HN user

sandwichsphinx

1,862 karma
Posts598
Comments61
View on HN
en.wikipedia.org 9mo ago

Dark Fibre

sandwichsphinx
2pts0
www.wsj.com 10mo ago

Is Gen Z Unemployable?

sandwichsphinx
4pts4
www.wsj.com 10mo ago

Bank of Canada's Macklem Warns U.S. Dollar Losing Appeal as Hedge

sandwichsphinx
4pts0
www.wsj.com 10mo ago

Why Marriage Is Increasingly for the Affluent

sandwichsphinx
19pts28
www.wsj.com 10mo ago

Fed Divisions Thrust Powell into Uncharted Territory

sandwichsphinx
4pts0
www.wsj.com 10mo ago

Dating an AI: How Much Is It Happening?

sandwichsphinx
2pts0
www.wsj.com 10mo ago

Americans Are Having Less Sex

sandwichsphinx
22pts5
www.wsj.com 11mo ago

'Work-Life Balance' Will Keep You Mediocre

sandwichsphinx
2pts1
www.wsj.com 11mo ago

Nobody’s buying homes, nobody’s switching jobs, America’s mobility is stalling

sandwichsphinx
156pts156
www.wsj.com 11mo ago

No More Offshore. Startups Look to Spend and Hire in US Due to Trump Tax Change

sandwichsphinx
13pts1
www.wsj.com 1y ago

In America's Return to the Office, Women Are Falling Behind

sandwichsphinx
3pts0
news.ycombinator.com 1y ago

Ask HN: What million dollar questions do you want answers for?

sandwichsphinx
1pts1
patents.google.com 1y ago

Efficient Document Clustering (2008)

sandwichsphinx
2pts0
www.wsj.com 1y ago

Al Gore Helped the U.S. Surpass Europe

sandwichsphinx
4pts0
www.bleepingcomputer.com 1y ago

Google Gemini flaw hijacks email summaries for phishing

sandwichsphinx
2pts0
www.wsj.com 1y ago

A Never-Ending Supply of Drones Has Frozen the Front Lines in Ukraine

sandwichsphinx
8pts2
www.technologyreview.com 1y ago

Ultra-Efficient Solar Power (2013)

sandwichsphinx
3pts0
arxiv.org 1y ago

ZipNN: Lossless Compression for AI Models (2024)

sandwichsphinx
4pts0
www.nature.com 1y ago

Computational understanding of Li-ion batteries (2016)

sandwichsphinx
2pts0
pip.raspberrypi.com 1y ago

Cooling a Raspberry Pi Device [pdf]

sandwichsphinx
2pts0
www.gutenberg.org 1y ago

Preface to a Dictionary of the English Language (1755)

sandwichsphinx
4pts0
www.aeaweb.org 1y ago

The Broad Decline in Health and Human Capital of Americans Born After 1947

sandwichsphinx
3pts6
www.wsj.com 1y ago

As Tesla Stumbles, SpaceX's Finances Go from Strength to Strength

sandwichsphinx
3pts1
pmc.ncbi.nlm.nih.gov 1y ago

Impulse Control in Pigeons (1974)

sandwichsphinx
4pts0
www.justice.gov 1y ago

DOJ Statement of Interest on Suppression of Competition Through Deplatforming

sandwichsphinx
11pts12
ieeexplore.ieee.org 1y ago

Error detecting and error correcting codes (1950)

sandwichsphinx
2pts0
www.mdpi.com 1y ago

The Role of Family in the Life Satisfaction of Young Adults (2024)

sandwichsphinx
6pts0
ieeexplore.ieee.org 1y ago

Standardization of the Ohm as a Unit of Electrical Resistance, 1861–1867 (2019)

sandwichsphinx
4pts0
www.wsj.com 1y ago

Pentagon Lifts Drone Restrictions, Sending Shares of Defense Stocks Higher

sandwichsphinx
3pts0
rift.stacktitan.com 1y ago

The Covert Hardware Implant: Part 1

sandwichsphinx
1pts0

I stopped using AI for code, everything it gives me requires manual checking and editing to work but now I have the overhead of not having the deep understanding I would otherwise have had if I figured out things myself from the start. I also stopped using AI as a sort of replacement for quick google searches in general, but I didn't go back to using google because now I feel those results are too shallow. I been trying to figure things out from primary sources as much as possible and thinking for myself since it seems to me the value going forward is having depth and breadth over the competition that you only get by doing things the very old and slow way.

Sora is here 2 years ago

not dishes but the other day I saw this recent papers for clothes

RoboHanger: Learning Generalizable Robotic Hanger Insertion for Diverse Garments

https://arxiv.org/abs/2412.01083

To overcome the challenge of limited data, we build our own simulator and create 144 synthetic clothing assets to effectively collect high-quality training data.

the strategy is simulation

My dad is getting older and often forgets to bring reusable bags when he goes shopping. As a result, he ends up buying new ones every time, adding to a growing stockpile of unused bags at home. On the other hand, I’ve adapted by buying less and only purchasing what I can carry by hand to avoid needing a bag altogether.

