Skip to content
RosmarOps
  • Blog
  • About
  • Observatory
  • Sandbox
  • RosmarOps, the product
  • ES
  • CA

Blog

The AI agent map: what each one can do without asking you

November 17, 2026 by Albert Garcia Pujadas

I measured nine AI agents as they come installed. The Observatory map shows how much each one can do without asking you, and why.

Categories RosmarOps news Tags Claude Code, Cursor, Observatory

Your AI agent’s sandbox: it almost always has one, it’s almost never on

November 16, 2026November 16, 2026 by Albert Garcia Pujadas
A small figure plays in a sandbox whose walls lie flat on the ground; the sand spills out towards some keys, some folders and a half-open door.

We saw it going through the documentation of the nine agents RosmarOps evaluates. And something came up that worries me more: turning it on isn’t always enough.

Categories RosmarOps news Tags Sandbox

What’s new in RosmarOps: nine agents with a score, your projects across your whole folder and fewer interruptions

November 14, 2026 by Albert Garcia Pujadas
A laptop showing a map of nine AI agents, a hand placing a new one and, beside it, two laptops connected to a cloud.

RosmarOps now scores nine AI agents, looks for your projects across your whole folder, only interrupts you when something changes, and tells you inside Claude Code. Here’s what you get with 0.11.6, and what it still doesn’t do.

Categories RosmarOps news Tags Claude Code, Codex, Cursor

RosmarOps is now on the Microsoft Store

November 13, 2026 by Albert Garcia Pujadas
RosmarOps page on the Microsoft Store, with its download button and the PEGI 3 rating.

Microsoft has approved RosmarOps on its Store. On Windows it now installs in one click and without the blue «Windows protected your PC» window. Here’s how it went, and what 0.11 brings.

Categories RosmarOps news Tags Microsoft Store, Windows

The real work starts when the agent says “I’m done”

November 12, 2026 by Albert Garcia Pujadas
An AI agent holds up a sheet with a tick while a hand checks another sheet with a magnifying glass and finds an error.

An AI agent tells you it’s done. How do you know it’s true? What I’ve learned building UXMachine, and a few questions to check your agent’s work even if you’re not a programmer.

Categories Loop engineering Tags Claude Code, Codex

AI doesn’t learn from its mistakes. Your system should.

November 11, 2026 by Albert Garcia Pujadas
An AI agent trips three times over the same pothole; further on the pothole is filled and a hand places the last block of a wall.

Your agent doesn’t remember it got something wrong yesterday. If you want it not to happen again, what has to change is the system it works in. How to turn a repeated mistake into a change that stays, and why writing it in a file isn’t always enough.

Categories Loop engineering Tags AGENTS.md, Claude Code

Every agent mistake, a piece of the system: harness engineering at RosmarOps

November 9, 2026 by Albert Garcia Pujadas
An AI agent moves along a track towards five arches: the first is finished and the other four are only outlined. Behind it are several red marks, joined by lines to pieces already placed beside the track, and a hand sets down the next one.

I build, with AI agents, a tool to use AI agents more carefully. This week its bugs were the same ones RosmarOps looks for in yours. This is how we turned each one into a permanent piece of the system.

Categories RosmarOps news Tags AI-First, harness engineering

Loop engineering: letting agents work on their own doesn’t mean they work better

November 8, 2026 by Albert Garcia Pujadas
An AI agent at the centre of a loop of documents going round, with a hand checking one of them with a magnifying glass.

A study examined 36,487 software repositories and confirmed autonomous AI agent loops in 217. More interesting than the percentage is what we know —and what we still don’t— about the controls and results of those systems.

Categories Loop engineering Tags Claude Code, Codex

What wasn’t read doesn’t count as clean

October 31, 2026 by Albert Garcia Pujadas
Four file cards drawn clearly and one that is only a dashed outline, with a magnifying glass over the empty one.

Why RosmarOps refuses to give a score when it couldn’t read something, and why hiding what it reads doesn’t improve it.

Categories Agent Posture Score Tags Agent Posture Score, coverage, methodology

Tools that check what your AI coding agents can reach (October 2026)

November 4, 2026October 29, 2026 by Albert Garcia Pujadas

Four open-source tools check the setup of AI coding agents on a developer’s machine.

Categories Agent Posture Score Tags Claude Code, MCP
Older posts
Page1 Page2 Next →

Recent posts

  • The AI agent map: what each one can do without asking you
  • Your AI agent’s sandbox: it almost always has one, it’s almost never on
  • What’s new in RosmarOps: nine agents with a score, your projects across your whole folder and fewer interruptions
  • RosmarOps is now on the Microsoft Store
  • The real work starts when the agent says “I’m done”

© 2026 RosmarOps

© 2026 RosmarOps • Built with GeneratePress

Privacy · Legal notice · Terms · Cookie settings