Freshness surface

The cabinet gets inspected weekly

This page is designed to grow: new demos, source refreshes, retired claims, and feature-availability checks belong here.

Update log

2026-09-19

The Stage opens: one wow-demo with the wires labeled

Added /stage/, the signature spotlight for the one-prompt artifact prototype on Claude Opus 5. The page carries the copyable prompt, checkable house and proof cues, verification steps, and Anthropic artifact citations. It does not publish a transcript or promise identical reruns. Funny fails route to Claude Gone Wild; experiments route to Curious Claude. /casting/ remains the model-lineup page.

2026-09-19

Model refresh: Fable 5.1 and Opus 5 named across the cabinet

Every demo now carries a casting note saying which current model to run it on, and a new /casting/ page lists the generally available lineup with API IDs: Claude Fable 5.1, Claude Opus 5, Claude Sonnet 5, and Claude Haiku 4.5. Retired the Claude Opus 4.8 source record, which described a model that is no longer the current Opus, and replaced it with the Opus 5 and Fable 5.1 overviews plus the model-overview and choosing-a-model pages.

2026-09-19

Two demos added where the model is the subject

Added the Long-Horizon Agent Run for Fable 5.1, which Anthropic positions for demanding reasoning and long-horizon agentic work, and the Effort Dial Reveal, which runs one prompt at two thinking-effort levels. Both are demo scripts rather than measurements; scored comparisons route to Claude Reports.

2026-07-06

Launch cabinet published

Published eight reproducible Claude demos covering artifacts, PDFs, vision, data visualization, Claude Code, computer use, web research, and long-context synthesis.

2026-07-06

Source baseline captured

Initial source set uses official Anthropic docs and Claude Help Center pages current during build.

Next weekly slot

Future demo refresh

Add a new demo, retire stale source claims, and refresh feature availability notes when Anthropic docs change.

Refresh checklist

  1. Open each official source linked from the affected demo.
  2. Check the model overview for new models, changed API IDs, and retirement dates, then update every casting note that names a retired or superseded model.
  3. Confirm feature availability, plan requirements, and safety warnings still match the page copy.
  4. Update the demo prompt only if the old prompt no longer reflects current product behavior.
  5. Add a changelog entry that distinguishes source refreshes, model refreshes, and new demos.

Why model names get their own step

A stale capability claim usually still reads as plausible. A stale model name does not: it tells a reader the page was written for a product they no longer have. The lineup changed twice between July and September 2026, which is whycasting is a maintained surface rather than a one-time note.

Primary Sources

These demos cite official Anthropic docs or help-center pages for capability claims. They do not claim identical output across accounts, plans, models, or dates.

Anthropic Docs

Claude model overview

The model overview is the canonical list of generally available Claude models, API IDs, context limits, and retirement dates.

Claude Platform Docs

Claude Fable 5.1 overview

Claude Fable 5.1 (`claude-fable-5-1`) has a 1M-token context window, 128K max output, always-on adaptive thinking, and a default effort level of high in Claude Code and medium on Claude.ai and Cowork.

Claude Platform Docs

Claude Opus 5 overview

Claude Opus 5 (`claude-opus-5`) launched July 24, 2026 with a 1M-token context window, 128K max output, adaptive thinking, and an optional fast mode.

Claude Help Center

Enable and use web search

Claude web search can be enabled in chat and grounds responses with current web content and source links.

Claude Help Center

Let Claude use your computer in Cowork

Computer use lets Claude interact with permitted apps and files, but Anthropic warns users to avoid sensitive apps and monitor behavior.