by Luis Rodrigues
What I’m building with AI + one hot take + five links worth your time.
AI is unlocking access to really technical tools
Ed, from Low Level on YouTube, has been hacking software professionally for 12+ years. This week he asked Claude Code to recreate a Linux calculator and let it run. He had the source, but gave the agent only the compiled app to see how good REA (Reverse Engineer Anything) was.
REA is one of the fastest-growing repos on GitHub. As I write this, it has over 333 thousand stars.
In 28 minutes, Ed had a functional copy of the original application with over 2,000 functions rebuilt from zero. Building this by hand would have taken several weeks.
And here's the point that most people are missing: the tools that REA uses aren't new. Ghidra has been public since 2019; IDA was created in the 90s. For decades, these tools were out of reach for most people because they required deep technical knowledge.
REA just enabled normal people to use agents to interact with these tools. The agent asks, the tools show what's inside the machine code, and the agent continues asking about the original app in a loop. That loop is reverse engineering, and anyone can run it now.
AI isn’t only creating new tools. It’s opening up the ones we already had.
Ghidra is just the first tool going viral; I'm sure in the next weeks and months you'll see other highly technical tools becoming available to non-technical users.
Access isn't the same as knowledge. Ed's code was generated in Python, and it had some interface bugs, but still it was a working calculator.
Why this matters if you're building, or getting ready to: the opportunity is connecting agents to specialized tools that are hard for non-technical people to learn. If you're getting ready to build, then stop asking what tool you should learn, and ask how to make it available to an agent.
Meet Zenzap: the easiest way to get real AI agents for real work
Most AI tools are built for one person talking to one chatbot. Real work happens in teams, across a dozen tools, with context nobody writes down.
Zenzap might be the best AI agent platform for real businesses. Set it up in a couple of clicks, connect your tools, and run agents right inside your team’s group chat. Automate reports, tasks, follow-ups, and approvals. It proposes, someone approves, and the whole team sees it happen.
Secure enough for any CISO to sign off on: admins control access, every action is logged, and your data never trains AI models.
On My Radar
The company that knows these tools best cut junior technical hires from 27% to 9%. To get in now, a junior needs good judgment more than they need to know the syntax.
Open-weight models now run most production tokens for under a quarter of the spend. The teams doing this well send the cheap volume to open models and pay frontier prices only where quality matters.
An agent can now explain how almost any app feature works, down to the binary. If a clever feature is the only thing protecting your product, it won't protect you for long.
Ramp spent $1.5M on tokens in a single week before learning that shared company context matters more than giving everyone the best model.
If five engineers on your team couldn't agree, fifty agents won't fix that. They'll make the same mistakes faster.
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/
