Rohan Varma
@rohanvarma
Building systems that remember, learn, and develop continuity over time.
3,741 subscribers
Memory Is More Than Context
Why persistent agents need structured experience, not endless transcripts
Building Systems That Remember
A practical architecture for long-running AI agents
What Comes After the Chat Window?
From episodic answers to software with continuity
The Coding Agent Landscape
Local terminals, cloud workers, and where each model fits
When Should You Self-Host Inference?
A decision framework for APIs, managed deployments, and GPUs
Agents Need Boundaries, Not Just Tools
Why capability without constraints creates brittle automation
The Cost of Remembering Everything
Retention, retrieval, and the discipline of forgetting
Evaluations for Long-Running Agents
Measuring continuity, recovery, and useful autonomy
A World Model Is Not a Database
Representing change, uncertainty, and cause inside an agent
Why Agents Fail Quietly
The hidden state transitions behind plausible mistakes
Designing Human Checkpoints
Where review adds safety without destroying momentum
From Prompt to Operating Loop
The architecture that turns a model into a persistent process