Portfolio Projects

Personal projects, designed and built end-to-end solo.

Each project below is a personal portfolio piece — built to demonstrate the skills on my resume. None are commercial products, businesses, or services for sale.

Schwarz-D AI Routing

A model-routing engine that picks the cheapest AI model good enough for each request — using the Schwarz D minimal surface as the cost/quality decision boundary. Live on this site, fully tested.

Problem. AI spend balloons because most requests get routed to a frontier model "just in case." A trivial regex classification burns the same credits as a state-space search.

Approach. Six stacked layers — regex pre-filter, truth-table constraints, logic gates, decision-tree shortcuts, Tiresias manifold demand (z derived from x and y), Schwarz-D surface threshold — plus a determination graph that resolves multi-step requests into a DAG of independently routed nodes.

Outcome. 23/23 passing unit tests. Live demo with interactive sliders. Downloadable policy JSON. Generic tiers in source, named real-model mappings in catalog.example.json.

  • Vanilla JS, zero runtime dependencies
  • Schwarz D minimal-surface math (triply-periodic)
  • Truth-table constraint engine with explicit AND/OR/NOT/IMPL gates
  • Regex pre-filter for zero-token flag inference
Open the router

KensGames.com

A full 3D game platform with deterministic simulation, real-time rendering, multiplayer orchestration, and procedural music synthesized live from manifold equations. Solo-built and maintained end-to-end.

Problem. Build a portal of polished browser games without a framework tax, without audio assets, and without per-game engineering drift.

Approach. Vanilla JS + WebGL2 + Web Audio. Every substrate (board, audio, AI, lobby) owns a single domain and communicates by publishing coordinates to a shared registry. Adding a game is adding a manifold descriptor, not a codebase.

Outcome. Eight shipped games on one substrate-driven rule engine (fasttrack, 4DTicTacToe, brickbreaker3d, starfighter, cubic3d, assemble, and more). 60fps target hit on mid-range hardware. Zero audio files in the repo.

  • Vanilla JS · WebGL2 · Web Audio API
  • Node.js auth/lobby server · PM2 · nginx · Ubuntu VPS
  • GitHub Actions CI/CD · rsync deploys
  • SQLite records layer for transactional state
Visit kensgames.com

TTLRecall.com

Rapid-build challenge platform leveraging AI-accelerated workflows for scenario generation, scoring, and feedback.

Problem. Show how fast a single developer can stand up a complete mock business platform when AI is treated as a power tool rather than a chatbot.

Approach. Directive-driven generation pipeline. Each scenario is produced, scored, and verified by a separate agent before reaching the user.

Outcome. Two complete platforms stood up in roughly 30 minutes each — including infrastructure, content, and feedback loop.

  • AI workflow automation
  • Scenario engine with verifier agents
  • Real-time scoring and feedback
Visit ttlrecall.com

Personal Search Engine App

A manifold-based search engine using the z derived from x and y substrate for indexing and retrieval. Demonstrates universal applicability of the manifold paradigm to information retrieval.

Problem. Traditional search engines rely on complex indexing structures and heuristic ranking. Can a simpler, mathematically-grounded approach achieve comparable results?

Approach. Use the manifold operation (z derived from x and y) as the fundamental indexing and retrieval mechanism. Documents are indexed as identity-modifier pairs, and search queries are resolved through manifold extraction.

Outcome. Demonstrates that the manifold paradigm applies beyond games and AI to information retrieval, with the same mathematical foundation (z derived from x and y) producing consistent, predictable results.

  • Manifold-based indexing (z derived from x and y)
  • Universal applicability to information retrieval
  • Same substrate as KensGames and dimensional programming
Learn about the manifold

DimensionalProgramming.com

The paradigm site — the canonical writeup of dimensional programming as a discipline, with the full breath-verb algebra and the deploy pipeline.

Problem. Most "AI-aligned" architecture writing is metaphor. There is no shared vocabulary for the structural choices that actually make AI systems behave.

Approach. A small set of primitives — identity, lens, delta, manifold, directive — with worked examples, a reference implementation (Python stdlib, append-only SQLite ledger), and the equation that grounds the whole thing: z derived from x and y.

Outcome. A reference writeup that grounds the doctrine used across the other portfolio pieces.

  • Dimensional programming as a discipline
  • Breath-verb algebra (fold / bloom / quadratic)
  • Append-only ledger ≡ manifold
Visit dimensionalprogramming.com

ButterflyFX.us

This site — an extended portfolio and working artifact of the directive engineering and dimensional programming on my resume.

Stack. Vanilla HTML/CSS/JS. The animated background is a single raymarched fragment shader rendering the Schwarz-D minimal surface (and a gyroid blend), reactive to mouse, scroll, and the routing-page sliders. No frameworks, no Three.js, ~150 lines of pure WebGL2.

Notable. The Ultimate AI Directive page is the canonical writeup of the manifold-based reasoning framework these other projects share. Every page is deployed via the same GitHub Actions + rsync pipeline as the rest of the portal.

  • Vanilla HTML / CSS / JS · WebGL2 raymarcher
  • Schwarz-D / gyroid triply-periodic minimal-surface shader
  • Static deploy via GitHub Actions to VPS
Read the Directive

Open Source — github.com/kenbin64

Public repositories spanning AI tooling, agent design, dimensional programming experiments, and game-engine work.

Includes the source for the kensgames portal, ButterflyFX, and standalone libraries and tools.

  • Multi-language (Java, .NET, Python, JS/TS)
  • Public commit history
View on GitHub

Substack — @kbingh

Technical writing on AI architecture, directive engineering, dimensional computing, and the structural choices that make AI systems behave.

  • Long-form essays and worked examples
Read on Substack