// JOURNAL — NOTES FROM THE STUDIO
Journal
What we learn building automation, web apps, and mobile apps — written plainly, for the people who hire us.
What Gemini 4 Argon means for software teams
Gemini 4 Argon lifts the output limit to 1M tokens at $2/$10 per million. What that changes for software teams shipping agents and automation.
Read →
How to build an MCP server for an internal tool
How to build an MCP server for an internal tool: a working FastMCP example, plus the auth, tool-catalogue and versioning decisions that come after it.
Read →
Apps for field teams: what actually gets used
A mobile app for field service teams lives or dies on offline behaviour and how few taps a job takes. What we have learned building them.
Read →
How long does it take to build an internal tool?
Six weeks to a first working version, and what actually consumes those weeks — plus the four things that make a build take three times longer.
Read →
How to choose an AI automation agency
How to choose an AI automation agency: the five questions that predict whether the project ships, and the red flags that predict it won't.
Read →
React Native or native: how we decide
React Native vs native app development, decided by four questions about your app rather than by framework preference.
Read →
What are System One models? TypeSafe's Jev, explained
System One models return typed decisions with calibrated probabilities instead of text. What TypeSafe AI's Jev is, and when to use it over an LLM.
Read →
When to replace a spreadsheet with an internal tool
The five signals that a spreadsheet has outgrown itself — and the cases where the honest answer is to keep it.
Read →
Context engineering for AI agents: why yours forgets
What is context engineering for AI agents? Curating what the model sees at each step — and why most 'our agent got dumber' complaints trace back to it.
Read →
GPT-6 Astra can use a computer. Should it use yours?
When computer use AI agents for business automation beat an API integration — and the many times a plain integration still wins.
Read →
Why automation projects fail (it's not the AI)
Why do automation projects fail? Usually because they automate a broken process — which just fails faster. How to find the real bottleneck first.
Read →
The review queue: keeping humans in the loop
The pattern behind every automation we ship: let software handle the 90% it's sure about, and give a human a fast, boring queue for the rest.
Read →
