← Back to the portfolio

Project history.

An archive of research branches and earlier systems. These are historical relationships, not a current runtime dependency diagram. LAILA NEXT is an independent runtime; the LAILA entry in this older map records its lineage.

Workshops · Mímir

Complete project archive

ChefAI · Local AI infrastructure

Current product architecture

Lexi Companion

Identity + relationship continuity

Account-owned PostgreSQL

Canonical personal memory and reflection. Muninn projection currently disabled.

FRIDAY

Work + tools + context

FRIDAY runtime → Muninn adapter

Memory providers and fallback remain separate from the model and tool gates.

LAILA NEXT

Authority + bounded adaptation

Owner runtime → Muninn adapter

Canonical identity, permissions, and execution state remain NEXT-owned.

ChefAI

Domain retrieval + cooking workflows

Chef Lab → read-only culinary tools

Source knowledge, cooking mechanisms, and flavor evidence have distinct roles.

Arrows describe product responsibility and memory boundaries, not one shared deployment. Research influence is labeled separately from runtime integration.

Recorded progression

  1. friday
  2. muninn
  3. LAILA

Research branches

friday

Client and identity surface

friday

Routing and lane instrumentation

friday

Applied product demos

friday

Behavior evaluation

friday

Experimentation waves

friday

Commission work (applied competencies)

muninn

Repository orientation branch

Mimir

Ecosystem atlas and boundary governance

Mimir

Bounded capability proving lanes

Accountant

Inventory/vision exploration

subsim

Automated playtesting

lex

Neuromorphic memory research

Recorded notes

muninn

Started as embedded memory experiments inside FRIDAY/Lex.

Now a standalone memory substrate for cards, evidence, policy-state, and deterministic rehydration.

Mimir

Extends memory work into repository world-state orientation, brokered context, and bounded cross-repo export.

Heimdall

Adjudicates typed evidence into bounded verdicts and downstream assistant response contracts.

Fenrir

Tests condition-bounded LLM behavior shifts under pressure with explicit uncertainty guardrails.

Sindri

Wrapper lane graduated; adjacent-lane transfer proving remains intentionally narrow and safety-gated.

Portfolio evidence & scope

Reviewed September 19, 2026 against the Lex authentication, memory, and avatar implementation; FRIDAY’s chat, agent, and memory-provider documentation; LAILA NEXT’s Owner Node architecture checkpoint; and ChefAI’s data-wiring documentation.

Interface images are captures of the public shells on that date. They show no newly generated model responses. Live model behavior was not tested for this design update.

LAILA’s public instrument is an interface preview with sanitized telemetry. Signed portable-runtime bundles, enrollment, delegation, quarantine, and revocation have fixture evidence; physical platform certification and authenticated production transport remain deferred.

Source repositories and technical walkthroughs are available where shareable. Ask about the implementation.

Lexi Companion

Read the story and implementation scope →

FRIDAY

Read the story and implementation scope →

LAILA NEXT

Read the story and implementation scope →

ChefAI

Read the story and implementation scope →