A thermometer outside your window shows a single number, but in reality it's warmer near the wall than by the window. The ocean is even more capricious: a satellite sees the whole surface, but its readings get skewed in areas of strong currents. Buoys are accurate, but they're spaced far apart. How do you know the water temperature where there's no instrument at all? Previously, scientists built one map and trusted it. Now, SAGE builds many maps at once, each slightly different from the others.

SAGE works like a translator who doesn't settle on a single rendering of a phrase but shows all possible meanings. First, it learns from 39 years of historical temperature maps, memorizing what warm and cold patches in the ocean typically look like. Then, given fresh data from satellites and buoys, it draws not one map but a whole fan. In places with abundant observations, all maps are nearly identical, meaning the temperature can be trusted. Where instruments are absent, the maps diverge, and this spread honestly says: here, we don't know.

Validation against buoys that were not part of the training showed that error dropped from 0.59°C to 0.17°C. That's like a thermometer that used to be off by half a degree now being accurate to within two-tenths. And when a ten-day weather forecast started not from a single map but from SAGE's fan of maps, accuracy improved by nearly five percent. The seasonal El Niño forecast for 2023, which was previously almost useless, suddenly became reliable.

Now oceanographers can see not only the average temperature but also how much confidence to place in that picture. For marine heatwaves, which kill corals and shift fish migration routes, this is especially important. In 2023, the ocean warmed so much that such events occurred three times more often than usual, and SAGE captured their intensity to within a tenth of a degree.