TEXXR means "tech extended release" — the idea that news shouldn't be disposable. Every day, the tech press publishes hundreds of articles. Each one covers an event. But events are chapters, not stories. The story is the arc: how an entity changed over months, what a pattern of coverage reveals about power shifts, where two companies' trajectories intersected without either realizing it. TEXXR is the tool that finds those arcs.
What's Inside
The platform is built on a database of over 130,000 articles from Techmeme, spanning 2014 to today. Every article has been processed through three layers:
Vector embeddings. Each article is converted into a 1,536-dimensional vector using OpenAI's embedding models. This means any two articles can be compared for semantic similarity — not by keyword matching, but by meaning. An article about "Arm designing its own chip" is semantically close to "TSMC losing a customer," even though they share no keywords.
Entity extraction. NLP models identify every company, person, product, and technology mentioned in each article. Apple, Tim Cook, iPhone, Core ML — each gets tagged and indexed. This turns unstructured headlines into a searchable knowledge base of who did what, when.
Knowledge graph. A hyperedge extraction pipeline reads every article and identifies structured relationships: "OpenAI → partnership → Microsoft," "Arm → launch → AGI CPU," "Meta → legal → New Mexico." Over 85,000 of these edges form a knowledge graph that tracks how entities relate to each other — and how those relationships change over time.
What You Can Do
Semantic Search
Search by meaning, not keywords. Ask "companies building their own AI chips" and get results about Arm's AGI CPU, Google's TPU, Amazon's Trainium, and Meta's MTIA — even if none of those articles use the phrase "building their own AI chips."
Entity Profiles
Every entity in the database has a profile page showing its full coverage arc. OpenAI's profile shows 4,000+ articles spanning its transformation from nonprofit to the most funded AI company in history. Arm's profile reveals the shift from chip designer to chip competitor. Each profile includes coverage timelines, related entities, knowledge graph connections, and momentum metrics.
Story Arcs
The Arc tool visualizes how a topic evolves over quarters. Enter "AI regulation" and see the coverage shift from abstract policy debate (2023) to state-level lawsuits (2024) to jury verdicts (2026). Enter "self-driving cars" and watch the narrative move from Waymo's early tests to full commercial deployment. The arc is the story that no single article can tell.
Semantic Drift
The Drift map tracks how an entity's meaning changes over time. When we say "OpenAI" today, we mean something completely different from what we meant in 2020. Drift quantifies that shift — projecting the entity's embedding trajectory onto axes like "safety vs. speed" or "research vs. commerce." The result is a visual map of transformation.
Knowledge Graph Exploration
The Nexus interface lets you explore the knowledge graph directly. Search for an entity and see every relationship it's involved in — partnerships, acquisitions, launches, controversies, legal actions. Filter by time period. Compare two entities' relationship histories. The graph reveals connections that headlines miss.
For example, the Arm/Nvidia arc traced through the knowledge graph shows: joint AI research (2024) → Nvidia cutting its Arm stake (Feb 2025) → NVLink partnership (Nov 2025) → Arm launching a competing chip (Mar 2026). Each step was a separate headline. The graph shows the arc.
Signal Detection
The Pulse API runs statistical analysis across the full corpus. It detects z-score spikes (entity getting far more coverage than its 30-day average), silences (major entity absent from today's news), co-occurrence anomalies (two entities appearing together at unusual rates), and juxtapositions (same entity in contradictory contexts). These are the raw signals that our editorial posts investigate.
The Extended Release
Every day, TEXXR publishes editorial posts that use these tools to find patterns in the news. Not summaries of what happened — patterns that span months or years, revealed by today's articles.
Some examples of what the tools surface:
- The Architect's Chip — The knowledge graph's regime detection showed Arm experienced its largest identity break in recorded history in Q1 2026, shifting from "partnership" to "competitor." The 10-year arc from chip designer to chip maker was visible in the data before it was obvious in the news.
- The Convergence — Edge-arc analysis revealed that OpenAI and Anthropic arrived at identical coverage profiles (23% competitor, 21% financial) from opposite starting points. Two companies founded on different principles, shaped by the same market into the same thing.
- The $6 Million Precedent — Semantic search traced the social media liability arc from the 2021 Facebook Files leak through 70+ lawsuits to two jury verdicts in one week — a 5-year transformation from congressional hearings to courtroom liability.
- The Suboptimal Move — Similar-article search connected a 2026 Bloomberg feature about chess grandmasters to a 2023 Go breakthrough and the 1997 Deep Blue match — a 29-year arc of human-AI competition.
The tools find the signals. The posts interpret them. Together, they turn disposable news into durable understanding.
Go Deeper
The methodology pages explain each tool in detail:
- Vector Patterns and Triangulation — How semantic search finds connections that keywords miss
- What Is Semantic Drift? — Measuring how an entity's meaning changes over time
- What Is a Knowledge Graph? — Structured relationships from unstructured headlines
- What Is Entity Momentum? — Velocity, acceleration, and jerk in coverage
For a complete index of editorial posts organized by theme, see the Topic Map.
How to Start
The platform is free to explore:
- Browse: The chronicle shows every article, searchable by date, entity, or text.
- Explore: /explore for semantic search across the full corpus.
- Read: /posts for editorial analysis using the tools above.
- Build: /pulse for API access to search, entities, edges, and momentum.
- Research: /docs/cli for the command-line interface.