Top all-time · #8844
fleetwood.dev
Personal blog · fleetwood.dev/postsaihardwareprogramming
fleetwood.dev is a personal blog at fleetwood.dev about ai, hardware and programming. Since 2021, 8 different Hacker News readers have linked it 32 times, most often “A case for client-side machine learning”. It is still writing: the last post was on 2025-07-05, and you can follow it by feed.
Too few sampled comments (2) to say how often HN recommends it.?
Why HN links it
Posts on running machine learning client-side in the browser, covering the WebNN API and running LLMs in the browser with WebGPU, including its limitations. Commenters link them as background.
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.
fleetwood.dev
Most-linked posts?
- A case for client-side machine learning3 people · 8 mentions · first linked 2022-10-31
- Layer Normalization as Fast as Possible3 people · 6 mentions · first linked 2023-12-12
- Domain specific architectures for AI inference3 people · 5 mentions · first linked 2025-03-08
- Running LLMs in the Browser with Rust and WebGPU2 people · 3 mentions · first linked 2023-04-26
What Hacker News says?
I described this system word for word 2 years ago, glad to see it come to fruition on the only software stack integrated enough to do it. https://fleetwood.dev/posts/a-case-for-client-side-machine-l...
Comment linking /posts/a-case-for-client-side-machine-learning
on “A look at Apple's technical approach to AI including core model performance etc.”Submitted You could have designed state of the art positional encoding as a story · 216 points
HN discussionRead this book for the first time recently, a huge amount of prescient ideas in there. I wrote a blog post exploring how far away glasses like Manfreds are: https://fleetwood.dev/posts/a-first-principles-analysis-of-c...
Comment linking /posts/a-first-principles-analysis-of-consumer-smart-glasses.mdx
on “Accelerando (2005)”Submitted Running LLMs in the Browser with Rust and WebGPU as a story · 8 points
HN discussionThis isn't correct. It's 100% feasible to run very capable models on consumer devices. I wrote up a feasibility investigation last year: https://fleetwood.dev/posts/a-case-for-client-side-machine-l...
Comment linking /posts/a-case-for-client-side-machine-learning
on “Stability AI Launches the First of Its StableLM Suite of Language Models”Submitted Domain specific architectures for AI inference as a story · 4 points
HN discussionhttps://fleetwood.dev/posts/a-case-for-client-side-machine-l...
Comment linking /posts/a-case-for-client-side-machine-learning
on “Ask HN: What's the idea behind the WebNN API?”Submitted Domain specific architectures for AI inference as a story · 4 points
HN discussionWhy were you skeptical of it originally? I wrote about this coming to pass over a year ago now: https://fleetwood.dev/posts/a-case-for-client-side-machine-l...
Comment linking /posts/a-case-for-client-side-machine-learning
on “Deno 1.39: The Return of WebGPU”Submitted Domain specific architectures for AI inference as a story · 3 points
HN discussionI wrote a similar blog post in Nov 22: https://fleetwood.dev/posts/a-case-for-client-side-machine-l...
Comment linking /posts/a-case-for-client-side-machine-learning
on “The Coming of Local LLMs”Submitted Domain specific architectures for AI inference as a story · 3 points
HN discussionAs an example, INT8 support in WebGPU would enable running quantized models, allowing larger LLMs to run locally in the browser. See Limitations section here: https://fleetwood.dev/posts/running-llms-in-the-browser
Comment linking /posts/running-llms-in-the-browser
on “WebGPU – All of the cores, none of the canvas”Submitted A first principles analysis of smartglasses as a story · 3 points
HN discussionSubmitted A first principles analysis of smartglasses as a story · 3 points
HN discussionSubmitted A case for client-side machine learning as a story · 3 points
HN discussionSubmitted A Rust based ML model analyzer as a story · 3 points
HN discussionSubmitted fMRI timeseries similarity with a 10x speedup using Rust as a story · 3 points
HN discussionSubmitted Arc Virtual Cell Challenge: A Primer as a story · 2 points
HN discussionSubmitted Arc Virtual Cell Challenge: A Primer as a story · 2 points
HN discussionSubmitted A first principles analysis of smartglasses as a story · 2 points
HN discussionSubmitted A case for client-side machine learning (2022) as a story · 2 points
HN discussionSubmitted A case for client-side machine learning as a story · 2 points
HN discussionSubmitted Layer Normalization as Fast as Possible as a story · 1 points
HN discussionSubmitted Domain specific architectures for AI inference as a story · 1 points
HN discussionSubmitted Layer Normalization as Fast as Possible (2023) as a story · 1 points
HN discussionSubmitted Layer Normalization as fast as possible (2023) as a story · 1 points
HN discussionSubmitted A first principles analysis of smartglasses as a story · 1 points
HN discussionSubmitted Layer Normalization as Fast as Possible as a story · 1 points
HN discussionSubmitted Layer Normalization as Fast as Possible as a story · 1 points
HN discussion