Architecture
Shared principles. Distinct designs.
Frankenstein, the energy proofs of concept, and Resactly share principles, not a single implemented engine. Each is designed for its own domain and is at its own stage.
Shared principles
- 01
Context
Carry forward the information that shapes a decision or a task.
- 02
Explicit assumptions
Keep inputs and assumptions visible and adjustable.
- 03
Human oversight
People stay in control of consequential actions.
- 04
Initiative-specific designs
Each initiative is designed for its own domain; there is no single shared engine.
Frankenstein
Current capabilityIdentity and accumulated context persist outside any single underlying model. The model provides cognition; the parts below can evolve independently without erasing continuity.
- Memory
- Durable, correctable memory carried across conversations.
- Project knowledge
- Ongoing work and its context, available in later sessions.
- Preferences
- How its creator likes to work, retained over time.
- Tools
- Web access, repository inspection, production-log investigation, controlled coding workflows — permissioned and bounded.
- Interfaces
- A production conversational interface, including voice.
Physical embodiment is a future ambition, not a current capability.
Explore Frankenstein →Energy
Under evaluationThree proofs of concept, each evaluated separately. Their usefulness depends on the quality of source data and on assumptions that teams can see and adjust. A connected planning process is a potential direction; today the POCs remain separate evaluations.
Location intelligence
Evaluating potential data center locations against regulations, power availability, redundancy, and infrastructure constraints.
Energy planning, cost forecasting & RFP response support
Exploring how disparate project information can be assembled into an energy plan, with supply-option comparisons, projected costs, and supporting material for an RFP response.
Data center configuration mapping
Exploring how a data center’s setup, redundancy, and backup requirements shape its infrastructure needs and costs.
Resactly
Intent capture, menu and context data, and recommendation are built and running in the live dining product. Measuring outcomes and learning from them are planned.
Resactly's design in detail →