Jonata Santos is a software entrepreneur who publicly describes building a portfolio of products at Olympus. He shares practical agent-development workflows, including how he preserves research and evaluates an agent’s work before accepting it.[1][2]
Start with Agent development workflow →
X Career
Ideas
In his own workflow post, Santos describes hosting an agent on a virtual server, adding a persistent memory store and accessing it from desktop and phone. Before implementation, he has the agent research the subject and save the source material, then surface unknowns and define evaluations that distinguish a successful run from an unsuccessful one. The agent iterates against those evaluations and saves what it learns. This is a documented personal recipe, with tool and infrastructure prerequisites, rather than a comparative test of productivity.[2]
Tools, writing & talks
Writing, research & talks
Start here
Santos’ checklist for research, durable memory, explicit evaluations and repeated implementation.
Start with Santos’ own workflow post for the order of operations: research, preserve context, define evaluations, then iterate. It assumes command-line familiarity, a server and access to the named agent tools; treat the stack as his recipe rather than a performance test.
Author[2]- Just Talk To It - Agentic Engineering Guidesteipete.me
A practical, no-hype guide to working effectively with AI coding agents for software development. Covers real-world workflows, tool comparisons, and best practices for managing multiple agents in parallel.
Peter Steinberger essayOriginal source