@justintroutman
southern drawl. applied crypto. disinfo research. tradecraft historian. street baller. bats left. electric + bass player. trail runner. 4x dad. 2 dogs. erdős-bacon #6. cryptography @ duke. security @ panw. justintroutman.com
Codes and Ciphers by Andrew Pennycook.
Well, it happened — one of my kids got into cryptography (independently).
I passed along a book I’ve had for well over two decades now, written by Andrew Pennycook for his daughter Veronica, back when books on codes and ciphers — especially for kids — were scant.
Now, another dad to his daughter.
For context, my Duke summer course ran roughly 10 days, on campus, with 6 in-classroom hours per day. We held multiple rounds of an AES-selection-process like field cipher demolition derby. (This was the capstone-like project.)
If you, or maybe even your high schooler, were to take a hands-on course around field cipher design (scaled down SPNs, Feistels, ARX, Lai-Massey, sponges), what would the optimal format be?
In person or Zoom?
2 days or 2 weeks?
2 hours per session or 4, or more?
Name your ideal format and cadence.
This will be a key part of upcoming workshops on field cipher design and lambrOS, allowing students to implement their own capstone designs. A separate YAML file makes it trivial to include the actual plaintext, along with modal qualities (e.g., allowed modes, keys, scales, et cetera).
A new evolution of lambrOS is that it can take variable-length input, pass it through a hot-swappable primitive (e.g., sponge MAC) and the output of that becomes input to the harmonic engine (itself using the same primitive for deterministic repeatability). Hear below: plaintext --> MAC --> MIDI.
lambrOS + #NASA meteor data + #strangerthings (www.instagram.com/p/DTAcUxZAUWQ/)
Also lots of GUI variations in the works, but I'll likely release the CLI version first (Spring '26), with some concise documentation. The GUI version will have "helpers" to simplify building full configs from a template, and also give music theory guidance when turning dials.
This means output is now much, much truer to the context surrounding the input (e.g., if input is a basketball sequence, it doesn't treat all passes equally). Inspired by nonlinear cryptographic functions, it also presents some sense of entropy, but with intention, thanks to enhanced user-dials.
Lots of logic optimizations, as well as turning YAML configs into full-fledged adaptors — pushing musical intent out of code and into declarative structure, without changing how the system sounds. The "shapeshifter" function replaces memoryless probability with path-dependent constraint dynamics.
One cannot go overboard with potatoes.
Enjoy this visualizer of a MIDI created from in-game NBA2K telemetry around Kayla McBride and Kon Knueppel sequences and processed through midrangeOS' new "shapeshift" harmonic layer, inspired by block cipher design (S-box, Hamming distance, avalanche). Intermodal composition (Mixolydian, Dorian).
What if, like salmon in our final days, we all return to Usenet?
crypto[graphy] is dope .
lambrOS upgrades: tightening reproducibility proofs and now using BLAKE2b. Ultimately, lambrOS now operates like a PRF, but with output that needs to be both deterministic and aurally pleasant, and the ciphertext is MIDI.
And then there's the concept of harmonic collision, which would be cool to analyze: two different inputs from some input type with different entropy parameters resulting in the same chord progression.
This makes it frustrating to have a productive discussion about election security. There's no simple narrative. Yes, there are security vulnerabilities in some election systems, and that's genuinely bad. But no, there's no evidence any US elections have been stolen or rigged this way (so far).
And this captures the true intent of lambrOS, harnessing the entropy of real-world signals and human creativity in a way that produces shapeable, repeatable, and intentional musical compositions. It's the ability to temper the dissonance of unpredictability in an aurally palatable manner. (8/8)
This kind of analysis helps form proofs around lambrOS' ability to adhere to the rules we set in our YAML config, like, "gravitate towards E Mixolydian, but allow some % of inter-modal play with E Dorian and A Lydian." Entropy dials can create a little or a lot of movement within those bounds. (7/7)
That's good reinforcement for the applicable semantic HFF property. And then there's the Harmonic Inference Problem (HIP), which is kind of the Bayesian inverse of HFF. If HFF is “given data → produce harmony”, then HIP is “given harmony → infer what parameters produced it”. (6/6)
There's also the more cryptographic kind of HFF, which can take input directly from a PRF and reduce its "soundness" back to the indistinguishability of the PRF. No efficient observer can tell whether the sequence came from random or from lambrOS, within a tonal boundary. (5/5)
This is a kind of semantic HFF, where consistency (similar inputs/fixed params/similar outputs) or indistinguishability (due to dynamic parameters) is provable. It provides bounds around entropy's influence and preserves lambrOS' ability to allow controlled shaping rather than random garble. (4/4)
The more we toggle the entropy dials, the less distinguishable two similar inputs, or even the same input with different parameters, become. We get closer to the indistinguishability found in cryptographic PRFs, and can even make a similar reduction for HFF (Harmonic Function Families). (3/3)
Beyond that, we also want consistency; in other words, similar inputs with identical parameters should yield similar outputs. This implies maximum control by the dials; the dials are where the human shapes the outcome, and for a composer, this human control is central to lambrOS' design. (2/2)
As lambrOS evolves, I'm applying a lot of rigor around the consistency and fidelity of how it produces MIDI from real-world signals. The first step was reproducibility; even with multiple dials for introducing entropy, it's imperative that any output can be reproduced with fixed parameters. (1/1)
This is a kind of semantic HFF, where consistency (similar inputs/fixed params/similar outputs) or indistinguishability (due to dynamic parameters) is provable. It provides bounds around entropy's influence and preserves lambrOS' ability to allow controlled shaping rather than random garble. (4/x)
And this captures the true intent of lambrOS, harnessing the entropy of real-world signals and human creativity in a way that produces shapeable, repeatable, and intentional musical compositions. It's the ability to temper the dissonance of unpredictability in an aurally palatable manner. (8/8)
There's also the more cryptographic kind of HFF, which can take input directly from a PRF and reduce its "soundness" back to the indistinguishability of the PRF. No efficient observer can tell whether the sequence came from random or from lambrOS, within a tonal boundary. (5/x)
The more we toggle the entropy dials, the less distinguishable two similar inputs, or even the same input with different parameters, become. We get closer to the indistinguishability found in cryptographic PRFs, and can even make a similar reduction for HFF (Harmonic Function Families). (3/x)
Beyond that, we also want consistency; in other words, similar inputs with identical parameters should yield similar outputs. This implies maximum control by the dials; the dials are where the human shapes the outcome, and for a composer, this human control is central to lambrOS' design. (2/x)
New lambrOS feature: YAML now has toggles for printing chord voicings, musical notation, and tabs. One future goal is to have a post-production feature that scans output for legal-but-jarring chord voicing, and can auto-polish ("rounding up/down") or present the user with alternatives.
