Research
The science behind representation superintelligence.
Mantis is made in Professor Manolis Kellis’s group at MIT CSAIL, where computation is used to reveal hidden structure in complex systems. This is the published work it carries forward, and the engineering it is built on.
Lineage
From the Kellis Lab at MIT CSAIL
- Nature 590, 300–307 · 2021Regulatory genomic circuitry of human disease loci by integrative epigenomicsBoix, James, Park, Meuleman, Kellis (opens in a new tab)
- Nature 570, 332–337 · 2019Single-cell transcriptomic analysis of Alzheimer’s diseaseMathys, Davila-Velderrain, et al. (opens in a new tab)
- Nature 518, 317–330 · 2015Integrative analysis of 111 reference human epigenomesRoadmap Epigenomics Consortium et al. (opens in a new tab)
Technical notes
The engineering behind the workspace
Each note is a measured change to how Mantis streams, decodes, draws and remembers maps, with a frozen baseline and identical output asserted before any timing.
- 10.4×First points on screenfaster to first points at 500,000 points
- 21.4×Decoding the mapless main-thread decode time at 250,000 points
- 5.39×Cluster membershipfaster at 100,000 points
- 2.0×Assembling the point mapfaster at 250,000 points, with half the insertions
- 256×Memory across Spacesless retained map state after 256 Spaces
- 33.5×Cache policyfaster on 5,000 paired set and get operations
- −99.1%Hidden panels at resttimer ticks per minute while hidden
- 40 → 0Steady labelslabel placement changes over a 40-frame pan
Latest
Releases
- Ask about Mantis answers live, from any page of the site.
- A new front page, opening on Mantis at work in its own workspace.
- Instant navigation: the sidebar stays put as you move between pages.
- Seven public dataset Spaces open to everyone.
- Continuum-1, Mantis’s own model, serves the Composer, with usage tiers.
Work with us
Bring representation superintelligence to your team.
Start in the workspace today, or talk to us about your data, your deployment and your questions.