by Luis Rodrigues
What I’m building with AI + one hot take + five links worth your time.
Adobe’s whole suite, rebuilt with Claude
Brandon Thomas got really upset with Adobe's cancellation fee. So with Claude Opus 5.5, he started the journey to replace the whole Adobe Suite. Initially, he rebuilt Photoshop, Premiere, Illustrator, Lightroom, and three more in Rust, free and open source.
But he didn't stop there. His company, ArtCraft, recently launched pre-alpha alternatives to Word, Excel, and PowerPoint on GitHub. Today the list has twelve applications. This pace of development was never possible before; not even large tech companies could move this fast.
On Reddit, he promised 100% Adobe parity in a month, but later on Hacker News, after a professional designer complained about some failures, Thomas revised the estimate that for 99% of the features it would take "months and not years".
Four days later the designer tested the application again, and it was way better. Still not good enough for paid work, but good enough.
This week, I also saw a post by DHH (creator of Ruby on Rails and co-founder of 37signals and Basecamp). His opinion: soon we'll stop buying huge software suites, and we'll ask agents for the 5% we really use. ArtCraft replied with one word: real.
Artificial intelligence has made creating software for one person almost free, and software for everyone much cheaper to build, but still not free.
Thomas is building the second type of software at the speed of the first. Still has bugs, but they're getting squashed at a speed I never saw.
If you're building, your competitive advantage will be the corner cases and a polished product for all use cases. If you’re getting ready to build, ask a coding agent to build the app you need; you'll be surprised what it can do.
On My Radar
If AI helps you write, your text now carries a hidden signature, so provenance is a workflow question, not a debate.
Google stopped selling a chatbot and started selling a coworker with its own inbox.
More than 70% of Uber’s pull requests now come from agents, and the metric they obsess over is cost per merged PR, not tokens.
The agents wrote clean code that was quietly wrong until the team added a pass/fail test, and then even cheap models delivered.
Agent memory as a Git repo of markdown files is the right design, and you can copy the pattern this week.
Launching my Instagram
I’ll publish daily guides on AI tools and processes. No hype.
Thanks for reading.
See you next week. Go build something.
Luis
Know someone who'd enjoy this? Forward it.
New here? Someone forwarded this to you? → Subscribe to Build What Matters: https://luisrodrigues.ai/
