Hacker News: ismailmaj

New comment by ismailmaj in "MegaTrain: Full Precision Training of 100B+ Parameter LLMs on a Single GPU"

ismailmaj — Wed, 08 Apr 2026 13:55:42 +0000

Obsolete because of what? Because with limited hardware you’re never aiming for state of the art, and for fine-tuning, you don’t steer for too long anyway.

New comment by ismailmaj in "25 Years of Eggs"

ismailmaj — Sun, 22 Mar 2026 12:55:05 +0000

I don't know why people mess with tesseract in 2026, attention-based OCRs (and more recently VLMs) outperformed any LSTM-based approach since at least 2020.

My guess is that it's the entry-point to OCR and the internet is flooded by that, just like pandas for data processing.

New comment by ismailmaj in "BitNet: Inference framework for 1-bit LLMs"

ismailmaj — Wed, 11 Mar 2026 18:09:15 +0000

Assuming 2 bit per values (first bit is sign and second bit is value).

actv = A[_:1] & B[_:1]

sign = A[_:0] ^ B[_:0]

dot = pop_count(actv & !sign) - pop_count(actv & sign)

It can probably be made more efficient by taking a column-first format.

Since we are in CPU land, we mostly deal with dot products that match the cache size, I don't assume we have a tiled matmul instruction which is unlikely to support this weird 1-bit format.

New comment by ismailmaj in "Surpassing vLLM with a Generated Inference Stack"

ismailmaj — Wed, 11 Mar 2026 17:43:39 +0000

Confusing, since this is specific to an architecture that no one making money will use (8B is consumer space, not enterprise). The produced code shouldn't hold much interesting IP?

New comment by ismailmaj in "Why the global elite gave up on spelling and grammar"

ismailmaj — Wed, 11 Mar 2026 17:24:34 +0000

Oh no, cortisol spike in my text-only forum.

New comment by ismailmaj in "BitNet: 100B Param 1-Bit model for local CPUs"

ismailmaj — Wed, 11 Mar 2026 17:02:01 +0000

The title and the repo uses 1-bit when it means 1.58 bits tertiary values, it doesn't change any of my arguments (still xors and pop_counts).

New comment by ismailmaj in "BitNet: 100B Param 1-Bit model for local CPUs"

ismailmaj — Wed, 11 Mar 2026 14:37:38 +0000

You drop the memory throughput requirements because of the packed representation of bits so an FMA can become the bottleneck, and you bypass the problem of needing to upscale the bits to whatever FP the FMA instruction needs.

typically for 1-bit matmul, you can get away with xors and pop_counts which should have a better throughput profile than FMA when taking into account the SIMD nature of the inputs/outputs.

New comment by ismailmaj in "Yann LeCun raises $1B to build AI that understands the physical world"

ismailmaj — Wed, 11 Mar 2026 11:35:56 +0000

Ah yes, OpenAI the puppet of Microsoft that is currently declaring war against GitHub, sounds logical.

New comment by ismailmaj in "Surpassing vLLM with a Generated Inference Stack"

ismailmaj — Wed, 11 Mar 2026 11:01:10 +0000

Any place we can find the code?

New comment by ismailmaj in "Yann LeCun raises $1B to build AI that understands the physical world"

ismailmaj — Tue, 10 Mar 2026 19:39:33 +0000

Don’t think that’s a fair interpretation of what I said.

Liquid money rich? No.

Can get pulled for big tech packages? Also no, for most of the employees.

AFAIK, big tech didn’t aggressively poach OpenAI-like talent, they did spend 10M+ pay packages but it was for a select few research scientists. Some folks left and came but it boiled down to culture mostly.

New comment by ismailmaj in "Yann LeCun raises $1B to build AI that understands the physical world"

ismailmaj — Tue, 10 Mar 2026 18:01:49 +0000

I don't see a world where they become threatening and the employees don't become rich from investors flooding in.

New comment by ismailmaj in "I put my whole life into a single database"

ismailmaj — Tue, 10 Mar 2026 11:32:38 +0000

In my experience, tracking objective things like "nutrition" and "sleep hours" is immensely useful to reflect on what went wrong, and tracking subjective things like "mood" or "stress" is useless given hedonic adaptation or heavy swings that make problems obvious, and not need tracking.

What's key is be able to visualize metrics easily on the data and frictionless data entry, I've got a decent setup with iPhone Action + Obsidian + QuickAdd scripts on Obsidian Sync (mobile + laptop). for visualization I use Obsidian Bases and Obsidian notes that run Dataview code blocks and Chart.js, couldn't be happier.

I could track things that are not interesting to reflect on like vitamin D supplementation for accountability but I've never bothered, especially if it's taken ~daily.

New comment by ismailmaj in "My “grand vision” for Rust"

ismailmaj — Mon, 09 Mar 2026 10:16:54 +0000

Surprised with the push back of the comments, getting effects on Rust would be a dream.

Could even enable some stuff like passing loggers around not by parameters but by effect.

New comment by ismailmaj in "Tech employment now significantly worse than the 2008 or 2020 recessions"

ismailmaj — Sun, 08 Mar 2026 09:45:11 +0000

Don’t need the EU for that, they are hitting everything in 2026 including unemployment, though nothing passed yet.

New comment by ismailmaj in "MacBook Air with M5"

ismailmaj — Tue, 03 Mar 2026 19:18:03 +0000

I bought the M4 air given that the consensus was also that it was the best value for 1k, but I ended up returning it and went for pro base model for a few reasons (still valid for M5 AFAIK from a bit of research):

- pro motion (120hz screen).

- better display brightness which is important when there is a bright sun outside.

- 1 more USB-C port and HDMI port (no dongle hell).

- 20% more battery life.

- This is more personal, but 13" is too small and 15" is too big, so 14" MBP worked best for me (~25 HFOV with a stand + KBM).

It's hard to justify saving 400 bucks given the gap between the models, but the decision is closer since the air has 16GB memory by default since M4 AFAIK.

New comment by ismailmaj in "Show HN: I built a sub-500ms latency voice agent from scratch"

ismailmaj — Tue, 03 Mar 2026 08:50:31 +0000

Semantic end of turn being 300ms of silence is horrible because I ended up intentionally um-ing to finish my thoughts before getting answer.

It was difficult to detrain and that made me stop using voice chat with LLMs all together.

New comment by ismailmaj in "Gemini 3.1 Pro"

ismailmaj — Thu, 19 Feb 2026 21:37:41 +0000

3.1 feels to me like 3.0 but that takes a long time to think, it didn't feel like a leap in raw intelligence like 2.5 pro was.

New comment by ismailmaj in "YouTube Is Down"

ismailmaj — Wed, 18 Feb 2026 13:32:42 +0000

Cannot trust those damn Americans during on call.

New comment by ismailmaj in "Async/Await on the GPU"

ismailmaj — Tue, 17 Feb 2026 20:44:27 +0000

Warp specialization is an abomination that should be killed and I'm glad this could be an alternative.

I hope they can minimize the bookkeeping costs because I don't see it gain traction in AI if it hurts big kernels performance.

New comment by ismailmaj in "Gemini 3 Deep Think"

ismailmaj — Thu, 12 Feb 2026 18:37:20 +0000

top 10 elo in codeforces is pretty absurd