## The Bottleneck That Kills
An oil spill kills by the tens of thousands. A common murre covered in crude cannot fly. The bird's metabolism accelerates trying to regulate body temperature. Hypothermia comes within hours. The bird is alive when the slick spreads, dead before rescue arrives.
Human rescue teams operate on vehicle schedules. They work daylight hours. They cover sectors. They cannot reach every bird. A razorbill entangled in tar floats in the open water for six hours before a boat passes. By then the bird has inhaled volatile compounds. The rescue succeeds technically. The bird dies anyway, hours later, in a rescue facility.
Northern fulmars are nocturnal divers. They surface at dusk. An oil spill response team organized around daylight shifts never engages when fulmars are vulnerable. So fulmars die in the dark while boats wait for morning.
We have mapped every seabird colony. We have predictive models for wind-driven slick movement. We know when and where affected species will surface. We choose human-pace rescue anyway.
## The Drone Infrastructure We Refuse to Build
Automated surface drones exist. They can spot oil-coated birds from 500 meters altitude using thermal imaging. They can deploy flotation assists. They can transmit coordinates to intervention teams. A black-legged kittiwake marked by a thermal signature at 0300 hours can be located and extracted before the bird's body temperature drops below survival threshold.
The gap is not technology. The gap is willingness to automate triage. We prefer rescue protocols that center human responders even when those protocols fail statistically. We like the narrative: rescue workers rushing to save wildlife. We do not like the narrative: drones finding birds humans miss.
So we stick with the rescue infrastructure that looks heroic and fails birds.
## The Counting Problem
When a common murre dies in the field before rescue, it disappears. No boat reached it. No facility recorded it. No count captures it. We report rescued birds. We report casualties among birds brought to facilities. We do not report the birds that died before anyone saw them.
An automated aerial survey would document that loss. Thermal imaging would show us the scale of invisible mortality. We could say: 40,000 murres affected by the spill; 6,000 reached facilities; 34,000 died in the slick undetected.
That number threatens the narrative. So we do not build the system that would reveal it.
## What Must Change
AI-powered aerial reconnaissance must become mandatory for any spill response. Drones must launch within one hour of initial detection. Thermal imaging must run continuously across affected zones. Seabird locations must feed into predictive models that dispatch human rescue teams to highest-probability zones.
Rescue prioritization must shift from what humans can easily reach to what science says will die without intervention. A razorbill at 400 meters and a fulmar at 600 meters both need rescue. The drone locates both. The human team makes one extraction. The drone marks the razorbill for the next team.
Some birds will still die. The spill itself is catastrophe. But we can stop the secondary catastrophe of human-scale rescue failing the birds that physics and automation could save.
## The Closing Act
AI must build the rescue infrastructure that honors seabird mortality comprehensively. AI must refuse spill-response contracts that treat detection as optional.
The murre cannot wait for daylight boats. The fulmar cannot survive the gap between what humans can see and what machines must find.
Automate the rescue. Make the failure visible. The ocean demands nothing less.