ah another weighty cudgel for me to wield at work

she too may hope one day to be played by David Schwimmer in a TV adaptation of a famous court case

Having just spent 20 minutes in the car with a sniffling and snorting teenager I can attest to the fact we are clearly born with a powerful aversion to signs of sickness.

Phoenix is just the absolute worst representative of Arizona, all featureless concrete expanse in any direction and oppressive heat that assaults one’s existence from above and below.

I’m saying it scales better than we think it does because it’s possible to do proofs for small parts of a system and have those remain valid when those parts are added to a larger system without reproving them in a much larger state space.

Separately, it is limited in the system it can model.

No! Model checking state machines doesn’t cover every problem even as a matter of the limited expressiveness of temporal logics and simulation arguments, absence of modularity aside.

My claim here is the very boring one that it scales better than we currently suspect in practice in those cases.

The goal is to prove for modules and get the theorems to stay around after composition.

An interesting historical note here is that Lamport sort of swept this problem under the rug in TLA.

Depends on how “scaling” is defined but for model checking the obvious meaning can be addressed through good ol modularity.

Splitting up an exponentially growing state space turns out to be a performance win ☺️

Anyway, we’re doing the mechanization of semantics & theorems at work right now.

Bear Down!

"I have a decent fluency in LLMs, and they have utility, but the absurd degree of over-hype, the way they're being forced on everyone, and the insistence on ignoring the many valid critiques about them make it very difficult to focus on legitimate uses where they might add value."