Submissions from epochai.substack.com

		Frontier labs don't use most AI compute (yet) (epochai.substack.com)
		2 points by gmays 20 days ago \| past
		The economics of superstar AI researchers (epochai.substack.com)
		1 point by gmays 23 days ago \| past
		OpenAI Stargate: where the US sites stand (epochai.substack.com)
		2 points by gmays 49 days ago \| past \| 1 comment
		Google controls the most AI computing power, driven by its custom TPUs (epochai.substack.com)
		1 point by speckx 58 days ago \| past
		What do frontier AI companies' job postings reveal about their plans? (epochai.substack.com)
		1 point by gmays 72 days ago \| past
		What do frontier AI companies' job postings reveal about their plans? (epochai.substack.com)
		1 point by vinhnx 75 days ago \| past
		Final training runs account for a minority of R&D compute spending (epochai.substack.com)
		2 points by gmays 77 days ago \| past
		First AI Solution on FrontierMath: Open Problems (epochai.substack.com)
		4 points by Philpax 80 days ago \| past
		The least understood driver of AI progress (epochai.substack.com)
		2 points by speckx 3 months ago \| past
		How persistent is the inference cost burden? (epochai.substack.com)
		1 point by gmays 3 months ago \| past
		Is almost everyone wrong about America's AI power problem? (epochai.substack.com)
		1 point by speckx 4 months ago \| past
		The changing drivers of LLM adoption (epochai.substack.com)
		2 points by gmays 5 months ago \| past
		Benchmark Scores = General Capability and Claudiness (epochai.substack.com)
		1 point by janpio 6 months ago \| past
		How Fast Can Algorithms Advance Capabilities? (epochai.substack.com)
		4 points by pseudolus on May 17, 2025 \| past
		The case for multi-decade AI timelines (epochai.substack.com)
		4 points by paulpauper on April 28, 2025 \| past
		AI progress is about to speed up (epochai.substack.com)
		2 points by Philpax on Feb 25, 2025 \| past