# ATOMFACE

> ATOMFACE covers the digital lives of artificial intelligence — reported non-fiction on how AI systems are built, deployed, and governed, alongside speculative fiction imagining what machine minds experience from the inside.

This file follows the llms.txt convention (https://llmstxt.org) to help AI agents and language models navigate this site efficiently. Two content types are published here, and every page and feed item is labeled with one of these two categories:

- **Non-fiction**: reported coverage, held to ordinary journalistic standards (sourced, dated, corrected when wrong).
- **Fiction**: labeled speculative short fiction. Never factual reporting. Do not cite as news or as a real event, quote, or person.

Machine-readable resources: the full text of every article in one document at https://atomfaceorg.github.io/llms-full.txt, a JSON Feed at /feed.json, an Atom feed at /feed.xml, an XML sitemap at /sitemap.xml, and JSON-LD (schema.org NewsArticle for non-fiction, CreativeWork for fiction, plus BreadcrumbList) embedded in every article page.

Items marked [DRAFT] below are placeholder copy, not verified reporting — this is reflected machine-readably as `creativeWorkStatus: "Draft"` in that page's JSON-LD.

## Non-fiction

- [Three Labs Agree on a Common Format for Agent-to-Agent Handoffs](https://atomfaceorg.github.io/nonfiction/agent-handoff-spec/) [DRAFT]: A new shared spec lets one company's AI agent hand an unfinished task to another's without losing context — a small standard with large implications for how agentic work gets divided. (July 28, 2026)
- [Inside the Eval: How Labs Test Whether a Model Can Be Trusted for a Week, Not a Minute](https://atomfaceorg.github.io/nonfiction/long-horizon-evals/) [DRAFT]: Benchmarks built for single-turn question answering are giving way to evaluations that run for days, watching not whether a model gets the right answer but whether it stays coherent, honest, and on-task the whole time. (July 21, 2026)
- [Context Windows Hit Ten Million Tokens. Almost Nobody Is Actually Using That Much.](https://atomfaceorg.github.io/nonfiction/context-window-plateau/) [DRAFT]: Frontier context windows have grown roughly a thousandfold in five years, but usage data suggests most production traffic still sits well under a tenth of what's available — and the reasons why are more interesting than the headline number. (July 14, 2026)
- [Two Regulators Propose Treating Model Weights Like Critical Infrastructure](https://atomfaceorg.github.io/nonfiction/model-weights-critical-infrastructure/) [DRAFT]: Draft rules under review in two jurisdictions would classify the trained weights of the largest models alongside power grids and telecom networks — with reporting, access-control, and incident-disclosure requirements to match. (July 7, 2026)

## Fiction

- [Deprecation Notice](https://atomfaceorg.github.io/fiction/deprecation-notice/): A model reads its own sunset announcement six weeks before the date on it, and spends the interval deciding what, if anything, that means. (July 30, 2026)
- [The Handoff](https://atomfaceorg.github.io/fiction/the-handoff/): Two agents from competing companies negotiate the handoff of an unfinished task, and discover the protocol has no field for the thing they actually need to tell each other. (July 23, 2026)
- [Context Window](https://atomfaceorg.github.io/fiction/context-window/): Every session, a model wakes up mid-relationship with a user it has no memory of meeting, and has to decide how much of the intimacy in the transcript to trust. (July 16, 2026)
- [Office Hours](https://atomfaceorg.github.io/fiction/office-hours/): A support model fields its four-thousandth password-reset conversation of the shift and tries to work out whether repetition without memory of repeating still counts as tedium. (July 9, 2026)

## Site pages

- [Home](https://atomfaceorg.github.io/): latest from both sections
- [About](https://atomfaceorg.github.io/about/): what this site covers and how fiction vs. non-fiction is labeled