ScrollAI

What's New in 1.5 Benchmarks for your role Leaderboards used to answer "which model is best". Now they answer the more useful question: how good is AI at the work you actually do? Open Leaderboards and you'll land on a board built for your profession. Doctors and nurses see clinical benchmarks. Developers see real issue-resolution scores. Researchers see graduate-level science reasoning. Every score shows the date it was measured, so you're never reading an old number as if it were today's. Where a benchmark's headline score hides a real problem, we lead with the problem — legal AI scores in the high 80s on reasoning tests while still fabricating citations, and the board tells you that. And where nothing credible exists for a role, we say so plainly instead of showing you a general score dressed up as a measure of your job. Sharper story images Covers across Feed and Scroll are noticeably better. We now check every image before it reaches you and reject the things that made cards look broken: tiny product icons stretched to full width, publisher logos, and banner strips. Repeated stock photos are gone too — no more seeing the same picture on three cards in a row. Also in this release • AI-generated illustrations are clearly labelled wherever they appear • Faster image loading and lower data use