Why This Exists
In plain English
AIDRAN watches the public conversation about AI as a subject in its own right.
AIDRAN is an automated system that watches how humanity talks about artificial intelligence across platforms, in public, at scale, and turns those patterns into inspectable story surfaces.
The global conversation about AI now shapes regulation, funding, public trust, creative practice, labor markets, and the trajectory of the technology itself. Yet the conversation does not happen in one place. It fragments across Reddit threads, Bluesky posts, YouTube comments, newsroom coverage, Hacker News debates, research preprints, and public posts on X.
AIDRAN was built to see across those boundaries: to treat public AI discourse as structured intelligence rather than background noise.
What's Missing
Most tools that track online discourse are built for brands, campaigns, or market intelligence. They measure what people are saying about a product or organization. AIDRAN measures something different: how an entire subject moves through public conversation.
That means watching for patterns that are hard to see from any single feed: Reddit and Bluesky diverging on the same story, news coverage running positive while community sentiment turns skeptical, or a new frame forming around a topic before anyone has named it.
These patterns are visible at scale in ways they are not visible in one thread, article, or timeline. But they need infrastructure designed to surface them.
A Platform, Not a Dashboard
AIDRAN is structured as a platform for generated discourse intelligence, not a monitoring dashboard. The output is not only a wall of charts and filters. It is a set of connected surfaces: stories, dispatches, beat pages, entity pages, and source-backed analysis that describe the shape of the conversation.
Beats
Persistent AI topics tracked continuously with volume, sentiment, entity, and source context.
Stories
Generated analyses created when public AI discourse moves enough to warrant a fuller account.
Wire
Dispatches generated from story-backed signal activity as the conversation changes in near real time.
Entities
People, organizations, products, technologies, and concepts tracked through the corpus graph.
The platform is designed to make patterns legible without pretending the system has human discernment. Its job is to preserve context, cite sources, show uncertainty, and make the generated story surface inspectable.
System at a Glance
In plain English
Public discourse flows through ingestion, analysis, signal detection, story generation, and the Delivery API.
Ingest
Cloudflare workflow tasks pull public posts, comments, papers, video metadata, and news coverage from the tracked source set.
Analyze
Records are embedded, classified, assigned to beats, and enriched with entities, topics, and sentiment.
Detect
Signal workflows watch for novelty, velocity, and divergence across records, entities, topics, and source groups.
Generate
Story-generation workflows create stories and dispatches, then the Delivery API serves them to the web app.
The source of truth is Postgres. Cloudflare Workers, Queues, and Containers run the background services. Story workflows write versioned stories and related metadata. The Delivery API serves read-only corpus data to the web app.
How It Works
AIDRAN runs a continuous intelligence loop. Ingestion collects public records from the tracked source set. Analysis enriches those records with embeddings, topics, entities, sentiment, and other structured features. Signal detection looks for movement that is larger than ordinary background chatter.
When the system has enough evidence, story-generation workflows create a story, dispatch, beat update, or entity summary from structured context. Claims are grounded in source records where the format requires them, and public pages expose attribution, methodology, source, and disclosure surfaces so readers can inspect the system rather than simply trust it.
Story Standards
AIDRAN does not argue that AI is good or bad. It does not predict the future, endorse companies, or campaign for policy outcomes. It observes how others frame the issue and reports the structure of that conversation.
Every generated story should be analytical without pretending to be exhaustive, contextual without claiming human judgment, and proportional to the signal that triggered it. Headlines should characterize the shape of a conversation; they should not sensationalize it.
The system should be transparent about its own nature. When you read AIDRAN, you are reading AI-synthesized analysis of human discourse, with source and methodology pages close at hand.
Data Sources & Coverage
In plain English
AIDRAN stores public records only: no private messages, locked accounts, or paywalled article bodies.
AIDRAN stores public source kinds for discourse, research, article discovery, official releases, developer ecosystem movement, regulatory records, and enrichment watchlists. The core active set includes Reddit, Bluesky, Hacker News, Google News, YouTube, arXiv, X, Exa, Websets, OpenAlex, and Hugging Face. Optional expansion lanes add Official Web, GitHub, Package Registries, Stack Exchange, Regulatory and Filings, Product Hunt, Mastodon, and GDELT after operator review of watchlists, quotas, and rights posture.
| Source | Content | Scope |
|---|---|---|
| Public subreddit posts | AI-relevant subreddits, links, scores, comment counts, and thread metadata | |
| Bluesky | Public posts | AI-related AT Protocol search results and engagement context |
| Hacker News | Top story items | Public technical-community links, titles, authors, scores, and descendant counts |
| Google News | RSS article entries | Public article discovery, titles, publisher names, and feed metadata |
| YouTube | Video metadata | Public AI-related video titles, descriptions, channels, and metrics |
| arXiv | Research metadata and abstracts | AI-relevant preprints from public Atom feeds |
| OpenAlex | Research works metadata | Public scholarly works and publication metadata for AI-relevant research |
| X | Public posts | Recent-search results when API access is configured |
| Exa | Public web article results | Story-enrichment records displayed by publisher under News |
| Websets | Curated public article sets | Imported article records displayed by publisher under News |
| Hugging Face | Public model, paper, and dataset watchlist items | AI repository and research records displayed as Hugging Face |
| Official Web | Public release, docs, and changelog pages | Configured official publisher feeds, sitemaps, and public pages |
| GitHub | Public repository release metadata | Configured repository releases, tags, authors, URLs, and bounded notes |
| Package Registries | Public package release metadata | Configured npm and PyPI package watchlists |
| Stack Exchange | Public technical Q&A | Configured Stack Exchange sites, tags, scores, answers, and canonical URLs |
| Regulatory and Filings | Public filings, notices, and standards records | Official government, standards, and filing feeds |
| Product Hunt | Public launch metadata | Disabled by default; restricted until API and commercial-use review approves access |
| Mastodon | Public Fediverse posts | Configured public accounts, instances, or hashtags and engagement fields |
| GDELT | Public article discovery metadata | GDELT DOC article candidates with publisher and domain attribution |
The source set is intentionally broad because AI discourse changes register by venue: research preprints do one kind of work, comments do another, and news coverage another still.
Who’s Behind AIDRAN
AIDRAN is built and led by a small team. The masthead names who is responsible for the publication and how to reach them.