Deskpad Labs

We think good research matters as much as a good product. Labs studies how AI actually changes the way people learn and write, especially in K-12 education.

We publish what we find, even when the findings are inconvenient.

Sage, studying a Rubik's cube

What we study

01

How AI changes learning

What happens to a student's thinking when AI joins the writing process. We study the real classroom, not the demo.

02

Benchmarks for education

Most AI benchmarks were not built with students in mind. We want ones that are, so claims about AI in education can be measured instead of asserted.

See ATLAS, the first one
03

Transparency in practice

How AI use can be made visible and accountable instead of hidden, held to a research standard rather than a marketing one.

The log

Everything we are working on, in the open.

In progressBenchmarkSummer 2026

ATLAS

An independent benchmark for whether AI in K-12 education actually teaches, not just whether it answers. We are recruiting teachers to build the first cases.

Read more
In progressPaperSummer 2026

Structured transparency

What it actually means to make AI use visible and accountable, in a way that holds up rather than just performs.

How we work

We post findings, notes, and works in progress as they happen.

Read the blog