Experiment
Test the question and define what useful evidence would look like.
Experimental systems online
Kingston Labs
Where financial AI ideas face real operating pressure.
Benchmarks, agent systems, prompt experiments and workflow prototypes—tested in public before the strongest ideas become part of Kingston.
Lab telemetryLive queue
Active lab tracks08
Evaluation stages05
Released intelligence stream01
How ideas move
Not every experiment should become a product.
Labs makes uncertainty visible. Each idea moves through a clear stage so visitors can distinguish early exploration from something released and usable.
Prototype
Turn the idea into a working interaction or controlled workflow.
Testing
Challenge reliability, usefulness and failure conditions on finance tasks.
Release
Move the strongest signal into Kingston with a clear operating purpose.
Experiment queue
Inspect what is moving through the lab.
LAB / 01Testing
Model benchmarks
Model Benchmark: Financial Document Q&A
Comparing how each model answers direct questions against a shared set of financial documents.
Evaluation in progress
LAB / 02Prototype
AI agents
Multi-Agent Research Assistant
A prototype that splits a research question across multiple models and reconciles the results.
Evaluation in progress
LAB / 03Experiment
Prompt experiments
Prompt Chaining for Risk Memos
Testing whether chained, narrower prompts produce more reliable risk memos than one large prompt.
Evaluation in progress
LAB / 04Beta
Beta tools
Kingston Copilot
An AI assistant that runs a Kingston workflow end-to-end and keeps the steps organised.
Evaluation in progress
LAB / 05Beta
Beta tools
Kingston Dashboard
A unified workspace surfacing model comparisons, active workflows, and research updates together.
Evaluation in progress
LAB / 06Testing
Community testing
Community Prompt Testing Pool
A shared pool where community members test draft prompts before they're added to the library.
Evaluation in progress
LAB / 07Prototype
Prototype workflows
Workflow Auto-Suggest
Suggests the most relevant Kingston workflow based on a short description of your task.
Evaluation in progress
LAB / 08Released
Research notes
Weekly Model Benchmark Notes
Short, dated notes tracking how model updates shift performance on financial tasks.
Available signal
Put a finance-AI claim under pressure