Not arguing "summers are cooling" and never have. They have definitely been warming. Just pointing out the strangeness of people both lamenting how cool/wet spring is and how it drags on forever, and yet others saying it's just become the start of summer. In what world does this make sense?

Certainly seems less likely with the ECMWF, EPS, and AI models not agreeing.

I guess 75 and sunny with no breeze at all can get a little warm in direct sunlight. We are getting closer! But then you just sit in the shade. Its not that hard. I like at least having the option to be in warm sun if I want.

AI gefs is pretty consistent

this is what Claude says about it: GFS AI (AIGFS) vs Traditional GFS NOAA just launched the AI Global Forecast System (AIGFS) in December 2025 National Oceanic and Atmospheric Administration, and early testing shows promising results: AIGFS advantages: Shows improved forecast skill over traditional GFS for many large-scale features National Oceanic and Atmospheric Administration Demonstrates significant reduction in tropical cyclone track errors at longer lead times National Oceanic and Atmospheric Administration Dramatically faster: completes a 16-day forecast in approximately 40 minutes using only 0.3% of the computing resources National Oceanic and Atmospheric Administration Current limitation: Version 1.0 shows degradation in tropical cyclone intensity forecasts, which future versions will address Inside HPC & AI The Hybrid Approach The most accurate option appears to be HGEFS (Hybrid-GEFS), which combines AI-based AIGEFS with the physics-based Global Ensemble Forecast System National Oceanic and Atmospheric Administration. Initial testing shows this hybrid model consistently outperforms both the AI-only and physics-only ensemble systems across most major verification metrics National Oceanic and Atmospheric Administration. Bottom line: For now, the AI version shows improved accuracy for large-scale patterns and tropical cyclone tracks, but the traditional GFS still has advantages for intensity forecasts. The hybrid model combining both approaches performs best overall.