đź’Ś Is your AI agent lying?
8 of 42 databases got it wrong
How many rows are in this table? It’s the easiest question you can ask a database. Put to 42 engines, eight came back with confident nonsense. This week we dig into which statistics you can actually trust, why your browser agent shouldn’t get to decide when it’s done, and how microVMs keep misbehaving AI agents boxed in. More from the community: The Linux golden image trick that spins up servers in minutes, and the dot commands that formatted documents long before Markdown.
On the podcast, a developer advocate who flipped burgers for 17 years before writing a line of code explains his one-tool AI habit, and why he’s not switching back. This is edition #86 of the We ❤️ Open Source newsletter. Made and curated with renewed energy by Jason & Todd.🙂
This week we'll look at...
Get to Know Torc
Met the Randstad Digital Powered by Torc team at All Things Open? Don’t let the connection end with the conference. Join their community to keep up with future events, conversations, resources, and opportunities.
Articles
How many rows are in this table? It sounds like the easiest question you could ask a database. We put it to 42 engines, and eight answered with confident nonsense, from hundredfold misses to permanent zeros. Find out which statistics you can trust before your dashboard fibs to you.
WHY WE ❤️ IT: Real servers, real numbers, and the guts to admit their own bugs. That's good, honest homework.
Long before Markdown, writers formatted documents by typing little commands that started with a dot. Jim Hall takes us back to RUNOFF in the 1960s, then shows how to rebuild it with a short Gawk script. It’s part history lesson, part coding project, and all kinds of fun.
WHY WE ❤️ IT: Jim brings 1960s text formatting back to life, with a whole lot of "so that’s where that came from."
Your agent clicked "Add to cart," said it was done, and the cart is still empty. Sound familiar? Skyvern’s team shares the second check that keeps their browser agent honest, from evidence-based prompts to prompt-injection guards. You can add one to your agent loop this week.
WHY WE ❤️ IT: Real bugs, real fixes, and a homework assignment from Suchintan that you’ll actually want to do.
Hardened images secure what you ship, but what happens once AI agents and AI-written code start running in production? Edera CMO Kaylin Trychon tells 808labs how giving every workload its own microVM keeps a compromise contained, and where the company is headed next: GPU isolation.
WHY WE ❤️ IT: Agents will misbehave sometimes. Edera makes sure one bad actor stays in its own box.
What's in your technology toolbox?
Tired of setting up the same server from scratch every time? Learn Linux TV shows how to build a Linux golden image on Linode, a reusable template with your packages, security settings, and configs baked right in, so new servers are ready to go in minutes.
WHY WE ❤️ IT: Jay bakes fail2ban into his golden image so every new server starts already guarding its own front door.
From theWe ❤️ Open Source Podcast
He flipped burgers for 17 years before he ever touched a line of code. Now he’s a developer advocate with a blunt Hacktoberfest hot take and a one-tool AI habit that saves him money and sanity. Here’s how he got there, and why he’s not switching back.
WHY WE ❤️ IT: Because "keep the most important thing the most important thing" is good advice whether you code or not.
Featured podcast shorts
Wrap Up
If you've made it this far (thank you!), check out everything we do, our YouTube channel with 1000's of open source talks, the many meetups we host around the southeast and NYC, the All Things AI spring event we co-host, and of course All Things Open, the largest open source tech conference on the US east coast.
We hope you learn something new this weekend!
How did you like the newsletter?
Update your email preferences or unsubscribe here
© 2026 We ❤️ Open Source
P.O. Box 5333
Columbia, South Carolina 29250, United States