UncoverAI
Follow

Release Notes

Release Notes record changes to the website. The Data Changelog records changes to tracked vendor data. Current version v1.4.0. Feed

v1.4.0

· Release Notes

A public record of website changes, separate from the log of vendor data.

  • New Feature Release Notes lists website changes by version, with the current version in the footer.
  • New Feature A roadmap shows planned website work and which dates are still targets.
  • Enhancement The vendor-data log is now titled Data Changelog, and it stays at the same address.

v1.3.1

· Monthly briefing signup

One email signup for the monthly briefing, and the maintenance notes live with the methodology.

  • Enhancement Subscribe is one email popup for the monthly Enterprise AI Briefing. It opens from the home page, the header, or a shared link.
  • Enhancement A personal email address can subscribe. The footer is the links, the research disclaimer, and the copyright line.
  • Enhancement How the index is maintained is on the Methodology page. The previous About address opens that page for now.

v1.3.0

· Cited proof on signals and sub-processors

Signal items and sub-processor changes point at the page that was checked.

  • New Feature Sub-processor diffs name what was added, with the source page and the date when the vendor states one.
  • New Feature Each Market Signals item ends with the document that was checked, including a saved snapshot when one exists.
  • Bug Fix Unpublished and future editions stay off the public site.

v1.2.2

· Shareable index tabs

A link to a model or platform tab opens that tab.

  • Enhancement Model and platform tabs have their own addresses, so a shared link opens the same tab.
  • Enhancement Sitemap dates follow the day the page content last changed.

v1.2.1

· Platform rows stay up

One held clause no longer hides the rest of a vendor, and the checked date stays on the last success.

  • Bug Fix A platform row stays published when one clause cannot be refreshed, instead of hiding the vendor.
  • Bug Fix The platforms checked date stays on the last successful refresh when a run does not finish.

v1.2.0

· Open-weight memory and model availability

Memory estimates for open-weight models, and a table of which platforms list each model.

  • New Feature Open-weight hardware estimates show hypothetical memory at FP16, FP8, and INT4, with the assumptions on the page.
  • New Feature Model availability lists which tracked models each platform publishes, with the page that was checked.

v1.1.0

· Workload fit, batch prices, and sunsets

The model index can be filtered by workload, and it shows batch prices and upcoming shutdowns when the vendor states them.

  • New Feature Workload-fit tags on the model index, computed from published prices and capabilities.
  • New Feature Cached and batch prices appear in the estimator only when the vendor page states them.
  • New Feature Upcoming sunsets show the shutdown date and the vendor notice.

v1.0.0

· Public launch

The first public version of the index: cited model prices, platform terms, and the monthly briefing.

  • New Feature Frontier model index with token prices, context windows, licenses, and a side-by-side compare.
  • New Feature Compliance and residency grid for tracked models, each cell cited to the vendor page that was checked.
  • New Feature AI platform terms, uptime commitments, and a sub-processor roster with snapshot diffs.
  • New Feature Market Signals, a monthly edition of enterprise AI events with sources.
  • New Feature A methodology for what counts as a source, and a corrections mailbox.

Roadmap

These are plans, not commitments. Dates and scope can change.

Platform terms, retention, and controls

. Planned as v1.5.0, which may change.

  • Per-platform SLA and terms pages. Amazon Bedrock, Azure AI Foundry, and Vertex AI come first: uptime SLA, service credits, and data and training terms, each with its source.
  • A zero data retention and abuse-monitoring matrix for each model.
  • Model IP indemnification and the guardrails a vendor requires.
  • Enterprise network, VPC, and key-management controls for each platform.
  • A sovereign and cross-region inference routing audit.
  • A refresh of the core platform SLA index.

Later

No date yet.

  • A latency and throughput baseline, covering time to first token and tokens per second.
  • A deployer matrix of enacted laws and the duties they trigger.
  • An on-demand information-security risk dossier.
  • A mapping onto the NIST AI Risk Management Framework and ISO 42001.