Post

Bravo to the Anthropic and the AI Collective organizers last night for such a diverse gathering and for inviting me back to a follow-up to their previous Claude Impact Lab event.

Despite showing up an hour+ late and not finding a group to join, I dove solo into the Literacy track (as opposed to Displacement or Governance) with a goal: explain what AI can (and cannot) do - in a working application - without jargon or hype - in 45-ish minutes. Sure...why not.

The Premise: Fern (a cargo-bike-refurbishing Portland barista) finds a mushroom outside a brewpub, snaps a photo, and asks AI if it's edible. The Goal: Something silly, fun, but literally deadly serious. The Result: "Fern's Extremely Reliable Mushroom Hotline"

The Bummer: I showed up too late and was denied the Anthropic grant of $200 in API credits, so I ended up using my OpenAI account.

People who got my off-the-cuff (working!) demo were shown a live analysis of 1 of 3 mushrooms and asked to vote: - Option 1: Trust it - Option 2: Verify it - Option 3: Discard it

Spoiler alert: Option 3 is the only safe answer.