Incidents mapped to where the labs operate — an OpenAI story lands in San Francisco, a Mistral story in Paris. Below, the state of AI capabilities: model and agent benchmarks plotted over time, from public data.
Threat-intel headlines, classified by capability and pinned to the provider whose models are named. Click a provider on the map — or a filter below — to isolate their incidents.
Models named in tracked incident reports. A mention is a detection, not an accusation — it marks where attention is going, and pins the story to that lab's city.
Each dot is a lab's operating location. A glowing ring means their models were named in tracked incidents — the wider the ring, the more mentions. No ring means quiet. This is an indicator board, not a battlefield: it answers who and where, so operators know which provider's status page to check first.
Frontier scores on the benchmarks that matter — coding agents, general assistants, terminal agents, research agents — tracked over time from public reports. When a curve flattens at the top, the frontier has moved to a harder test.
—
| Date | Model / system | Score | Source |
|---|
Curated frontier snapshots — representative scores compiled from public reports, not an exhaustive leaderboard. Dates are report months, approximate. Methodologies differ between benchmarks; compare within a curve, not across them. awesome-llm-bench tracks the full picture.
The radar below watches the command center's own fleet — the same view operators see inside. Flip back to Providers for the world.
When an incident report names a model, the story pins to that provider's operating city — San Francisco for OpenAI, Paris for Mistral, Hangzhou for DeepSeek. Rings show attention, not attacks.
Coding agents, general assistants, terminal operators, research agents — each curve is the best reported score over time. Flat at the top means the benchmark is saturated and the frontier moved on.
Everything on this page is observed from outside the execution path — the same principle the command center applies to your fleet. Monitoring that can't be dragged down with the ship.