Local utility
ExploringTokenTrim
A macOS image utility for reducing visual payload before it enters an AI workflow.
Local SwiftUI utility · measured examples · preview pendingPublic index · current work
Builds with a reason to exist.
A curated selection of systems and experiments from the lab. These are public because they have a clear use, a visible artifact, or a useful question behind them.
Featured Builds
The lab’s current questions.
Four active systems represent where TTL is working now: knowledge, memory, evidence-backed intelligence, and executable operating models.
Knowledge systems
BuildingKnowledge Library
A local-first pipeline for turning a book collection into structured, searchable knowledge.
Resumable phases · graph · retrieval →Agent infrastructure
BuildingFuli Memory Core
Inspectable local memory infrastructure for AI agents, with durable records and fallback retrieval.
SQLite · MCP · offline benchmark →Competitive intelligence
BuildingEnergyRadar
Energy-market monitoring that keeps screenshots, source pages, and changes attached to every insight.
Playwright · evidence store · diffs →Operating systems
BuildingSimTwin
A visual business twin for modelling organizations and running human and AI workflows with approval boundaries.
Graph model · Agent OS · Reality metrics →More From The Lab
Other credible systems and experiments.
On-device AI
BuildingArk
A private AI workspace for iPhone, using local models for chat and file analysis without cloud dependence.
SwiftUI · MLX Swift · case study →Writing tool
ShippingAIMd Studio
A privacy-first Markdown editor for focused writing with local storage, iCloud sync, and export.
Live product host · privacy policy →Learning system
BuildingMather
Adaptive mathematics practice for a real Portuguese 5th-grade curriculum, refined through study sessions.
SwiftUI · curriculum data · case study →Reading system
ShippingCognito
An AI reading companion for EPUBs and PDFs with contextual chat, semantic search, and private libraries.
iOS screenshots · architecture →Multimodal systems
ExploringXSKD
A local multimodal assistant experiment spanning vision, voice, and on-device model paths.
Measured latency notes · partial voice tests · preview pendingCuration rule
Small, public, evidence-backed.
The index favors work that is understandable externally, supported by a real artifact or page, safe to publish, and worth maintaining. Older experiments remain accessible through their existing URLs without being presented as current builds.