S&P 500 7,489.72 +0.70% NASDAQ 25,373.85 +1.00% DOW 52,485.03 +0.50% RUT 2,604.40 -0.34% BTC-USD 63,514.00 +0.80% ETH-USD 3,412.60 -0.65% 10-YR 4.12 -0.72% 2-YR 3.86 +0.52% VIX 14.92 -2.67% GOLD 3,288.40 +0.43% WTI 71.86 -1.26% NATGAS 3.14 +1.95% NVDA 188.34 +1.90% MSFT 512.60 -0.28% AAPL 231.09 +0.46% AMZN 228.77 +1.16% GOOGL 196.42 -0.45% META 704.15 +1.41% TSLA 297.50 -2.11% AMD 172.88 +2.38%

S&P 500 7,489.72 +0.70% NASDAQ 25,373.85 +1.00% DOW 52,485.03 +0.50% RUT 2,604.40 -0.34% BTC-USD 63,514.00 +0.80% ETH-USD 3,412.60 -0.65% 10-YR 4.12 -0.72% 2-YR 3.86 +0.52% VIX 14.92 -2.67% GOLD 3,288.40 +0.43% WTI 71.86 -1.26% NATGAS 3.14 +1.95% NVDA 188.34 +1.90% MSFT 512.60 -0.28% AAPL 231.09 +0.46% AMZN 228.77 +1.16% GOOGL 196.42 -0.45% META 704.15 +1.41% TSLA 297.50 -2.11% AMD 172.88 +2.38%

Home / Guides / AI & tech

Claim check True

Does more compute reliably make a better model? The scaling claim, tested

We plotted three labs' published benchmark scores against disclosed training compute. The line holds better than the memes suggest.

May 22, 2026 · 10 min read · by the JMM Research desk

Draft entry. The full analysis for this piece ships with the data pipeline; the dataset it grades against is being assembled.