iiinova — where your digital self keeps evolving

Christopher Suter @cgs.bsky.social · 31/08/2026

It's never too early to start talking about the lengths of coastlines

000

Christopher Suter @cgs.bsky.social · 22/06/2026

Sorry, maybe the term is "inharmonic" not anharmonic. en.wikipedia.org/wiki/Piano_a...

en.wikipedia.org

Piano acoustics - Wikipedia

100

Christopher Suter @cgs.bsky.social · 22/06/2026

A fun fact about pianos is that the larger strings become anharmonic and must be tuned with their fundamentals in the wrong place so that the actual harmonics are in tune with the higher notes. Each piano has its own "tuning curve". There's good (pricy) software that will automatically find it.

110

Christopher Suter @cgs.bsky.social · 10/06/2026

I was pretty proud of this one too: using magnatiles to reproduce some of the more abstract geometrical puzzles from The Witness to make them more intuitive for 5yo:

5yo traces in the solution to a geometric puzzle in the game The Witness, following the solution found by building the puzzle in the floor from magnatiles.

000

Christopher Suter @cgs.bsky.social · 05/05/2026

There's some literature on Koopman operator theory applied to brain data, in the same ballpark. MZ is beautiful stuff but challenging in practice I think, as @patrickmineault.bsky.social alludes to. Choosing a coarse graining is hard and then you get some pretty gnarly integro-differential eqns.

020

Christopher Suter @cgs.bsky.social · 05/04/2026

Oh that's the distribution but there's also a distribution transformation ("bijector") iirc

010

Christopher Suter @cgs.bsky.social · 11/03/2026

Pretty pleased with this kids' book showing proper Lagrangian mechanics on a whiteboard

010

Christopher Suter @cgs.bsky.social · 09/03/2026

I can only imagine how that must have felt

120

Christopher Suter @cgs.bsky.social · 04/03/2026

If they were getting closer they'd be blue dots

3681

Christopher Suter @cgs.bsky.social · 29/01/2026

I agree it's coarse, but making resource allocation decisions under significant uncertainty is definitely a central part of both science and engineering. One could defensibly call this betting.

120

Christopher Suter @cgs.bsky.social · 09/01/2026

It's usually a bit hard to find, sometimes it's in accessibility settings, but most sites let you disable auto playing video. At least then you have a bit more choice.

000

Christopher Suter @cgs.bsky.social · 25/12/2025

We discovered a new way to tile the play space with magna tiles tonight! God I love these things.

Planar tiling with various magnatile shapes (large squares, small equilateral triangles, and long isosceles triangles)

150

Christopher Suter @cgs.bsky.social · 24/12/2025

Why is the default mathjax renderer one that fails on the prettiest symbols? anyway...option-click (or right-click or whatever), select Math Settings > Math Renderer > Common HTML to make the mathcal's not be boxes, and the mathbb's have proper serifs, and the Greek letters be elegant, &c &c &c...

100

Christopher Suter @cgs.bsky.social · 16/12/2025

Chocolate Jesus is a nice touch

030

Christopher Suter @cgs.bsky.social · 09/11/2025

I think the best answer is to import into Books, which also gives a decent interface for navigating documents in the hundreds of pages. Download the PDF, find it in Files, "share", from the list of apps select the triple dots, scroll down to the bottom, click Books. Easy peasy 😒

010

Christopher Suter @cgs.bsky.social · 09/11/2025

My kingdom for an iOS pdf reader that doesn't lose track of the page you're on

140

Christopher Suter @cgs.bsky.social · 03/11/2025

On mac or iphone, press and hold 'h'. A popup appears with a barred option.

040

Christopher Suter @cgs.bsky.social · 24/08/2025

Yes! It's a beautiful, simple story and the music irresistible. There are no references to get (probably some mythology I'm oblivious to)

000

Christopher Suter @cgs.bsky.social · 09/08/2025

I enjoyed David krakauer's recent visit to ML Street Talk (indeed, it rather increased my estimation of the show itself). And all of the Sean Carroll Mindscape catalogue is thoroughly enriching. I especially like the solo episodes on various topics (emergence, reality of time, crisis on physics)

000

Christopher Suter @cgs.bsky.social · 06/05/2025

idk these are definitely finite cardinals

020

Christopher Suter @cgs.bsky.social · 22/04/2025

Not exactly that, and doesn't have graphical DAGs, but you might find Inference Gym interesting. We host it as a "spinoff" of TFP (it's maintained by a TFP maintainer). Github: github.com/tensorflow/p... Tutorial Colab: colab.research.google.com/github/tenso...

colab.research.google.com

Google Colab

010

Christopher Suter @cgs.bsky.social · 14/04/2025

It's tired, Vince what do you want

030

Christopher Suter @cgs.bsky.social · 23/02/2025

Closely related, I was dissatisfied, a while back, with all available models' renderings of ASCII Groucho Marx (e.g. chatgpt.com/share/67babe...)

chatgpt.com

ChatGPT - ASCII Groucho Marx

Shared via ChatGPT

020

Christopher Suter @cgs.bsky.social · 20/02/2025

I think "see" is doing a lot of work in Vince's assertion.

011

Christopher Suter @cgs.bsky.social · 26/01/2025

Was gonna say 45² and then I felt kinda sick

010

Christopher Suter @cgs.bsky.social · 10/01/2025

Also "Don't Think Twice it's Alright"

010

Christopher Suter @cgs.bsky.social · 07/01/2025

the story of multi-shot continuations in haskell is fun. i can't fathom what the C++ equivalent would entail 🫣

110

Christopher Suter @cgs.bsky.social · 04/01/2025

I have no idea what OR is and at this point I'm too afraid to ask

100

Christopher Suter @cgs.bsky.social · 04/01/2025

Hmu if you get into Naesseth and want to talk SMC sometime!

010

Christopher Suter @cgs.bsky.social · 04/01/2025

The devil is in the details, of course. SMC is kinda magical and its analysis is (to me) very challenging. I know how to use it but I don't grok why it works (or, more importantly, how and when and why it fails). I'm hopeful this book provides alternative perspectives that make it clearer.

110

Christopher Suter @cgs.bsky.social · 04/01/2025

runs the dynamics forward to get a future distribution. F-K seeks to invert this procedure. It succeeds by employing the path integral approach that Feynman and others popularized in QFT. IIUC this in turn lends itself to MC simulation methods.

210

Christopher Suter @cgs.bsky.social · 04/01/2025

I understand this book as drawing a connection between particle methods like SMC and the path integral formulation of QFT. IIUC Feynman-Kac is to inverse problems (eg bayesian inference) as Focker-Planck is to simulation. For given dynamics, F-P takes an initial distribution and...

110

Christopher Suter @cgs.bsky.social · 03/01/2025

I'll have to check that out. I read permutation city years ago. Good stuff

010

Christopher Suter @cgs.bsky.social · 03/01/2025

I suspect at bottom it's computation (whatever that means). But hey I'm just a drunk under a street light...have you seen my keys?

110

Christopher Suter @cgs.bsky.social · 03/01/2025

I think if we're talking about precision of measurement, then energies at small scales become important. But if we're just talking about, say, storing digits of π to higher and higher precision, I suppose we can just keep making more and more black holes to store them all.

110