February 17, 2026
Green checks, red flags
Show HN: Continue – Source-controlled AI checks, enforceable in CI
AI code cop goes full GitHub—devs ask “rebrand or remix?” and beg for a demo
TLDR: Continue adds an AI “code cop” that reads rule files and labels pull requests green or red, even suggesting fixes inside GitHub. Commenters are split between intrigue and skepticism, asking if it’s a rebrand, demanding CSV metrics and a demo, and comparing it to existing GitHub tools—clarity and proof are the big asks.
Move over human reviewers—Continue just rolled up like an AI traffic light for code changes, slapping green checks or red flags on every pull request (that’s a proposed code change). You write simple rules in text files, it scans the changes, and if it finds trouble, it even suggests a fix you can accept inside GitHub. It’s basically an AI hall monitor for your team’s code.
But the community immediately kicked off a mini-identity crisis. One top comment squinted at the logo and asked, “Is this the same Continue that did local AI coding agents?” Translation: is this a rebrand or a remix? Another voice called it a “more configurable” take on code review bots—and then pulled out the receipts request: do you export metrics to CSV? They even dropped the docs link and demanded a brief demo because, as one wag put it, “demo or it didn’t happen.”
Comparisons flew fast, with skeptics asking how this differs from gh-aw. The vibe: excitement for an AI code cop you can script with your own rules, but nerves about yet another GitHub robot. Expect memes about “the PR babysitter,” jokes about green-check dopamine, and a brewing showdown over what’s new vs. what’s marketing. The crowd wants clarity, metrics, and a live show—stat.
Key Points
- •Continue runs AI-driven checks on every GitHub pull request and reports results as status checks.
- •Checks are defined as markdown files in .continue/checks/ with a name, description, and prompt.
- •A provided example “Security Review” check flags secrets, missing validation, unsafe SQL concatenation, and sensitive logging.
- •The system evaluates only the PR diff and provides pass/fail with suggested fixes when issues are found.
- •A quickstart guides users through creating checks, connecting GitHub, and seeing them run on a PR.