As an extended toy universe Lego is great but as STEM skills and intuitions catalyst it is a cul de sac of arrested development. There is a weird way it directs your attention away from all the hard problems, which are partly encapsulated away, partly curated out entirely.

Blogger at ribbonfarm.com, running summerofprotocols.com, host of /gloom ☀️

It is an extremely expressive medium for modeling some closed-world subsets of engineering in a way that achieves a certain stylized skeuomorphic fidelity at the expense of insight into base principles. Technic mitigates that somewhat but still feeds the wrong motivations.

Toy universes are like standardized tests. Being good at standardized tests mainly demonstrates that you have a talent for standardized tests. Being good at Lego mainly demonstrates that you’re good at Lego. True of toy-like programming languages/environments for kids too.

I’m not a software developer but is it fair to say the history of programming as a Lisp —> C feature migration flow (I think Paul Graham made up that theory?) is a Lego —> real world migration flow?

I’m developing a broader theory that Lego-like escaped realities are fundamentally unstable and their abstraction stack has to leak first them to persist. They last as long as the leakiness mitigates the tendency of the notional non-leaky ideal to float off into bullshit

Yes the joy and meditative quality is the main value. Legos catalyze poiesis even if it’s sometimes at the expense of praxis. For me the grittier, more realistic Meccano sets struck the right note.

From my recent talk https://studio.ribbonfarm.com/p/towards-a-metaphysics-of-worlds

I have to write this up properly. The headline idea is what I’ll call the leakproofness-degeneracy tradeoff. An escaped reality is degenerate with respect to the higher dimensional reality it is embedded in proportion to its leakproofness, and will require proportionate external intervention to stabilize.

If you like 2/3 triangles, it is: Pick 2 of 3 — leakproofness, non-degeneracy, stabilizability. 

Might be my best triangle yet

Though I didn’t develop this idea for the purpose, it provides a good frame for my critique of AGI/alignment theology. The problem is, it is a Lego model of AI pretending to be the real thing, and claiming to hit 3/3. And as with Lego it leaves out absolutely critical features of the real-world thing it aims to model

Stub to myself to work in how VMs, containers, intermediate representations etc fit into this theory. Plus the maintenance required to make it all work such as garbage collection, and the performance cost

Rather appropriately, I was simul-posting this thread on bluesky and the reply conversations in the two places caused the mirroring to break. The abstract idea of the thread leaked and broke down. This warpcast version is now officially the canonical version.

There is a like Lego-like streak in OOP philosophy but unlike Lego, it enjoyed decades of practical use because it was a leaky mixed paradigm. Smalltalk = Lego probably. C++ = mixed paradigm. Java = mighty striving to keep it Lego like in production that arguably failed by bloat.