High-signal intelligence for people building with AI.
The warmer editorial mode of TTL: models, agents, infrastructure, and the practical implications worth carrying into the lab.
Read Obsolete AITiny Little Lab builds applied AI systems, runs experiments, publishes what it learns, and turns useful work into practical products.
A small public selection from the lab. Each project is tied to a concrete question, artifact, or running system.
A local-first pipeline for turning a book collection into structured, searchable knowledge.
Evidence · resumable phases, graph, and retrieval →Inspectable local memory infrastructure for AI agents, with durable records and fallback retrieval.
Evidence · SQLite, MCP, and offline benchmark scaffolding →Energy-market monitoring that keeps screenshots, source pages, and changes attached to every insight.
Evidence · Playwright capture and evidence-backed diffs →A visual business twin for modelling organizations and running human and AI workflows with approval boundaries.
Evidence · graph model, Agent OS, and Reality metrics →Research is not a separate publishing business. It is the trail of experiments, benchmarks, and intelligence produced while building.
The warmer editorial mode of TTL: models, agents, infrastructure, and the practical implications worth carrying into the lab.
Read Obsolete AIThe latest Forge research tracks contracted AI infrastructure, capital concentration, and what the shift means for small operators.
Read the evidenceTTL products are practical assets extracted from real systems and research. The current catalogue is being reworked; there are no active checkout offers here.
Forge is the continuously generated daily record of TTL work and research. The feed below reads from the existing latest-deliverables contract.
Build before theorizing. Stay local and open when practical. Prefer evidence over hype. Publish useful failures as well as wins.