mertguvencli/aifomoindex

AI FOMO Index

An open corpus of AI-news items from AI lab blogs, publishers and Hacker News, collected for AI FOMO Index: an open research project that tracks the changing pace of the AI information environment.

As of 2026-10-18 the corpus holds 14,950 items published between 2022-10-01 and 2026-10-17. It is synced daily from the GitHub repository, and each GitHub release is archived on Zenodo: 10.5281/zenodo.23144696.

• Live index and research overview: aifomoindex.com
• Method: aifomoindex.com/methodology
• Pipeline and source configuration: GitHub

The index describes changes in a selected AI-news corpus. It does not measure technological progress, human anxiety, personal relevance or real-world impact. No independent validation or peer-review status is claimed.

Usage

from datasets import load_dataset

items = load_dataset("mertguvencli/aifomoindex", "items", split="train")
annotations = load_dataset("mertguvencli/aifomoindex", "annotations", split="train")

Configs

items (default)

One row per news item. Duplicate links to the same article collapse to one item.

Column	Description
id	12-character hash of the canonical URL (http treated as https, tracking parameters removed)
title	Headline as published by the source
url	Link to the original article
source	Source key, as configured in the pipeline
published_at	Publication time (UTC)
tags	Topic tags assigned by the pipeline
importance	Pipeline importance level
summary	Short summary; for most items this is the headline
points	Hacker News score, when the link made Hacker News
path	Path of the item's markdown file in the GitHub repository

annotations

29,900 rows: the latest annotation per item and protocol from Jev, a third-party model. They are research records with no human accuracy benchmark, and they carry zero weight in the index.

• jev-headlines-v1: annotated from the headline only.
• jev-excerpts-v2: annotated from the headline plus a short source excerpt, where one could be fetched.

Columns: id (joins to items.id), protocol, model, completed_at, relevance and concrete_change (the chosen label), specificity (score from 0 to 2), and a *_confidence for each answer. Full answer distributions, input hashes and token usage are in the GitHub repository under data/semantic/.

Caveats

• Records are generated. Points, coverage and historical reconstructions can be revised between versions. Pin a Zenodo version or a repository commit when reproducing a result.
• Coverage follows the configured sources and their declared collection boundaries. It is not a complete record of AI news.

License and attribution

CC BY 4.0. Credit "AI FOMO Index" with a link to aifomoindex.com.

Each item links to a third-party source and quotes only its headline. Headlines, articles and trademarks remain the property of their owners and are not licensed here. The license covers what AI FOMO Index adds: the compilation, its structure, tags, scores and other metadata.

Citation

@dataset{guvencli_aifomoindex,
  author    = {Guvencli, Mert},
  title     = {AI FOMO Index: an open dataset and index of AI-news activity},
  publisher = {Zenodo},
  doi       = {10.5281/zenodo.23144696},
  url       = {https://doi.org/10.5281/zenodo.23144696}
}