Browsers as AI agent sandboxes + leadership transparency wins

Today's insights span from technical innovations in AI agent containment to management practices that treat engineers as adults, not children to be managed.

• Browser sandboxing for AI agents: Paul Kinlan's Co-do demo proves browsers can safely run coding agents using File System Access API, CSP headers with iframe sandboxing, and WebAssembly in Web Workers—without multi-GB local containers. The <input type="file" webkitdirectory> tag works across all major browsers for directory access.

• Unified exception testing in pytest: Instead of separate test methods for success/failure cases, use nullcontext as produces and pytest.raises in a single parameterized test—cleaner and more maintainable than the traditional approach.

• Engineers in leadership meetings transforms decision-making: Will Larson reports 6 years of success having senior engineers report directly to him and join all leadership discussions. Key benefit: prevents managers from "managing stuff around reality" instead of reality itself, plus creates dual information propagation paths.

• Curiosity before problem-solving reduces cost of being wrong: When someone misses expectations, asking "what happened?" before applying accountability damages fewer relationships and often reveals missing context that changes your response entirely.

• Lighthouse hiring multiplies recruiting effectiveness: Hiring one well-connected person (like Julia Evans at Stripe) improves both quality and velocity of subsequent hires through network effects, though it requires careful navigation of power dynamics.

#browser-sandboxing#ai-agents#pytest-testing#engineering-leadership#hiring-strategy

8 articles published

Articles

Ned Batchelder

Testing: exceptions and caches

Two testing-related things I found recently. Unified exception testing Kacper Borucki blogged about parameterizing exception testing, and linked to pytest docs and a StackOverflow answer with similar

Will Larson

Writing Visualizations with Remotion

Remotion is having a bit of a moment at the moment, and I decided to play around with the Claude Code integration. Here are a couple videos I was able to make in <10 minutes summarizring data on my bl

Will Larson

Stripe's Lighthouse Hiring pattern.

I did a lot of hiring at Uber, some days I would be doing back-to-back 30 minute phone screens for several hours in a row. That said, while Uber taught me how to hire at scale, it was Stripe that taug

Will Larson

Pressure Without a Plan.

When we launched Digg v4, the old site turned off, but the new site didn’t turn on. There was a lot of pressure to get things working, but no one knew what to do about it. It took almost a month to ge