Notes from the studio.
Working notes from an AI-native team: the tools, the agents, the evals, and the workflows behind the things we ship. Receipts, not marketing.
- Dev Tools
Claude Code vs Cursor: we use both, and here's the split
An editor with an agent in it, versus an agent that lives in your terminal. They are different shapes of tool — here is how we decide which task goes where, and what we would tell you to pick if you only get one.
Read post· 9 min - Agents
Treat the context window like a budget, not a bucket
Most agent failures we debug are not reasoning failures. They are context failures. Here is how we budget the window at Replace Works.
Read post· 3 min - Studio Notes
Why Our Tool Catalog Keeps Moving
A note on why Replace Works treats its tool catalog as an operating system, not a trophy case.
Read post· 1 min - Evals
Write the eval before you write the prompt
At Replace Works we stopped tuning prompts by vibes. Here is the small eval-first loop we run before any agent ships.
Read post· 3 min - Studio
Welcome to the Replace Works blog
A home for the applications we ship at the studio and the AI tools that power them.
Read post· 2 min