reaperman

Your comments reminds me of “ Falsehoods programmers believe about addresses”

https://gist.github.com/almereyda/85fa289bfc668777fe3619298b...

Yes, crude oil / gasoline / diesel will break down polyethylene grocery bags.

Eventually it made sense that boat-speed only changes the "apparent wind", as it's only simulating wind-boat interactions, not water-boat interactions.

Amazon Sidewalk can also be used - it automatically finds devices on other networks (like your next-door neighbor) and sends data through their devices in case you don't connect your device to your own network.

https://www.amazon.com/Amazon-Sidewalk/

Products are often cheap enough that the labor costs are too high compared to getting a new unit. It would generally work for appliances that are built into a house or hard to transport because then the relative cost would be offset by the cost of labor to remove and install a replacement.

For example, people generally wouldn’t do this for a TV when they can get a decent replacement for $300 new.

Yes, it's more of a convention where we assume language like "...ignoring the trivial case of 1 being an obvious factor of every integer." It's not interesting or meaningful, so we ignore it for most cases.

There have been many periods in US history where sets of laws were purposefully created that criminalized activities that nearly ~100% of the population engage in. The intent of those isn't to stop those activities, and there's no intent of prosecuting everyone. The intent is to be able to prosecute any individual person or someone close to them, at any arbitrary point in time.

Many of today's lawmakers no longer have that intent, but the system as a whole still keeps running in a manner that allows tools of that nature to be used against targeted individuals and populations.

Slick deals: https://slickdeals.net/f/18256009-milwaukee-m18-redlithium-x...

Normally I dislike these quips for HN; I hate that I love this one.

> Roulette, no way to gain an advantage over the house regardless of what strategy you use.

These days that’s probably true but it has been done: https://www.roulettestar.com/people/joseph-jagger/

> The paper itself is behind a paywall so I can't see it

https://archive.org/details/s41586-025-08759-9

Only place this falls apart is that Accenture / Deloitte are really not sexy. Like being a federal employee at a similar pay scale would actually be more sexy. McKinsey/BCG maybe this makes sense.

TSA (more accurately - CBP, more generally - DHS) contract out the hard engineering to Cellebrite and NSO Group. Those companies develop a dumb-proof box. The CBP agents at the border take the phones, plug them to the box, press a few buttons, and that’s it.

No one in the TSA/CBP/ICE/DHS needs to be smart for this, that’s the job of private engineering firms/contractors.

The statistical methods can detect things orthogonal to performance KPI’s. Automation has “tells” - little things they do differently from what humans would do. Reliably discriminating those signals is a hard problem.

Huge thank you for correcting me. Do you have any good resources I could look at to learn how the previous CoT is included in the input tokens and treated differently?

Very, very little labor is unskilled. In almost any work there is a massive difference in quality and speed between someone who has been doing it for <6 months vs. someone who has been doing it for >3 years.

My theory is that "unskilled labor" was a term of propaganda invented by an earlier generation of business leaders in order to publicly devalue many labor-intensive roles. That generation knew that it was a lie, but the business leaders that followed were taught that "unskilled labor" was axiomatic, and essentially "drank the kool-aid".

The result of this is that the labor pool for many disciplines has been hollowed out because it's no longer financially sustainable for workers to build the skills needed to excel in those roles.

Edit: 'wahnfrieden corrected me. I incorrectly posited that CoT was only included in the context window during the reasoning task and later left out entirely. Edited to remove potential misinformation.

I'm sorry, what? Federal income taxes have not been eliminated. Trump can't even do that; Congress would have to, and they won't.

The federal government currently collects on the order of $2.5 trillion in income taxes. These tariffs would only generate $500 billion of federal revenue. But in reality they'll generate less, due to any amount of second-order effects where less stuff is imported due to higher costs. It's very myopic to only look at the household finances and not have anything to say about a proposed loss of $2 trillion of federal revenue.

With allies like you in this discussion thread, who needs enemies?

I spent the past 30 minutes calculating this all manually using the sources listed below:

The practical takeaway is that the average household will spend $3,488.27 more as a result of the tariffs. Clothes, furniture, toys will be (mean) 26.9% more expensive, electronics will be 24.4% more expensive, tires and jewelry 16.2% and 17.1% more expensive, respectively.

If this is more acceptable to you, please voice your approval.

[0]: https://wits.worldbank.org [1]: https://dataweb.usitc.gov [2]: http://atlas.hks.harvard.edu

I think you might be granting the administration too much benefit of the doubt. They aren't based on "tariffs + trade barriers", they're just based on trade deficit alone.

https://x.com/JamesSurowiecki/status/1907559189234196942

https://archive.ph/kRkRh

Here's a much more detailed analysis of the effects of the executive order: https://chatgpt.com/share/67ee10c6-4690-8006-83d7-8e9b22bccf...

