Hacker News Reader

SethTroon ESP32-C3 Wireless Adventure: A Comprehensive Guide to IoT [pdf]

I love my ESP32-S series. I use them in tons of LED projects[1]. Sadly it seems WS2812b are fundamentally incompatible, under FreeRTOS, with wifi due to timing glitches from interrupts.

In my next two project I'm going to have to run two 30ft cables because I can figure out how to get low latency wifi to work on a Raspberry Pi 2040 (where PIO is fabulous, but 500ms request latency is killing me) or without glitches on a ESP32.

[1] https://anima.haus/events/seacompression

SethTroon Deepmind Alphadev: Faster sorting algorithms discovered using deep RL

LLVM merge: https://reviews.llvm.org/D118029 Benchmark: https://bit.ly/3AtesYf

Benchmark seems to be in the range of a 1-5% improvement for 80% of sizes.

SethTroon Show HN: Word2vec Algorithm in ~100sloc with NumPy

Right out of the bag lots of wasted lines

    word_bag: dict[str, int] = dict()  # Multiset

    for line in raw_corpus:
        words = line.split()
        
        for word in words:
            if word in word_bag:
                word_bag[word] += 1
            else:
                word_bag[word] = 1

    keys_to_drop = []
    for k, v in word_bag.items():
        if v < min_frequency:
            keys_to_drop.append(k)

    for k in keys_to_drop:
        del word_bag[k]

    pprint(word_bag)
    print(len(word_bag))

    vocabulary = word_bag.keys()
    return set(vocabulary)

Can be a python one/two liner

    # word_bag = Counter()  # defaultdict(int)
    word_bag = Counter(word for line in raw_corpus for word in line.split())
    return set(word for word, count in word_bag.items() if count >= min_frequency)

SethTroon Numbers every LLM developer should know

Another factor in favor of 8 bits is that inference speed will generally be 2x faster than a larger model at 4 bits.

SethTroon PEP 684 was accepted – Per-interpreter GIL in Python 3.12

Moderately different; Particularly with respect to how to performance.

If I read this correctly immortal objects stop changing refcount which prevents cache line invalidation while `gc.freeze()` just stops cleaning up objects with zero refcount.

SethTroon AMD Zen2 ymm registers rolling back

https://lkml.org/lkml/2023/2/22/982

Seems to be a known errata which was fixed by a microcode update

SethTroon LLaMA: A foundational, 65B-parameter large language model

(1022362 + 82432) gpu-hours / 2048gpus / 5 months ~= 15% uptime.

That's only 0.08 nines of availability!

I remember in one of their old guidebooks a lot of struggle to keep their 64 machine (512 gpu) cluster running this was probably 4x the machines and 4x the number of cluster dropouts.

SethTroon The Busy Beaver Challenge

Some discussion of remaining infinite BB(5) machines yesterday in [1][2]

[1] https://www.sligocki.com//2023/02/02/skelet-34.html

[2] https://news.ycombinator.com/item?id=34689068 and

SethTroon Builder's Remedy goes into effect in many California cities tomorrow

No, they choose a percentage of income for housing, I think 30% or 35%, and calculate from that so 50K * 1.2 * 0.3 = 18K/year = $1500/month

SethTroon Seven years on, what do we know about the disappearance of flight MH370? (2021)

Should have [2021]

SethTroon Google to reduce workforce by 12k

Do you know the subject line of the follow up email (for those laid off)?

SethTroon GNU Terry Pratchett

Previous discussion: https://news.ycombinator.com/item?id=12844057

SethTroon More evidence for problems in VM warmup

A good blog post has lots of hard parts (layout, scope, visuals, audience). Laurence Tratt, you nailed it for me! I loved the details you put in "Benchmarking methodology" and the clean layout with instructive visuals.

SethTroon Zendo: The Game of Inductive Logic

You can 3D print cones if you have access to a 3d printer. I've made a few larger and smaller sets for work, travel, home.

https://www.thingiverse.com/thing:3784941

SethTroon A C++17 thread pool for high-performance scientific computing

The paper has only a single benchmark reported from a single system where they report

> In our performance test, we see a speedup of 18.2x, saturating and even surpassing this estimated theoretical upper bound.

IMHO if you exceed your theoretical bound that's a sign you didn't go a good job analyzing the situation.

