1/n. What's great about Neurosynth? Fast, automated meta-analyses of the imaging literature, with access to >500k peaks auto-culled from >14k papers. What's the downside?

2/n. Noisy study selection (off-target papers) and lumping of contrasts (e.g. activation/de-activation; Exp vs. Control vs. Implicit Baseline)...

3/n. Cool, cool, so why not run a gold-standard manually curated meta-analysis?

4/n. They're not very scalable. Extremely time-consuming. What if you could have the best of both worlds? An analysis that was seeded by computer-identified papers, but enabled manual curation/tuning and automagically produced a reproducible pipeline and PRISMA plots?

5/n. Thanks to alejandro (@neurozorro.bsky.social) tal yarkoni and an amazing team of collaborators and trainees, it does! Welcome to the world of Neurosynth Compose! compose.neurosynth.org