Hacker News: hislaziness

New comment by hislaziness in "The current state of LLM-driven development"

hislaziness — Tue, 12 Aug 2025 02:10:30 +0000

Would it be more appropriate to compare LLMs to Autotunes rather than pianos?

New comment by hislaziness in "How to break the 'AI hype cycle'"

hislaziness — Mon, 21 Jul 2025 23:54:28 +0000

Akamai CTO Robert Blumofe offers four tips for business leaders striving to foster AI fluency by empowering employees with the right tools and best use cases.

useful insights but definitely a PR article.

New comment by hislaziness in "What the Fuck Python"

hislaziness — Sun, 20 Jul 2025 01:56:36 +0000

Great. I enjoy these kind of articles. My all time favorite book for 'C' is Expert C Programming: Deep C Secrets.

New comment by hislaziness in "OpenAI claims gold-medal performance at IMO 2025"

hislaziness — Sun, 20 Jul 2025 01:45:59 +0000

Terence Tao on the matter - https://imgur.com/a/terence-tao-on-supposed-gold-imo-sMKP0bm

New comment by hislaziness in "Controllable Fast and Slow Thinking by Learning with Randomized Reasoning Traces"

hislaziness — Wed, 23 Oct 2024 05:03:31 +0000

As I understand, the LLM uses the techniques of searchformer - https://arxiv.org/abs/2402.14083. To do "slow thinking" doing a A* search using a transofrmer.

Controllable Fast and Slow Thinking by Learning with Randomized Reasoning Traces

hislaziness — Wed, 23 Oct 2024 03:51:45 +0000

Article URL: https://arxiv.org/abs/2410.09918

Comments URL: https://news.ycombinator.com/item?id=41921635

Points: 44

# Comments: 15

New comment by hislaziness in "$2 H100s: How the GPU Rental Bubble Burst"

hislaziness — Fri, 11 Oct 2024 04:03:18 +0000

It is not just MSRP, management and operations cost too. The article goes into the details of this.

New comment by hislaziness in "$2 H100s: How the GPU Rental Bubble Burst"

hislaziness — Fri, 11 Oct 2024 03:57:26 +0000

The details are in the article. They have done the math.

New comment by hislaziness in "$2 H100s: How the GPU Rental Bubble Burst"

hislaziness — Fri, 11 Oct 2024 02:49:05 +0000

TLDR: Don’t buy H100s. The market has flipped from shortage ($8/hr) to oversupplied ($2/hr), because of reserved compute resales, open model finetuning, and decline in new foundation model co’s. Rent instead.

Is the AI infra bubble already bursting?

New comment by hislaziness in "Task-Switching Experiment (2015)"

hislaziness — Mon, 07 Oct 2024 09:43:20 +0000

The screen seems to be stuck at Wait a second... for me.

New comment by hislaziness in "Reflection 70B, the top open-source model"

hislaziness — Sat, 07 Sep 2024 01:20:18 +0000

I tried a few local LLMs. None of them could give me the right answer for "How many 'r's in straberry. All LLMs were 8-27B.

New comment by hislaziness in "Looming Liability Machines (LLMs)"

hislaziness — Sun, 25 Aug 2024 01:42:28 +0000

I know you mean this in jest, but we are much closer to this than we would imagine, the use of LLMs to process communication / translation is becoming ubiquitous. We are 1 bad translation away from a disaster.

New comment by hislaziness in "Psssst! Your date of birth can be a random number!"

hislaziness — Thu, 15 Aug 2024 07:14:15 +0000

I also use some email providers ability to have +xyz at the end of the username. So for a registration I would for user.name+sitea@domain.com. has helped me track spam and leaks in the past.

New comment by hislaziness in "Llama 3.1 in C"

hislaziness — Wed, 24 Jul 2024 05:19:23 +0000

Cool. I will try it out. I tried the same with ollama, the non english part needs a lot more polish. Do you see the outcome being any different?

New comment by hislaziness in "CLISP Port to WASM"

hislaziness — Wed, 24 Jul 2024 02:23:19 +0000

This is pretty awesome. How did you do it? Any blog on the detials?

New comment by hislaziness in "Mistral NeMo"

hislaziness — Fri, 19 Jul 2024 00:04:36 +0000

The model description on huggingface says - Model size - 12.2B params, Tensor type - BF16. Is the Tensor type different from the training param size?

New comment by hislaziness in "Mistral NeMo"

hislaziness — Fri, 19 Jul 2024 00:01:19 +0000

I just checked huggingface and the model files download is about 25GB but in a comment below someone mentioned it is 8fp quantized model. Trying to understand how the quantization affects the model (and RAM) size. Can someone please enlighten.

New comment by hislaziness in "Mistral NeMo"

hislaziness — Thu, 18 Jul 2024 16:03:14 +0000

isn't it 2 bytes (fp16) per param. so 7b = 14 GB+some for inference?

New comment by hislaziness in "[dead]"

hislaziness — Mon, 15 Jul 2024 14:29:05 +0000

The title does not do justice to the article. It talks about a new classification system that OpenAI has introduced for LLMs

New comment by hislaziness in "ThankYouHN: 14 Years"

hislaziness — Fri, 07 Jun 2024 04:11:47 +0000

Same here. Coincidentally I joined in June 2009 and this is the only social media I indulge in.