Every domain agent and intelligence site built on the Bulkhead τ substrate.
Each domain is a deterministic agentic cockpit, gated by its own charter and grounded in real data.
Detect logical fallacies — 48 fallacies, 15 case studies, 11 meta-principles, 17 counter-strategies.
Garmin activity and wellness analysis from GDPR export data. 32 tools across 4 variations, with TennisAgent integration.
MediaPipe pose landmarks overlaid on a real serve clip, frame by frame — pixels to pose directly, no sensors or simulation.
Phone video of a serve with Zepp handle IMU and LabWired peak-detection firmware edges overlaid — not a simulated 3D body.
Forehand and backhand clips with MediaPipe pose overlays. Serve demos stay separate; stroke-mode has no LabWired origin stream.
The same tracked pose, isolated on an open field instead of drawn over video — a motion study, joints only.
Both overlays on the same 90 frames — MediaPipe skeleton and the Zepp/LabWired telemetry panel, one render pass, not two videos stitched together.
Golf swing analysis using the Zepp Universal sensor. 31 tools across 4 variations with performance profiling.
Personal tennis analytics cockpit. 189 tools for multi-sensor analytics across Apple Watch, Zepp, Babolat, and Garmin.
Repeatable tennis answers — tour-level reference lookups.
Parable-driven reasoning domain.
Plant-based recipe domain with progressive tools for lookup, planning, discovery, and personalization.
Arizona water report domain prototype.
Stanford Stats 315A Statistical Learning coursework domain, based on Elements of Statistical Learning (ESL II).
OSAT market analysis, Arizona facility thesis, advanced packaging trends for Amkor (AMKR).
OSAT advisory prototype.
Honeywell landing page with Bulkhead τ design principles and links to the full domain site.
Standalone Honeywell landing page for product visibility, revenue context, and executive dashboards.
Deterministic medical device QMS compliance workflows. 23 tools across 4 variations.
CPU/GPU-first semiconductor market intelligence with deterministic analysis workflows.
Geo-matrix companion hub for the Semiconductors domain.
"Trading at the Close" ML pipeline with deterministic feature engineering, model CV, and drift monitoring.
Geo-matrix companion hub for the Optiver domain.
Quantum Computing Inc. product analysis and sales intelligence.
IC manufacturing yield analysis with a physics-based optimization model and deterministic agentic cockpit.
Regulated cardiac device performance reporting — 92 million US implants, 18 years, government-mandated data.
Agentic cockpits for the WorldQuant University Applied Data Science Lab curriculum. 41 tools across 7 projects.
WorldQuant University AI Lab curriculum — deep learning for computer vision.
Financial portfolio analysis with tax optimization, options trading, and risk assessment.
Strategic solutions demo-consulting landing page.
U.S. macro conditions, fiscal position, and world-economy linkages.
Climate data analysis — 93 tools across 8 variations covering temperature, CO2, methane, sea level, and Arctic ice.
Meta-agent orchestrating 14 production domains with 649+ tools via intelligent routing.
Why strong agentic coding systems still collapse in familiar ways. Eight failure families, six evidenced from Bulkhead τ operations.
Papers 1.16 & 1.19: terminal capture corrupts protocol scores, and stronger models can override literal substrate inspection.
Semiconductor fab yield simulation using the Rough Volatility Hypothesis — defectivity as a dynamic path, not a static mean.
Detailed failure-mode results from Bulkhead τ boundary testing.
Which roles a local model is fit for, and which it isn't.
The useful unit is not the naked local model — it is the harnessed domain system.
Paper 1.17: how OpenClaw wraps Bulkhead τ as an operator access layer without replacing it as the authority.
OpenClaw as the operator wrapper over Bulkhead τ: access, compression, monitoring, and workflow discipline without crossing the authority boundary.
Harnesses, routing, repair, and the operating regimes where system shape settles the result before model size does.
Bulkhead τ prototype domain.
Rough volatility modeling used as a machine-learning benchmark.
What a prepared local 3090 setup can still do offline — a harness boundary paper.