This document describes who reviews the site’s product and model lists, what they check, and how often. It is a living process, not a one-time audit.

Why this matters

AI model versions change frequently. A table that lists “Claude Sonnet 3.5” when Anthropic has released several successors looks wrong to anyone who knows the space — and erodes reader trust in everything else on the site.

The pages most at risk are those that name specific models or list specific products. These are not opinion pieces; they are factual claims. When the facts change, the pages must change. An unreviewed list signals neglect.

What to review

Three pages on this site contain factual claims about specific products and model versions that will become stale:

PageWhat it containsRisk level
_docs/concepts/what-is-an-ai-coding-assistant.mdProduct list: Copilot, Cursor, Windsurf, Amazon Q DeveloperMedium — products rarely disappear, but relative positioning shifts
_docs/copilot/models.mdModel table: GPT-5.4, Claude Sonnet/Opus, Gemini 2.5 Pro, Grok Code Fast 1High — model versions change frequently, new models add regularly
_docs/landscape/index.mdModel provider table + AI coding assistant table + cloud coding agent tableHigh — all three categories change; cloud agents are newest and fastest-moving

The last_reviewed: front matter field on each of these pages records when it was last verified. Update this field every time you complete a review, even if no content changes.

Review cadence

Quarterly review — every three months.

  • Monthly is too frequent for the amount of change the review requires; it would consume disproportionate effort.
  • Annually is too slow. The AI model landscape can shift significantly in a quarter.
  • Quarterly strikes the right balance: meaningful change has usually occurred, but the review is predictable and bounded.

Suggested schedule: reviews in January, April, July, and October.

Who does the review: Cereal researches what has changed; Emmanuel makes the content updates. Both should sign off before changes are merged.

Quarterly review checklist

For each of the three pages above, complete the following steps:

Models table (copilot/models.md)

  • Check the GitHub Copilot model docs for the current model list
  • Verify each model in the table is still available in GitHub Copilot
  • Add any significant new models (focus on generally available, not preview-only)
  • Remove or annotate models that have been deprecated or withdrawn
  • Check model version names — providers rename and version models (e.g., “Sonnet 3.5” → “Sonnet 4.5”)
  • Update the last_reviewed: date in the front matter

Product list (what-is-an-ai-coding-assistant.md)

  • Check that all listed products are still active and maintained
  • Check if any major new AI coding assistant has reached significant market adoption
  • Verify the stated differentiators are still accurate (e.g., if Cursor has added agent mode, update the description)
  • Update the last_reviewed: date in the front matter

Landscape page (landscape/index.md)

  • Repeat the model provider table check (same models as copilot/models.md)
  • Review the AI coding assistant table for accuracy
  • Review the cloud coding agents table — this category moves fastest
  • Check third-party agents mentioned (Claude coding agent, Codex agent) — are they still in public preview, or generally available?
  • Update the last_reviewed: date in the front matter

Trigger events — do not wait for the quarterly cycle

Some events require an immediate review, outside the quarterly cadence:

  • A major new model is released (GPT-6, Claude 4, Gemini 3, etc.) — check whether it is available in Copilot and whether any existing model has been superseded
  • A major new AI coding assistant launches or an existing one shuts down — update the product tables immediately
  • GitHub announces a change to Copilot’s model availability (new models added, old ones removed) — update copilot/models.md immediately
  • A cloud coding agent moves from public preview to general availability — update the landscape and relevant concept pages; remove or update preview notices

When a trigger event happens: create a GitHub issue, assign it, and resolve within one week.

Suggested GitHub issue template

When creating a review issue (triggered by cadence or event):

## Model & Product Review — [Month Year]

**Trigger:** [Quarterly cadence / New model release: X / Other]

**Pages to review:**
- [ ] `_docs/copilot/models.md`
- [ ] `_docs/concepts/what-is-an-ai-coding-assistant.md`
- [ ] `_docs/landscape/index.md`

**Assigned to:** Cereal (research) + Emmanuel (content update)
**Due:** [Date — within 1 week for trigger events, flexible for quarterly]

Notes for future maintainers

  • The last_reviewed: front matter field is the single source of truth for when a page was last checked. Always update it, even for a “no change needed” review.
  • Do not remove models from the table without checking whether the page’s prose still references them by name.
  • When model names change, search the entire _docs/ directory for the old name — it may appear in prose on other pages, not just in the model table.
  • The GitHub Copilot model docs are the authoritative source for what is currently available. Use them, not third-party sources.

Updated: