Top all-time · #5448
Epoch AI
Company blog · epochai.substack.comaidata
Epoch AI is the company blog at epochai.substack.com, writing about ai and data. Since 2025, 15 different Hacker News readers have linked it 25 times, most often “First AI Solution on FrontierMath: Open Problems”. It is still writing: the last post was on 2026-10-30, and you can follow it by feed.
Too few sampled comments (8) to say how often HN recommends it.?
Why HN links it
HN readers mostly link Epoch AI's Substack posts as sources on AI compute, data centers, AI company economics, and FrontierMath results. The comments cite these posts rather than judging the writing.
Summarised by Claude from a sample of 8 HN comments that link here. Each comment is marked recommending, critical or neutral (citing without judging); how this works.
Insights and analysis on AI trends, developments, and research from Epoch AI.
Most-linked posts?
- First AI Solution on FrontierMath: Open Problems2 people · 2 mentions · first linked 2026-03-23
- What do frontier AI companies' job postings reveal about their plans?2 people · 2 mentions · first linked 2026-03-28
- Frontier labs don't use most AI compute (yet)2 people · 2 mentions · first linked 2026-05-22
What Hacker News says?
Epoch AI | REMOTE | Chief Technology Officer | Full-time | $200K-$300K Epoch AI is a research institute investigating artificial intelligence's trajectory and societal impact. We develop a comprehensive, publicly accessible knowledge base on AI that informs policymakers, industry leaders, and society. For examples of our work, see our recently released compute-centric model of AI automation and its economic effects…
Comment linking /
on “Ask HN: Who is hiring? (April 2025)”Submitted How Fast Can Algorithms Advance Capabilities? as a story · 4 points
HN discussionBorcherds has been working with Epoch AI since at least 2024: https://epochai.substack.com/p/ai-math-chat-1-thinking-smart... Now he issues the standard taking points about looms and horses and claims he could not have foreseen it? He is an AI shill, just like Tao.
Comment linking /p/ai-math-chat-1-thinking-smarter-not
on “AI use leads UC Berkeley math professor to see 'no point' in grading homework”Submitted The case for multi-decade AI timelines as a story · 4 points
HN discussionIt's quite clear that these companies do make money on each marginal token. They've said this directly and analysts agree [1]. It's less clear that the margins are high enough to pay off the up-front cost of training each model. [1] https://epochai.substack.com/p/can-ai-companies-become-profi...
Comment linking /p/can-ai-companies-become-profitable
on “Claude Opus 4.6”Submitted OpenAI Stargate: where the US sites stand as a story · 2 points
HN discussionThe wild thing to me, is that they're serving $47B run rate worth of requests on maybe 2-3 GW of compute currently [1], of which only a fraction goes to inference, vs R&D and training. Obviously there have been complaints on token limits and such so they're stretched a bit thin, but nonetheless. Hard to imagine what a world with 100GW of compute looks like. [1] https://epochai.substack.com/p/frontier-labs-dont-use-m…
Comment linking /p/frontier-labs-dont-use-most-ai-compute
on “Anthropic raises $65B in Series H funding at $965B post-money valuation”Submitted First AI Solution on FrontierMath: Open Problems as a story · 2 points
HN discussionthis is outdated stuff from 3 years ago. > If you trained an llm only to map some input to the output of John Steinbeck this is literally not possible because the llm does not get generalised reasoning ability. this is not a useful hypothetical because such an llm will simply not work. why do you think you have never seen a domain specific model ever? if you wanted to falsify this claim: "llm's cant reason" how woul…
Comment linking /p/gpt-54-set-a-new-record-on-frontiermath
on “Grammarly is using our identities without permission”Submitted AI progress is about to speed up as a story · 2 points
HN discussion> The near billion users OpenAI has They're losing market share and the growth of active user plateaued. They captured all the normies who learned about llms on TV but these people will never spend a cent as you said. They're not even on the top 10 most used llms on openrouter anymore: https://openrouter.ai/rankings At the current pace anthropic will make more money than openai soon: https://epochai.substack.com/p/a…
Comment linking /p/anthropic-could-surpass-openai-in
on “How Will OpenAI Compete?”Submitted Will financing bottleneck AI compute? An Anthropic case study as a story · 1 points
HN discussionhttps://epochai.substack.com/p/first-ai-solution-on-frontier... >The newly-solved problem came from Will Brian, who had placed it in the Moderately Interesting category. It is a conjecture from a paper he wrote with Paul Larson in 2019. They were unable to solve it at the time, or in several attempts since. Brian had this to say.
Comment linking /p/first-ai-solution-on-frontiermath
on “Epoch confirms GPT5.4 Pro solved a Frontier Math Open Problem for the first time”Submitted Frontier labs don't use most AI compute (yet) as a story · 1 points
HN discussionSee also: https://epochai.substack.com/p/introducing-the-frontier-data... https://epoch.ai/data/data-centers
Comment linking /p/introducing-the-frontier-data-centers
on “Open database of large AI data centers, using satellite and permit data”Submitted The economics of superstar AI researchers as a story · 1 points
HN discussionSubmitted Google controls the most AI computing power, driven by its custom TPUs as a story · 1 points
HN discussionSubmitted What do frontier AI companies' job postings reveal about their plans? as a story · 1 points
HN discussionSubmitted What do frontier AI companies' job postings reveal about their plans? as a story · 1 points
HN discussionSubmitted Final training runs account for a minority of R&D compute spending as a story · 1 points
HN discussionSubmitted The least understood driver of AI progress as a story · 1 points
HN discussionSubmitted How persistent is the inference cost burden? as a story · 1 points
HN discussionSubmitted Is almost everyone wrong about America's AI power problem? as a story · 1 points
HN discussionSubmitted The changing drivers of LLM adoption as a story · 1 points
HN discussionSubmitted Benchmark Scores = General Capability and Claudiness as a story · 1 points
HN discussion
