Top all-time · #7852
Manifest AI
Company blog · manifestai.com/articlesaiscience
Manifest AI is the company blog at manifestai.com, writing about ai and science. Since 2024, 10 different Hacker News readers have linked it 10 times, most often “Symmetric Power Transformers”. No feed was found, so read it on the site itself.
Too few sampled comments (2) to say how often HN recommends it.?
Why HN links it
HN readers link Manifest AI's articles on Power Attention and Symmetric Power Transformers, which describe attention variants for LLMs with linear training cost and constant inference cost. Comments only announce or cite them.
Summarised by Claude from a sample of 2 HN comments that link here. Each comment is marked recommending, critical or neutral (citing without judging); how this works.
Our mission is to train a neural network to model all human output.
Most-linked posts?
- Symmetric Power Transformers3 people · 3 mentions · first linked 2024-08-15
- End of Transformer Era Approaches2 people · 2 mentions · first linked 2025-10-30
- Power Retention – Drop-In Replacement for Flash_attention2 people · 2 mentions · first linked 2025-09-24
What Hacker News says?
At Manifest AI we have just released our open-source CUDA kernels to implement Symmetric Power Transformers, as described in our paper from back in August: https://manifestai.com/articles/symmetric-power-transformers... Since this is a variant of a linear attention, you get linear cost when training (as opposed to quadratic in regular attention), and constant when doing inference. This is especially attractive for l…
Comment linking /articles/symmetric-power-transformers/
on “Power Attention: Efficient CUDA Kernels for Symmetric Power Transformers”Submitted Linear transformers are faster after all as a story · 116 points
HN discussionHere’s an article I found from Manifest AI, the company behind Power Retention: https://manifestai.com/articles/what-is-power-retention/
Comment linking /articles/what-is-power-retention/
on “PowerRetention: a drop-in replacement for FlashAttention in LLMs”Submitted Symmetric Power Transformers as a story · 18 points
HN discussionSubmitted End of Transformer Era Approaches as a story · 15 points
HN discussionSubmitted Brumby-14B-Base: The Strongest Attention-Free Base Model as a story · 7 points
HN discussionSubmitted Power Retention – Drop-In Replacement for Flash_attention as a story · 7 points
HN discussionSubmitted Symmetric Power Transformers as a story · 3 points
HN discussionSubmitted Compute-Optimal Context Size as a story · 3 points
HN discussionSubmitted Release: Power Retention as a story · 2 points
HN discussion
