33 articles tagged with 'AI'

We’ve spent a few weeks with personal AI assistants that work in the background, connected to the tools we use to run MarsBased. The most useful thing so far: they flag what we didn’t ask them to look for.

Decision models could make AI practical in applications where cost and complexity have kept it out. Here’s what they are and why we’re keeping an eye on them.

AI is making code cheaper and faster to write, but building software that actually works, scales, and stays secure still takes broad expertise and someone accountable for the outcome. Here's how we see the role of development agencies changing in the AI era, and how MarsBased is adapting to it.

We ran Claude Design through its paces by importing one of our client's Figma design systems. The import worked, prompting took a few tries to get right, and the export options left us with some questions.

AI gathers review evidence better than any human, but lacks context. Here’s why human judgment still drives our final performance ratings.

Why multi-repo setups are vital for Claude Code, and their trade-offs.

We put the new Fable AI model to the test on a real production project: how we compressed two weeks of development into just five hours, and why true success still depends on solid architecture and human review.

A non-technical PM's ground-level view of using Claude day to day: writing itself into status reports, explaining unfamiliar tools, sharpening QA, and drawing the line between understanding a system and building one.

We shipped llms.txt on marsbased.com and checked the data. Honest verdict: low cost, useful for docs, not a SEO play, and almost no bots read it yet.

A few months ago we quietly added a new section to the MarsBased Handbook laying out exactly how we use Claude Code day to day, from research to planning to implementation. We never wrote about it here, so we're fixing that now.

From Slack noise to curated briefings: How Claude Cowork transformed how I oversee projects and team alignment.

How we connected Claude to our remote stack to eliminate bias and bureaucracy from performance reviews, while keeping human judgment firmly at the center.