Agents
Don’t Make the Agent Run Every Step
AI automation works more cleanly when routine operations stay in code and agents enter only where the work calls for judgment.
Episode 67 guest
Anthony
Anthony is an independent researcher who left the corporate world in 2024 to work in Bitcoin full time.
About
Anthony is an independent researcher who left the corporate world in 2024 to work in Bitcoin full time. He an alumni of Chaincode’s 2025 ₿OSS program and holds a PhD in computer science from the University of South Australia.
Conversations
Ideas from the conversation
AI Models & Infrastructure
Hosted AI Without the Hardware Hurdle
Hosted frontier AI reaches small consumer tasks more easily because local capability begins with a meaningful hardware hurdle.
AI Models & Infrastructure
The bandwidth limit on local AI
A local AI system needs enough memory to hold a model, but its bandwidth helps determine how quickly that model can produce an answer.
Agents
When Agents Need a Shared Layer
A common layer can coordinate several agent sessions through board tasks, hourly catch-up work, defined business scopes and deliberately uneven access.
Business & Markets
When Bigger Competitors Help
In an unfamiliar category, larger rivals can create an opening for a smaller company by teaching buyers what the category is before it presents a capability they can't match.
Privacy & Security
When Wingman's encrypted sync got too complex
Pete considered dropping Wingman's encrypted record sync after its need for a live key created a difficult design choice inside his self-hosted app.
Knowledge & Memory
When Your News Builds a Memory
A personal news system becomes more useful as it turns individual stories into a connected history that can resurface context without creating another reading queue.