Top all-time · #7406
Finbarr Timbers
Personal blog · finbarr.ca · on HN as fnbraiprogramming
Finbarr Timbers's personal blog at finbarr.ca covers ai and programming. Since 2017, 11 different Hacker News readers have linked it 24 times, most often “Large language models aren't trained enough”. It is still writing: the last post was on 2026-04-15, and you can follow it by feed.
Too few sampled comments (2) to say how often HN recommends it.?
Why HN links it
HN readers link Finbarr Timbers' posts on LLMs, such as one arguing LLMs are not trained enough under Chinchilla-optimal reasoning and one on how llama.cpp works. Commenters use them to back points on model training and inference.
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.
Most-linked posts?
- Large language models aren't trained enough3 people · 4 mentions · first linked 2023-02-28
- Five Years of GPT Progress2 people · 2 mentions · first linked 2023-04-02
- What, Is the GIL?2 people · 2 mentions · first linked 2026-04-11
- How Is LLaMa.cpp Possible?2 people · 2 mentions · first linked 2023-08-15
What Hacker News says?
True. There was a good blog post published about this a few weeks ago: https://finbarr.ca/llms-not-trained-enough/ Money quote for those who don't want to read the whole thing: ''' When people talk about training a Chinchilla-optimal model, this is what they mean: training a model that matches their estimates for optimality. They estimated the optimal model size for a given compute budget, and the optimal number of…
Comment linking /llms-not-trained-enough/
on “Cerebras-GPT: A Family of Open, Compute-Efficient, Large Language Models”Submitted Five Years of GPT Progress as a story · 6 points
HN discussionSEEKING WORK - Remote preferred - Based in Edmonton, AB, Canada. Willing to travel. Machine Learning Engineer, focusing on image recognition & NLP Technologies: Python, PyTorch, Tensorflow, Keras, Flask, Bash, Linux, Postgresql, MySQL, HTML, CSS, Git. Github: https://github.com/finbarrtimbers Email: finbarrtimbers a-t gmail dot com Site: https://Finbarr.ca I specialize in developing machine learning proof-of-concept…
Comment linking /
on “Ask HN: Freelancer? Seeking freelancer? (July 2017)”Submitted How Is LLaMa.cpp Possible? as a story · 4 points
HN discussionLocation: Calgary, Canada Remote: Yes Willing to relocate: Yes Technologies: Python, R, SQL, Bash, machine learning. Résumé/CV: http://finbarr.ca/cv.pdf Email: finbarrtimbers@gmail.com I'm a data scientist/machine learning engineer with two years of experience in a combined engineering/machine learning role. I've spent the last two years solving real world business problems, doing everything from doing pure analysis…
Comment linking /cv.pdf
on “Ask HN: Who wants to be hired? (April 2017)”Submitted Notes on “Training ImageNet in 1 Hour” as a story · 3 points
HN discussionSEEKING WORK - Remote preferred - Based in Edmonton, AB, Canada. Willing to travel. Machine Learning Engineer, focusing on image recognition & NLP Technologies: Python, PyTorch, Tensorflow, Keras, Flask, Bash, Linux, Postgresql, MySQL, HTML, CSS, Git. Github: https://github.com/finbarrtimbers Email: finbarrtimbers a-t gmail dot com Site: https://Finbarr.ca I specialize in developing machine learning proof-of-concept…
Comment linking /
on “Ask HN: Freelancer? Seeking freelancer? (September 2017)”Submitted Useful bash oneliners as a story · 3 points
HN discussionLocation: Edmonton, Canada, but looking to relocate Remote: Yes Willing to relocate: Yes, particularly to the Bay Area, or NYC. I'm a Canadian citizen so I can easily get a TN-1 visa. Technologies: Python (sklearn, Numpy, Pandas, Tensorflow, PyTorch, Keras), R, SQL, Javascript, C++ Resume: https://finbarr.ca/cv.pdf Email: finbarrtimbers [a-t] gmail [d-o-t] com I'm currently working as a data scientist/ML Engineer fo…
Comment linking /dedup
on “Ask HN: Who wants to be hired? (September 2017)”Submitted What, Is the GIL? as a story · 2 points
HN discussionSEEKING WORK - Remote preferred - Based in Edmonton, AB, Canada. Willing to travel. Machine Learning Engineer, focusing on image recognition & NLP Technologies: Python, PyTorch, Tensorflow, Keras, Flask, Bash, Linux, Postgresql, MySQL, HTML, CSS, Git. Github: https://github.com/finbarrtimbers Email: finbarrtimbers a-t gmail dot com Site: https://Finbarr.ca I specialize in developing machine learning proof-of-concept…
Comment linking /
on “Ask HN: Freelancer? Seeking freelancer? (August 2017)”Submitted The Bitter Lesson as a story · 2 points
HN discussionAccording to this [1] article (current top of hn) memory bandwidth is typically the limiting factor, so as long as your batch size isn't huge you probably aren't losing too much in performance. https://finbarr.ca/how-is-llama-cpp-possible/
Comment linking /how-is-llama-cpp-possible/
on “Why host your own LLM?”Submitted Large language models aren't trained enough as a story · 2 points
HN discussionYeah, you’re totally right. I actually wrote a follow up essay about that: https://finbarr.ca/llms-not-trained-enough/ I think the conversations were partly (largely?) a snapshot in time. I was talking to people in February/March, and all of this was much less thought through at the time. But you’re totally right. You want something like Llama, where you train a smaller model longer than Chinchilla would predict.
Comment linking /llms-not-trained-enough/
on “Five Years of LLM Progress”Submitted Five Years of LLM Progress as a story · 2 points
HN discussionSubmitted Large language models aren't trained enough as a story · 2 points
HN discussionSubmitted Notes on “Random Search for Hyper-Parameter Optimisation” as a story · 2 points
HN discussionSubmitted Unit tests make you write down your assumptions as a story · 2 points
HN discussionSubmitted How we designed our machine learning application as a story · 2 points
HN discussionSubmitted What, exactly, is the GIL? as a story · 1 points
HN discussionSubmitted 5 Years of GPT progress [pdf] as a story · 1 points
HN discussionSubmitted Bug Report Deduplication as a story · 1 points
HN discussion
