gpt-6IMAGE
Testing the rumored Astra token-saving method burned 19% of quota on setup
To trim AGENTS.md down to 5-10 lines, the author tuned his environment on Astra High and spent 19% of his quota on the setup alone before the actual test began.
テツメモ｜AI図解×検証｜Newsletter
@tetumemoArchived 2026.09.20Originally posted 2026.09.19
テツメモ｜AI図解×検証｜Newsletter @tetumemoOriginally posted 2026.09.19
📝早速話題のGPT-6 Astraのトークン消費を劇的に抑えると噂の方法を仕込むため、GPT-6 AstraのHighで自分の環境を調整 目標はAGENTS mdを5〜10行にするだけ けど、その調整作業にもうトークン19%も使ってしまったから、ここからが実際に検証フェーズ！
This post quotes another post: x.com/tetumemo/status/2098656305309421766
x.com/tetumemo/status/2101258603676225935Archived 2026.09.20 10:10