Experimental systems
before they
mature
into product.
Labs is the experimentation layer where satellite demos, SAM 2 water-body pipelines, synthetic data prototypes, simulation, and early runtime patterns get pressure-tested before they graduate into product, research, or public-interest surfaces.
Choose a lab lane to inspect the current experiment surface.
Impact
Water intelligence is one visible public-interest program right now. Labs work matters when it sharpens sensing, forecasting, and public-trust workflows there.
Games & Simulation
Game environments double as systems research: mechanics, agent behavior, and player response become test beds for broader platform ideas.
Experimental
Runtime experiments, operator tools, and early concepts live here while Labs tests whether they can become product, research, impact, or learning artifacts.
Simulation
Digital twins, climate workflows, and physics-informed AI. Testing agent behavior, water-risk scenarios, market dynamics, and organizational models in controlled virtual environments.
Data
Synthetic data generation and intelligence pipelines for training, testing, and distillation, shaped into inspectable experiment outputs before they graduate into product lanes.
Multi-Agent Orchestration
Coordination patterns, runtime experiments, and team-level agent loops belong here when Labs is pressure-testing how many agents can work together coherently.
What Labs is pressure-testing now.
Labs publishes the experiments that are mature enough to name, but not always mature enough to ship. Each item has a current status and a graduation path into product, research, impact, or learning.
Homepage Chat Runtime Demo
A live-facing interaction pattern that lets visitors type into a product-style interface while seeing how agents coordinate behind the answer.
Graduates to PlatformRuntime Control
Operator review, agent-team status, validator posture, and trust-boundary inspection for autonomous work that cannot become invisible.
Graduates to Higher-trust planningGame Dev Empire
A management simulation used to test decision loops, stateful systems, and how agent behavior changes under visible constraints.
Graduates to Labs gamesSAM 2 Water Segmentation Baseline
A satellite-image workflow for water-body change detection using public imagery, visible caveats, and a defensible segmentation baseline before Observatory claims become product copy.
Graduates to Research + ImpactSynthetic Data Pipelines
Domain-specific generation, evaluation, and distillation workflows that create safer test surfaces before product data is involved.
Graduates to DataControl-Plane Evaluation
Benchmarks and method notes for when agent organizations need approval paths, visible state, and durable handoffs instead of hidden automation.
Graduates to ResearchRecent from the Lab
ALCUB3 Runtime V5
Universal 4-stage agent loop with memory layers, parallel tool execution, and auto-compaction now powering all division leaders.
Apr 2, 2026Edge AI: Gemma 4 Benchmarks
Local inference benchmarks on consumer hardware, focused on quantization tradeoffs, latency, and model quality at smaller parameter counts.
Apr 5, 2026Game Dev Empire: 112 Tests Passing
Management simulation reaches milestone with 66 files, 17K+ lines, and 45+ features across 14 industry verticals.
Apr 1, 2026Digital Twin Market Dynamics
Agent behavior testing in controlled virtual environments. Organizational models and market simulations running 24/7.
Mar 30, 2026Experimental work only matters if it changes the system.
Labs is the exploratory layer. The work becomes valuable when it sharpens the platform, improves Impact, strengthens Research, or turns into something people can actually learn through Learning Hub.
Runtime, memory, and controlled execution
Experimental systems land in AI Agent, AI Workers, and Higher-trust planning once they become stable enough to support real work.
See the core systemWater intelligence as public-interest systems work
Satellite demos, Observatory workflows, BasinKit, and applied sensing work move from Labs into Impact when they can support inspectable public value.
See ImpactMethods, benchmarks, and public evidence
When an experiment is ready for outside scrutiny, Research turns it into methods, reports, and evidence people can actually inspect.
Read the methodTeach the operating patterns
The guided layer turns experiments and validated results into practical pathways so people can actually learn the system instead of just admiring it.
Learn the systemWhere experiments go when they harden.
The point of Labs is not to accumulate demos. It is to move useful patterns into the right public surface with enough evidence, documentation, and UX clarity to hold up.
Early signals and controlled tests
New ideas start as explicit experiments with a question, input, output, owner, and a destination lane before they are shown as anything more durable.
Exploration surfaceSimulation and working proofs
Promising experiments become prototypes when they can demonstrate a workflow, data path, or interface pattern in a bounded test environment.
Prototype surfaceEvidence and publication objects
Validated experiments become method notes, benchmark reports, or source-backed publications in Research.
Publication surfaceRuntime and orchestration patterns
Chat runtime, memory continuity, and multi-agent coordination graduate into AI Agent, AI Workers, or Higher-trust planning once they can support customer work.
Product surfaceArchived or superseded explorations
Experiments that do not meet the evidence bar are retired or kept as archived learning so customers only see active claims where the proof supports them.
Archive pathMission-ready systems
Water intelligence, satellite-backed observation, field sensing, and public-interest workflows move into Impact when they can be inspected by partners and non-technical readers.
Mission surfaceLessons, paths, and projects
Patterns that people can repeat become courses, modules, and project-based learning paths inside Learning Hub.
Teaching surfaceFollow the work into Research or Learning Hub.
Research shows the evidence behind what worked, Learning Hub explains the patterns people can apply, and the product pages show what is available today.