SethTroon Almost 1/3 of San Francisco charging stations are broken

Title is a bit sensational: Should really be "1/4 of SF charging stations are broke". They rounded 22.7% up to 33.3% bypassing the more logical/sensibleness 25%

SethTroon DarkPattern.games: Find Healthy Mobile Games

I hate absolute ratings (e.g. 5/5 with 1 vote being ranked higher than 4.99/5 with 100 votes).

All the top games seem to have 1 report giving everything positive ratings so that other fabulous games with more ratings but say a 4.97 rating are lost.

https://www.evanmiller.org/how-not-to-sort-by-average-rating...

SethTroon A Minimum Viable Computer, or Linux for $15

https://oshpark.com/ is my favorite for low volume runs. You get 3 copies for $5/in^2

SethTroon BitTorrent v2 (2020)

It is highly unlikely that the duplicate portions of the file will have an offset that's a multiple of 2^16 which would be required for chunks to have matching hashes. On the client side you could theoretically run lbfs over your files but on the swarm side this isn't going to happen

SethTroon The Matrix looks dramatically different on Hulu versus on HBO Max

In practice yes, Museums occasionally clean & restore their paintings which shifts the color balance, grain, texture similar ways.

There's a some great pre/post conservation photo in this article: http://blogs.getty.edu/iris/as-layers-of-old-varnish-are-rem...

SethTroon No evidence that parachutes prevent injuries when jumping out of planes (2003)

They did a follow up randomized experiment where people jumped out of planes (at 2ft) with and without parachutes.

https://healthjournalism.org/blog/2019/02/breakthrough-resea...

SethTroon Implementing secure leaderboards for my game

The start packet should contain something from the server like the level seed, so you can't "start" without it.

SethTroon Show HN: ChessCoach – A neural chess engine that comments on each player's moves

> White is now down a piece, but he has a strong pawn center, and the bishop pair. White is in trouble. The position is still

With no bishops on the board...

SethTroon The Greatest Art Thief (2019)

> more than 99 percent of art thieves are motivated by profit rather than aesthetics. This is why art crimes are typically solved on the back end, when the thieves try to sell the work. But with Breitwieser, law enforcement's chief strategy—poring over art-market data, waiting for the stolen items to reappear—is dead on arrival.

SethTroon Analyzing the legal implications of GitHub Copilot

I find the first argument, that if you're project is in GitHub then they have the right to train in it, weak. Plenty of projects are hosted elsewhere but have been mirrored by random users (e.g. not the copyright holder) to GitHub

SethTroon Boring Problems Need Attention Too

I diffed down to understand why Open Office (1663 consistent with Microsoft Office) differs from Google Docs (1655).

"ago--never": G docs 1 word vs OpenOffice 2 words.

"tiger-lillies--what": G docs 1 word vs OpenOffice 2 words (IDK what this should "really" be)

"Wanting?--Water": GD 3 words vs OO 2 words

In this case the disagreement springs, exclusively, from if the docs engine believes that double hyphens make a compound word (and potentially handling punctuation in the middle of such a compound word)

SethTroon Finding a CPU Design Bug in the Xbox 360 (2018)

Discussion from 2018: https://news.ycombinator.com/item?id=16094925

Side note, I'm happy to call Bruce a friend in the real world where he's an equally talented juggler and runner.

SethTroon Eu_sst: 20% probability for large satellite collision

Leolabs is tracking as well. Their data suggests a slightly lower but >2% chance of collision

https://twitter.com/LeoLabs_Space/status/1380549507390259203

SethTroon Can My Water Cooled Raspberry Pi Cluster Beat My MacBook?

I have a 2 socket E5-2650 v2 for 16 cores which is as fast as a modern Ryzen above (but 50% slower on real world tasks)

0.8 seconds for 10,000

28.16 seconds for 100,000

SethTroon All my servers have an 8 GB empty file on disk

Same idea as this game development legend

https://www.dodgycoder.net/2012/02/coding-tricks-of-game-dev...

> he had put aside those two megabytes of memory early in the development cycle. He knew from experience that it was always impossible to cut content down to memory budgets, and that many projects had come close to failing because of it. So now, as a regular practice, he always put aside a nice block of memory to free up when it's really needed.