: Side-by-side tables of VLAs and world models, generated from one dataset and kept up to date.
Landscape roadmap: One map of the field from 2022 to today, with a learning path through it.
Close readings: Paper-by-paper readings of frontier systems, ending with DreamZero and Dream-1.

Three questions a robot model can answer

Family	Learns	Answers
VLA	$p(a \mid o, l)$	What should I do?
World model	$p(o' \mid o, a)$	What happens if I do this?
World action model (WAM)	$p(o', a \mid o, l)$	What should I do, and what will it look like?

Get new lessons, cards and the full roadmap poster when they ship. No spam.

Sign-ups open soon. Until then, write to contact@dpintelli.com.