🛠️ Devlog #37

Wave 17: Number Theory, Statistical Mechanics, Neuroscience and Search 2.0

📅 June 2029 ⏱ ~9 min read 🌊 Wave 17 Retrospective

Wave 17 pressed into some of the most technically demanding territory we’ve covered to date: the Riemann hypothesis and post-quantum lattice cryptography, renormalization group fixed points and universality classes, and the Hodgkin-Huxley equations that describe how every action potential in every brain on earth is generated. On the infrastructure side, this wave brought a complete rebuild of the platform search system — replacing keyword substring matching with an inverted-index model, adding tag faceting, and introducing a three-tier difficulty classification across all 345 simulations.

Platform Stats

345 Simulations
75 Categories
107 Blog Posts
17 Content Waves
35 Spotlights
27 Learning Posts

Wave 17 Content

🔢

Spotlight #34 – Number Theory & Cryptography

From the prime number theorem and Riemann’s zeta function through RSA key mechanics, elliptic curve group law, NIST PQC finalists (CRYSTALS-Kyber/Dilithium), and interactive Schnorr zero-knowledge proofs — the mathematical foundations that secure every TLS connection. ~13 min.

📖

Learning #27 – Statistical Mechanics & Phase Transitions

Partition functions and Boltzmann statistics, the exact 2D Ising solution (Onsager 1944, critical exponents β = 1/8), Landau order-parameter theory, Wilson’s renormalization group and universality, the fluctuation-dissipation theorem, and Metropolis & Wolff Monte Carlo near the critical point. ~14 min.

🧠

Spotlight #35 – Neuroscience & Neural Circuits

Hodgkin-Huxley conductance equations and action potential generation, quantal synaptic transmission and short-term plasticity, Hebbian learning and spike-timing-dependent plasticity, Wilson-Cowan neural mass oscillations, connectome graph theory, and The Virtual Brain whole-brain simulation framework. ~13 min.

Search 2.0: Inverted Index Architecture

The original search system performed a linear scan across all simulation titles and descriptions on each keystroke, which worked adequately at 50 simulations but became noticeably slow at 345. More importantly, it could not rank results by relevance, support multi-term queries correctly, or filter by structured metadata. Search 2.0 addresses all of these limitations.

Inverted Index

At build time, all simulation metadata is tokenised: titles, descriptions, category names, and tags are split on whitespace and punctuation, lowercased, and written into a searchIndex.json where each token maps to a posting list of simulation IDs. At query time, the user’s input is tokenised identically; for each query token, the corresponding posting list is retrieved and the intersection (AND) or union (OR) of IDs is computed in O(k log n) time where k is the posting list length.

🔍

Full-Text Search

All 345 simulation titles, descriptions, tags, and category names indexed into a token → ID inverted index. Sub-10 ms query latency at full scale.

🏷️

Tag Faceting

Tag chips rendered as toggle filters that narrow the result set by intersection. Multi-select facets allow pendulum AND chaos combinations.

📊

Relevance Ranking

TF-IDF-inspired scoring: title matches weighted 3×, description 1×. Results sorted descending by score; ties broken by simulation date.

Instant Results

Results update on every keystroke with 16 ms debounce. Zero network requests: the entire index is loaded once as a static JSON file.

Schema Changes

Each simulation record in simulations.json now carries two new fields:

The difficulty field drives a three-tier badge system visible on simulation cards and in search results. The tag field populates the facet side-bar and improves full-text recall for domain-specific queries.

Simulation Difficulty Classification

Beginner

No prerequisites beyond secondary school physics or mathematics. Interactive controls encourage exploration. ~40% of simulations.

Intermediate

Familiarity with calculus, basic mechanics or introductory programming helpful. Parameters have physical significance. ~45% of simulations.

Advanced

University-level mathematics or physics assumed. Models involve PDEs, statistical mechanics or quantum formalism. ~15% of simulations.

Learning Paths

One of the most requested features has been curated sequences of simulations that build conceptual understanding progressively. Six initial learning paths are now live:

Each path page displays a horizontal stepper showing the user’s progress (stored in localStorage), links to each simulation, and a brief rationale for why that step follows from the previous one. Paths are defined in a new shared/data/learning-paths.json file and rendered client-side.

Search Performance Audit

Wave 18 Preview

Spotlight #36

Plasma Physics & Fusion

Tokamak confinement, Lawson criterion, MHD stability, ITER engineering and the physics of Z-pinch and inertial confinement fusion.

Learning #28

Topology & Manifolds

Surfaces, genus, Euler characteristic, fundamental groups, covering spaces, fibre bundles, and the classification of compact surfaces.

Spotlight #37

Materials Science & Dislocations

Crystal defects, Burgers vector, dislocation motion and strain hardening, phase diagrams, and semiconductor band engineering.

Devlog #38

Fuzzy Search & Mobile UX

Edit-distance fuzzy matching, phonetic name matching, improved touch controls across all simulations, and PWA offline reliability improvements.

← Previous Spotlight #35 – Neuroscience & Neural Circuits Back to Blog → All Posts