Hello, world — the pipeline works
This is the first post on AI Evals AI. Its only job is to prove the whole pipeline works: writing in MDX, committing to GitHub, and having Cloudflare build and publish the site automatically. Real content comes next.
A code block
Rendering code correctly matters for a blog about testing. Here’s a tiny example of the kind of thing I’ll write about — a trivial assertion:
function evaluate(output, expected) {
return output.trim() === expected.trim();
}
console.assert(evaluate(' pass ', 'pass'), 'should ignore surrounding space');
A small table
Tables are handy for eval results. Here’s a throwaway one:
| Check | Expected | Status |
|---|---|---|
| Domain resolves | yes | ✅ |
| HTTPS active | yes | ✅ |
| RSS feed valid | yes | ✅ |
What’s next
With the pipeline confirmed, the next posts will get into the actual topics: evaluating AI systems, testing agents, and what quality engineering looks like when outputs aren’t deterministic.