Release notes are often short. Diffs are often huge. This is the missing middle: what actually changed in v0.106.0, in plain terms, and who should care. This post is an experiment: we had our planning agent Themistocles read the tag diff and draft a first-pass summary. We edited it for clarity — but we didn’t do a full manual code review ourselves. If you need ground truth, jump to the diff in Sources. If you spot anything off, tell us —...

The first time I vibe coded, it felt like cheating. In the best way. A proof of concept up in minutes? What? I remember thinking: is this the future? And honestly, that part still applies. For prototypes, demos, and experiments, vibe coding is pure momentum. It compresses hours - sometimes days - into a single burst of “let’s see if this idea even works.” But then I did what everyone does after the magic trick works once: I tried to use the...

Agents are moving from “answer questions” to “do work”: filing tickets, changing settings, running scripts, editing content, deploying changes. That shift is less about model quality and more about agency—the ability to take actions in real systems. The catch is simple: the moment an agent can act, you inherit the failure modes of automation and the ambiguity of language. “Human-in-the-loop” isn’t a conservative stance. It’s how you ship...

AI can absolutely help you write UI tests faster. The trap is thinking the “writing” part is the hard part. In practice, most UI test suites fail for the same reasons: brittle selectors, non-deterministic data, leaky environment state, and tests that don’t actually prove the user outcome. AI can speed up all of that too—but only if you use it inside a workflow that’s grounded in real fixtures, intentional mocks, and predictable data seeding....

Accessibility work usually doesn’t fail because teams don’t care. It fails because it’s time‑expensive, scattered across tickets, and full of “we’ll come back to it.” AI can help—a lot—but only if you treat it like what it is: a drafting and pattern‑finding tool, not a compliance oracle. The goal of this post is simple: move faster on accessibility without shipping confident mistakes. AI is a multiplier, not a judge AI is great at turnin...