Technology Watch (AI Models)

Technology Watch tracks upcoming shutdown and retirement dates for AI models across nine providers, so a model your product depends on doesn't go dark without warning. It lives as the AI Models tab on the Cloud Status page — not a separate menu item.

Why it exists

Uptime monitoring catches a service that is down right now. It cannot catch a dependency that is scheduled to disappear in six weeks. AI providers retire models on their own timeline — sometimes with only a few weeks' notice — and a hardcoded model name in your codebase becomes a silent production outage the day the provider flips it off. Technology Watch is operational risk monitoring: it surfaces deprecations early enough that swapping the model is a planned change, not an incident.

How it works

Data is self-collected. Syschecks scrapes each provider's official deprecation/retirement page on its own schedule — there is no dependency on a third-party feed. Nine providers are covered:

OpenAI
Anthropic
Google
Vertex AI
AWS Bedrock
Cohere
Groq
xAI
Azure

Each provider is re-scraped every few hours. New or changed shutdown dates are recorded as deprecation events; alert thresholds are re-evaluated on their own hourly pass.

Upcoming Model Shutdowns

The AI Models tab shows a table of every model with a confirmed future shutdown date, sorted by urgency (soonest first):

Column	Meaning
Provider	Which of the nine sources reported the model
Model	Model identifier, with its suggested replacement(s) if the provider listed one
Deprecated	Date the deprecation was announced
Shutdown date	Date the model stops working
Days left	Calendar days remaining until shutdown
Status	Urgency bucket, see below

Status buckets

Status	Days left
Critical	14 days or fewer
Warning	15–90 days
Deprecated	More than 90 days

"Deprecated" is the least urgent bucket

Despite the name, the Deprecated status is used for shutdown dates more than 90 days out — it just means the provider has announced an end-of-life, not that it is imminent. Watch the Days left column, not the label, when triaging.

Watching a model

You get alerted about a model in two ways:

Star a row in the Upcoming Model Shutdowns table — the star toggles watching that exact model.
Add one manually with + Watch a model: pick a provider, and optionally a specific model. Leave the model empty to watch every model from that provider, including ones that don't have a scheduled shutdown yet — useful for models you depend on today, so you're alerted the moment a shutdown is first announced.

Use the provider pills, the search box, and the Watched filter to narrow the table down to what matters to you.

Alerts

Watched models are checked against shutdown-date thresholds: 90, 30, and 7 days before the shutdown date. Crossing a threshold sends a notification through your configured notification channels.

Never pages on-call

Technology Watch alerts are informational only. They go out through your notification channels but never open an incident, never trigger escalation, and never page on-call — a model deprecation is a planned migration, not a 3 a.m. emergency.

API

Watches and the deprecation catalogue are also available over the API:

GET /api/organizations/{org_id}/lifecycle-watches

[
  {
    "id": 12,
    "platform": "openai",
    "vendor": "openai",
    "resource_id": "gpt-4-0314",
    "created_at": "2026-05-01T09:12:00Z"
  }
]

resource_id is optional on create — omit it to watch every model from that platform. GET /api/lifecycle/events (the Upcoming Model Shutdowns feed) and GET /api/lifecycle/catalogue (the model picker's source list) are read-only and not scoped to an organization.