About
The Lab
Library 451 is an independent AI lab. Not a consultancy, not a research group. A workshop where AI meets real operational problems and working tools come out.
The starting point was close to a decade in institutional real estate: fund management, portfolio oversight, investor reporting. The lesson from those years was clear: 80% of the work is mechanical. Reports, data reconciliation, compliance, document processing. The remaining 20%, judgment, strategy, the actual decisions, is where the value lives. AI doesn’t replace the 20%. It eliminates the 80%.
That wasn’t just an observation. It was a before-and-after moment. The kind of shift you recognize immediately. Like watching science fiction from the 90s quietly become real. Not someday. Now. The gap between what AI could do and what the world was actually using it for was enormous. And it was closing fast, with most people watching from the sidelines.
Library 451 was the decision to stop watching.
Conviction
The real measure of AI won’t be what it generates. It will be what it makes possible for the people who learn to direct it.
Trajectory
Early 2024
Fine-tuning
First experiments training custom models on Hugging Face. Sector-specific datasets, small-scale runs. Learning where fine-tuning delivers, and where prompting alone is enough.
Mid-2024
Applied automation
Scrapers, Telegram bots, n8n workflows. The question shifts from “can AI do this?” to “can it do this reliably, unattended, at 3am?”
2025
Systems
MCP servers connecting AI agents to real-world data sources. Prototypes that graduate to production. The projects stop being experiments and start being infrastructure.
2026 →
Autonomous agents
Autonomous agents that monitor, decide, and execute without human intervention. The lab becomes the factory.
The name comes from Fahrenheit 451, a novel about the destruction of knowledge. This is the opposite project. Building the tools that make knowledge actionable, and continuing because what’s happening right now in AI is bigger than any one industry, any one team, or any one use case. And it’s just getting started.