EmpowerEd 2026

A podcast-style conversation with Jeff Plourd and Jon Laven, the two former teachers who founded Snorkl. They met as first-year teachers in Las Vegas in 2012. For about a year their tool had no AI in it at all: students recorded their thinking and gave each other feedback. Then ChatGPT arrived in November 2022 while they were holding a pile of student recordings nobody had time to watch.

Benny asks them why a recording of a student thinking is harder to fake than a finished product, where a student's voice recording actually goes and which country holds it, and what a teacher in this room does on Monday morning. The founders then demo the tool and take questions from the room.

Teachers take one question from their own subject and rewrite it so it asks students to explain their reasoning. Then they build it in Snorkl from a blank screen and choose how students will answer: typing, drawing, talking, or talking while drawing. Talking while drawing is the hardest for a student to fake with AI, so that is the option the programme asks teachers to try first.

By the end of the session every teacher has one question built from scratch, in their own language, and the cohort keeps improving it over the next three weeks. In trios, teachers answer each other's questions, so by Week 3 every question has real responses on it.

Snorkl will draft its own marking criteria from any question. Accepting that first draft is the beginner mistake. If the guidance is vague, the tool only checks whether the answer was right. If it is specific, the tool can comment on how the student reasoned. This week teachers add example answers and marking points to their Week 2 question, then read what the tool said about the real responses their colleagues left.

Teachers ask first whether they can trust the tool's comments. In this programme a teacher reads and edits every comment before a student sees it, and this week each teacher edits at least one.

Instead of reading thirty answers one at a time, teachers read the pattern. The centrepiece is a live class summary of the cohort's own question, built from around thirty real answers left by the people in the room over the previous two weeks.

From what they actually see, each teacher writes a marking guide of three or four lines, and one sentence on what they do when a student uses AI on the task. The position on honesty has to work without detection software, because detection gives false positives and punishes the wrong students.

In Week 5 there is no teaching. Every teacher presents their finished task in three minutes, in small rooms of five, and each room picks one task to show the whole group.

The shared collection is then handed over to the cohort, sorted by subject, so a teacher in Jakarta can adapt a task built in Warsaw. Teachers can also pair with someone in another country and trial each other's task next term.