Hacker News: trekhleb

New comment by trekhleb in "Claude Fable is relentlessly proactive"

trekhleb — Fri, 12 Jun 2026 15:51:39 +0000

This article gave me another nudge towards running Claude in a Docker container.

I made a thin Docker container wrapper "claude-pod" recently for my personal usage here: https://github.com/trekhleb/claude-pod

However, I wasn't using it that often, just because of that additional friction of running Claude via `PORTS="3000 5173" claude-pod` instead of just `claude`, etc.

But now I have more motivation for the containerisation :D. Not a 100% defence from the potential glitches, though, but still something...

New comment by trekhleb in "The Smallest Brain You Can Build: A Perceptron in Python"

trekhleb — Mon, 08 Jun 2026 03:32:48 +0000

Nice and minimalistic

I played with similar approach in JavaScript and built a NanoNeuron https://github.com/trekhleb/nano-neuron (it is more verbose than Python though)

Self-Parking Car Evolution

trekhleb — Wed, 25 Mar 2026 05:26:49 +0000

Article URL: https://trekhleb.dev/self-parking-car-evolution/

Comments URL: https://news.ycombinator.com/item?id=47513631

Points: 3

# Comments: 1

A unified map of California's passenger rail systems

trekhleb — Sun, 22 Mar 2026 17:03:18 +0000

Article URL: https://trekhleb.dev/cali-vibe/transit/

Comments URL: https://news.ycombinator.com/item?id=47479583

Points: 3

# Comments: 0

Show HN: CaliVibe – Explore California Neighbourhoods

trekhleb — Fri, 13 Mar 2026 14:25:09 +0000

Article URL: https://trekhleb.dev/cali-vibe/

Comments URL: https://news.ycombinator.com/item?id=47364920

Points: 1

# Comments: 0

Dijkstra's Algorithm in JavaScript

trekhleb — Sat, 08 Nov 2025 22:11:07 +0000

Article URL: https://github.com/trekhleb/javascript-algorithms/blob/master/src/algorithms/graph/dijkstra/README.md

Comments URL: https://news.ycombinator.com/item?id=45860503

Points: 1

# Comments: 0

A map of every recorded battle based on Wikipedia

trekhleb — Sat, 08 Mar 2025 15:28:22 +0000

Article URL: https://battles.nodegoat.net/viewer.p/23/385/scenario/3/geo/

Comments URL: https://news.ycombinator.com/item?id=43300940

Points: 8

# Comments: 4

Top Tech Jobs

trekhleb — Sun, 02 Feb 2025 21:47:40 +0000

Article URL: https://www.trueup.io/

Comments URL: https://news.ycombinator.com/item?id=42912171

Points: 1

# Comments: 0

Tech Jobs Radar

trekhleb — Mon, 06 Jan 2025 15:40:55 +0000

Article URL: https://jobs-radar.com/

Comments URL: https://news.ycombinator.com/item?id=42611457

Points: 2

# Comments: 0

A collection of essential TypeScript types

trekhleb — Sat, 30 Nov 2024 20:26:43 +0000

Article URL: https://github.com/sindresorhus/type-fest

Comments URL: https://news.ycombinator.com/item?id=42283808

Points: 1

# Comments: 0

Shitcode Principles (2020)

trekhleb — Tue, 26 Nov 2024 21:54:46 +0000

Article URL: https://github.com/trekhleb/state-of-the-art-shitcode

Comments URL: https://news.ycombinator.com/item?id=42250474

Points: 1

# Comments: 0

System Design Sketches

trekhleb — Tue, 26 Nov 2024 10:35:19 +0000

Article URL: https://okso.app/showcase/system-design

Comments URL: https://news.ycombinator.com/item?id=42244441

Points: 1

# Comments: 0

New comment by trekhleb in "Ask HN: Is distributed LLM training in browsers (WebRTC and WebGPU) possible?"

trekhleb — Fri, 22 Nov 2024 22:50:50 +0000

What I mean is training something like GPT-3 in a distributed manner using a large number of regular browsers or laptops with average WebGPU support/power and WebRTC for communication.

Does it even make sense to ask this? Is it reasonable or feasible?

I understand there are many nuances, such as the size and source of the training data, the size of the model (which would be too large for any browser to handle), network overhead, and the challenge of merging all the pieces together, among others. However, speculative calculations suggest that GPT-3 required around 3x10^22 FLOPs, which might (very speculatively) be equivalent to about 3,000 regular GPUs, each with an average performance of 6 TFLOPs, training it for ~30 days (which also sounds silly, I understand).

Of course, these are naive and highly speculative calculations that don’t account for whether it’s even possible to split the dataset, model, and training process into manageable pieces across such a setup.

But if this direction is not totally nonsensical, does it mean that even with a tremendous network overhead there is a huge potential for scaling (there are potentially a lot of laptops connected to the internet that potentially and voluntary could be used for training)?

Ask HN: Is distributed LLM training in browsers (WebRTC and WebGPU) possible?

trekhleb — Fri, 22 Nov 2024 22:19:17 +0000

Comments URL: https://news.ycombinator.com/item?id=42217823

Points: 4

# Comments: 5

Postgres and MySQL: The Main Differences

trekhleb — Fri, 22 Nov 2024 10:47:44 +0000

Article URL: https://twitter.com/hnasr/status/1859366187534254274

Comments URL: https://news.ycombinator.com/item?id=42212729

Points: 3

# Comments: 0

Training Baby GPTs in Browser

trekhleb — Mon, 11 Nov 2024 20:37:54 +0000

Article URL: https://trekhleb.dev/homemade-gpt-js/

Comments URL: https://news.ycombinator.com/item?id=42110243

Points: 7

# Comments: 1

New comment by trekhleb in "Homemade GPT JS – A Tensorflow.js Re-Implementation of MinGPT"

trekhleb — Sun, 10 Nov 2024 20:35:29 +0000

Thanks for the feedback! WebGPT is good. Looks like it is a vanilla JS? I used TensorFlow.js to offload all the troubles of working with tensors, gradients, and WebGPU integration to it. Along with a possibility to train the model in the browser it also helped to keep the actual GPT code pretty concise (<300 lines). Hopefully it will make easier to learn the model architecture itself for those who’re interested.

Homemade GPT JS – A Tensorflow.js Re-Implementation of MinGPT

trekhleb — Sun, 10 Nov 2024 19:19:05 +0000

Article URL: https://github.com/trekhleb/homemade-gpt-js

Comments URL: https://news.ycombinator.com/item?id=42101998

Points: 2

# Comments: 2

System Design Sketches

trekhleb — Wed, 17 Jul 2024 12:43:48 +0000

Article URL: https://trekhleb.dev/blog/2024/system-design-sketches/

Comments URL: https://news.ycombinator.com/item?id=40985250

Points: 3

# Comments: 0

New comment by trekhleb in "Watch cars evolve using genetic algorithm"

trekhleb — Mon, 29 Apr 2024 06:45:16 +0000

It is a very visual and entertaining visualization, I love it.

It inspired me to experiment with a genetic algorithm in "Self-parking car evolution":

https://trekhleb.dev/self-parking-car-evolution/