This feels (but I cannot say if it "is" without proper data) like a lose-lose situation: individuals either end up purchasing more bags (which defeats the purpose of reusables) or, like me, change their shopping habits in ways that might not be ideal for businesses themselves who would wish I spend more. I’ve also noticed more people skipping supermarket trips entirely and ordering everything online from places like Walmart or Amazon, which often deliver in boxes or reusable bags anyway.

It seems there isn’t a big-picture solution beyond individuals like me reducing consumption as much as possible—not necessarily out of environmental consciousness, but because it’s now more of a hassle to consume unnecessarily.

cards up to 128TB and up to 985MB/s speeds

this is impressive, seems that SDUC is just a specification so it applies to both full-size and micro SD cards. I don't know of many devices that still support full-size but I hope to one day see a 128TB microSD card. Not sure how they'll figure it out, last I check the biggest size Sandisk sells is 2TB tho the article mentions we'll be seeing 4TB+ cards soon.

What worries me, given my time with the raspberry pi, is reliability. They seem prone to random failure, not that I've personally ever had a card die on me (and it seems that even in failure they lock into read only mode so you don't lose data?), but it seems that there's no standard way to check the health of the card like you can with SSDs using SMART data. This makes me hesitant to trust larger-capacity SD cards for data tho I know of some industrial cards that seem to have some kind of support for viewing card health but I feel safer with small capacities since there's less to lose if something goes wrong but faster speeds are always a plus.

what problems do you have that could potentially be solved by developers?

an offline Grammarly alternative, ideally as a progressive web app that I could add to my home screen without needing to install anything from the App Store. I find the built-in spell check on my iPhone underwhelming, so I’ve disabled autocorrect and predictive text.

One of the most effective motivators I’ve seen is combining math and computer science with game development. Language models on their own can be a bit dry, but if you integrate them as agents in a game world alongside math-based challenges, it could be really engaging. I remember Raspberry Pi’s Educational Minecraft; using a setup like that, where you train characters to have their own personalities, could be an exciting way to teach both

I'm currently subscribed to both ChatGPT Plus and Claude Pro, but if I had to choose only one, I'd stick with ChatGPT. The message limits are more flexible, especially with unlimited mini-usage, which lets me experiment and iterate freely. With Claude, I frequently hit the message limits, which makes me hesitant to use it at all—like saving items in a game only to never use them, ultimately defeating the purpose.

What is something you hate doing that could be solved by software?

I dislike how much hassle 3D scanning is and exporting the model out to share easily online in an open file format

Something you'd be willing to pay for?

No more than a dollar

I noticed the same thing, especially with many of those old authoritative sources that seem to have disappeared or suffered from link rot. It reminds me a lot of the movie Soylent Green, where basic repairs had to be done custom because there were no replacements. Outside of academic journals and printed books, I don’t bother searching for information because it’s just not worth the effort given how hard it is to find reliable sources. The only way I’ve been dealing with this is by relearning my fundamentals. There are only so many ways hardware and software can be designed correctly, so I rely on recurring patterns, motifs, improving my reverse engineering skills, and adopting a clean-room design approach.

An unsolved software challenge I’ve noticed being discussed on HN is the difficulty in creating a competitive alternative to Nvidia’s CUDA and the rest of its toolchain. Intel and AMD making strides, but nothing substantial so far. I understand that this is an incredibly difficult problem, especially considering the significant investment Nvidia has made over time, which can’t be matched overnight

Ever since I read that thread [0] about how AI companions are helping reduce loneliness, I started using ChatGPT and Claude just for casual conversations, similar to how I used to chat with friends on Discord. I initially used them for coding help, but now I often max out my message limits just chit-chatting, especially since the larger models are great for that. The only thing holding me back is the message limits—ChatGPT and Claude often tell me I’ve hit a cooldown, or when the conversation gets too long, I’m prompted to start a new one and lose context. As these models improve, I think I’ll use them more and more for conversation, treating them as characters to talk to rather than just tools like most people do.

0. https://news.ycombinator.com/item?id=41613513

The Scandinavian country already has a minimum age limit of 13 in place. Despite this, more than half of nine-year-olds, 58% of 10-year-olds and 72% of 11-year-olds are on social media, according to research by the Norwegian media authority.

I think my life today would be significantly better if I hadn’t been exposed to unsupervised internet access at around 10 years old with the Nintendo DSi web browser. Who knows how people growing up on TikTok brain-rot streams will turn out. I don’t look back fondly on my early experiences with the internet.