Verify every AI model ID your app calls still exists - shippingszn launch readiness

AI providers retire old models on a schedule. When they do, any call to that model ID stops working and returns an error — so an app that hardcoded a model name silently breaks the moment the shutdown date passes, usually with no warning to you.

This is a brand-new way for AI-built apps to die, and it's on a clock. Across 2026, OpenAI, Anthropic, and Google all retired waves of model IDs (OpenAI has hard shutdown dates on July 23 and October 23; Anthropic retires Claude Opus 4.1 on August 5; Google's Gemini 2.0 is already gone and 2.5 is dated for October 16). If your AI feature points at a dead model, it doesn't degrade gracefully — it just starts throwing errors for every user until you notice and fix it. AI builders love to hardcode a specific dated model; that's the trap.

The shippingszn CLI includes automated checks for this launch-readiness control.

The Launch Fix Kit keeps scan-specific findings, file and line evidence, AI-builder punch-list tasks, and verification steps tied to the paid report.

Search your code for every model ID: gpt-, o1/o3/o4-, claude-, gemini-, text-embedding-, and anything passed as a `model` argument to a provider SDK.
Check each against the provider's deprecation page (platform.openai.com/docs/deprecations, docs.claude.com model deprecations, ai.google.dev deprecations). Note which are retired, which have a shutdown date, and which are current.
Stop pinning dated snapshots (anything ending in a date like -2024-05-13 or -0613). Point at the current stable model instead.
Move the model name into config or an env var so you can change it in one place without a code hunt.
Add a pre-launch smoke test that calls each provider with its real configured model and fails if the model is gone.

Run the scanner | Unlock Launch Fix Kit

Canonical URL: https://shippingszn.com/i/model-freshness/