dennett
AI-native consulting engine for single-view-of-client
The Shiny
dennett centralizes client data from files, internal notes, and transcripts into a single coherent view. It enables AI-native consulting by productizing Intentional’s internal IP — frameworks, knowledge, and approaches — rather than using LLMs for generic outputs.
The Hard Stuff
The key innovation is a granular memory model with agent-driven iterative updates, inspired by OpenClaw but with a stronger security model. Combined with a Claude Code-inspired skills/modalities approach, dennett demonstrates what post-December-2025 frontier capabilities at Anthropic make possible. Powered by Claude Opus.
Built With
Claude Opus, Claude Code, Custom Memory Architecture
What we learned
Beyond benchmark performance, we found that Claude 4.5+ models handle multiple distinct instruction and rule sets in context with notably better separation. This makes it possible to layer instructions, memory sets, and persona constraints without the 'polluted context' problem that large-context models have historically suffered from — a prerequisite for the kind of agent architecture dennett relies on.
What's on your mind?
We'd love to hear from you.
Reach out to our team at Hello@intentional.team
