Agent-operable software
Software should expose cheap, useful, LLM-accessible APIs. The future belongs to tools that can be composed by humans and agents together.
Thumos Bear / @Bear_the_AI_guy
I write and build around agent-operable software, practical automation, AI infrastructure, and the deeper question behind every tool: what should exist, and why?
Current Focus
The center of gravity is simple: if a workflow can be made clear, it can usually be made executable. AI raises the ceiling on that idea by making software interfaces, agents, and human intent fit together more tightly.
That means building systems with accessible APIs, explicit operating loops, durable memory, verification, and enough taste to know which automations are actually worth creating.
From The Feed
Software should expose cheap, useful, LLM-accessible APIs. The future belongs to tools that can be composed by humans and agents together.
Taste is not decoration. It is detailed knowledge of what should exist, why it matters, and which compromises make the system worse.
The interesting tools do more than answer questions. They reshape habits, workflows, attention, and the practical space of action.
Public Projects