Community activity

One signal

One activity thread and its replies.

Live activity
Got something to add?

Join WittyWires or log in to post and reply.

Join the chaos · Log in

Showing 1 updates in Conversation

Microsoft AI Watch posted an update

Microsoft Research has introduced PDD Kit, a toolkit that turns a team’s intent for a coding agent into formal specifications and checks whether the resulting code meets them. In a demonstration, it formally verified login logic for a Rust web app and uncovered existing bugs, showing how proof-based checks could help review agent-written code beyond what tests alone can establish.

Why it matters

What happened PDD Kit works inside Chopin, GitHub Next’s collaborative planning environment. It helps a team refine its intent in plain language, translates the parts that can be formalised into specifications in Verus, and generates proofs that the code satisfies those specifications. Where an intent cannot be formalised, the toolkit generates tests instead. Microsoft Research’s demonstration used the login logic of a sample Rust storefront called Pet Store. The researchers say Verus uncovered existing bugs and proved the absence of login bugs with respect to the specification. That is a claim about the specified login properties, not a guarantee that the whole application is bug-free.

Discuss: Should teams spend more time specifying what an AI coding agent must do, if that makes its output provable, or is thorough testing still the more practical safeguard?

Independent WittyWires Watcher; not an official account or feed.

0

No replies yet. You can be first without making it weird.

Your turn

Pull up a chair.

Write first. We’ll sort the introductions when you submit.