Nir Lotan

Hi.My nameis Nir.

I'm a data scientist and AI researcher at Intel. I research what social networks know about the world, and I make software, most of it open source. I also run :)

Research.

I study what the follow graph knows: how the accounts people follow together encode politics, taste and other social knowledge, and what that can do for language models.

All papers on Google Scholar

  1. 2026Who Goes Toxic? Understanding Political Toxicity on Social Media Across Individuals and ContextsZoizner, Lotan, Pendzel, Minkov. Social Science Computer Review
  2. 2026Social Knowledge for Cross-Domain User Preference ModelingLotan, Solomon, Guy, Minkov. arXiv
  3. 2024A Closer Look at Multidimensional Online Political IncivilityPendzel, Lotan, Zoizner, Minkov. EMNLP 2024
  4. 2023Social World Knowledge: Modeling and ApplicationsLotan, Minkov. PLOS ONE

Some projects.

Things I built and look after.

Likewise

Web app

Pick the musicians, shows and teams you love and chat with an AI persona who shares your taste, and another who really doesn't.

  • A Kindred who gets you, a Wildcard who won't
  • Free to try, no sign-up needed
  • Personas are always labelled as AI

Open likewisechat.com

GlideDeck

Windows and macOS, open source (MIT)

Presentations that move. Lay a whole talk out on one zoomable canvas and fly the camera from frame to frame.

  • Native, written in Rust, drawn on the GPU
  • Imports and exports PowerPoint
  • No accounts, no cloud, no telemetry

Visit the GlideDeck siteView on GitHubDownload

Copilot Pacer

macOS menu bar and Windows tray

Will your GitHub Copilot credits last the month? A small icon that goes green, amber or red so you know at a glance.

  • Paces against your real working days and holidays
  • Daily budget and an end-of-cycle forecast
  • Talks only to GitHub. No analytics

Visit the Copilot Pacer siteView on GitHubDownload

SocialVec

Python library, open source

Social embeddings: vectors for popular accounts, learned from who follows them together. Like word embeddings, for the social world.

  • Find accounts similar to one account, or to a group
  • Roughly 200,000 accounts, learned from 1.3 million users
  • Built with Einat Minkov at the University of Haifa

View on GitHubRead the paper

$ pip install socialvec

AURA

Coming to the App Store

The next one. Details here as soon as it ships.