Tools & experiments
Some questions need building.
Alongside client work, we make tools and explore technical ideas of our own. They give us something concrete to test, teach and learn from.
01
A tool for exploring visual ideas
02
Products
Tools we have built and run. Each status label says where it actually stands.
Content · Live
Content Agents
Multi-agent pipelines that research, write, illustrate and publish to a live CMS on a cadence.
ViewVoice · Live
VoiceReady
Speaking practice that analyzes how you actually sound, not just what you said.
ViewIntake · Live
Conversational Forms
Fill a long form by talking or chatting while the structured form fills behind the scenes.
ViewWorkforce · Live
SkillsNav
Voice interviews that build a skills inventory, find the gaps and suggest what to train and who to hire.
ViewPlatform · In build
Trellis
A maintained framework underneath, AI-generated code on top, and a validation pipeline in between.
ViewAnalytics · In build
Semantic Signal
Measuring whether large language models cite a brand, and how, across roughly fourteen metrics.
View03
Research
Questions we are investigating, with the method and limitations written down.
Research · In daily use
Autonomous build systems
Can a specification be built by a fleet of agents under verification gates, with people kept where judgment lives?
ViewResearch · Active
Multi-model orchestration
When orchestrating several models beats using the best one alone, and how to tell.
ViewResearch · Active
AI visibility & search
A model-agnostic framework for measuring whether an LLM cites a brand, and why.
ViewResearch · Working architecture
Clinical triage architecture
Separating the model that talks to a person from the model that makes the decision. An architecture study, not a clinical product.
ViewResearch · Active
Language-emergence simulation
An agent-based simulator for how a shared language emerges, or fractures, between populations.
View04
Separate from client work
This section is for tools and experiments we can publish. Client systems belong in the work section, with the client’s approval and an accurate account of our contribution.
Have a question that needs an experiment?
Bring us the question and what an answer would help you decide.