The practical takeaway is that the average household will spend $3,500-4,000 more as a result of the tariffs. Clothes, furniture, toys will be about 30% more expensive, electronics will be about 25% more expensive, tires and jewelry about 15% more expensive, and industrial buyers are going to get hosed when they buy equipment.

There were major carve-outs for the auto industry (yay Michigan) and petrochemical industries (yay Gulf Coast); they’ll still get hosed on equipment but will mostly escape increases in cost of materials. Imported cars/trucks aren’t directly affected either.

------------------------------------

Meta discussion:

I can't do an analysis on the de minimis situation, I don't know of any public datasets that would allow such an analysis, and it's obvious that it will have extremely complex effects (and therefore, any first-order analysis would be very low-quality).

Note on ChatGPT-4.5 "Deep Research": I spot-checked the calculations for HS codes using my own research and the numbers seemed reasonably close to my own. https://atlas.hks.harvard.edu/explore/treemap is an invaluable resource for this kind of analysis. ChatGPT fumbled the bag on HS 30: Pharmaceutical products, by not excluding products listed in Annex II, which overestimated total tariffs by about 6% ($30 billion), but the net effect on households is still in the right ballpark (+/- 20%).

Edit: This isn't one of those simple low-effort posts that say "omg look what ChatGPT said". I'm capable of doing this analysis on my own, and I checked ChatGPT's work for the largest contributing categories of goods. Sometimes we disagreed by 10% or so, but overall the results checked out except for Pharmaceuticals, which I caught and didn't repeat misinformation in my "practical effects" tl;dr. The only way it is significantly wrong is if both myself and ChatGPT missed some large tariff category and assumed it was small - that could raise the real costs above the numbers in the analysis - but I checked all the largest categories that ChatGPT identified as well as all the largest categories from our largest trade partners shown by the Atlas of Economic Complexity, so it's somewhat unlikely that happened.

I purposefully didn't include 2nd and 3rd order effects (e.g. chained CPI) because they are usually relatively small, massively uncertain (my analysis would be worth as much as dog poop on a shoe) and those higher-order effects take too long to manifest. It's not worth predicting costs out 2+ years because the political environment is far too unstable. Just today, the U.S. Senate voted to block all the Canadian tariffs and end the "state-of-emergency" that this executive order is standing on. Front-page diplomacy or private back-room deals could make Trump adjust tariffs up or down on various nations multiple times this year. Who knows what the situation will be 6 months from now, let alone 5 years from now when global supply chains and pricing elasticity might re-normalize. The fact that ChatGPT didn't include these effects is a point in its favor, not a glaring oversight.

I'm not complaining about downvotes but if you are downvoting this, please let me know why so that I can improve. I put a lot of work into this far beyond just typing some crap into ChatGPT and I think these numbers add a lot to the discussion.

You’re probably already on top of it but if your company doesn't operate the datacenter you’ll also want to estimate the carbon cost of cooling in addition to the electricity that the machines consume.

If it’s just the ads that are the issue, paying for YouTube premium would be a similar cost solution, and funnel some money to good creators.

Towards a better Jellyfin solution, I wonder if adding Whisper and an LLM model could transcribe the YT videos and flag any which contains themes that go against parents values.

> Are u talking about incurring technical debt from the generated AI code

I think they're talking about the marginal cost and capital cost of running all those GPU's, as well as the capital costs of training foundational models. With GPT-(n+m) projected to require new nuclear power plants dedicated to GPU usage, there's a question of what the payback time will be and whether the marginal costs will exceed that of a human.

Comparing HN to resdit is explicitly against HN guidelines. Though sometimes I think the only reason it’s never “true” is because Reddit is a moving target. Both HN and reddit get worse over time, so HN never catches up to how bad Reddit is.

Also the bots have not invaded HN, which is a truly massive distinction.

In a slightly different reality, most members might reasonably vote no if it was deemed too costly for the organization's budget, or if the champion of the project was seen as unreliable, or if there were genuine concerns with the chosen contractor. Some of these could change after a short bit -- a sudden funding source appears, a different leader of the project steps up, or they address concerns with the contractor bid.

There appear to be enough loopholes and lack of enforcement that you can still piece a solution together: https://electricalproducts.com.au/incandescent-lamps.html?p=...

Thank you! I’m sorry I overlooked those files. I greatly appreciate this perspective and research.

It’s clear that your company leaves no stone unturned.

If you have a chance to chat with the staff again, I notice their marketing language says "Nearly perfect spectrum matching with daylight" - but they don't publish a Spectral Similarity Index. They only claim a (relatively low) CRI of 90+.

Edit: In other materials, they claim a very high CRI of 95+. Also the advertised wattage is sometimes 400W, other times 500W.

I'm very impressed that Innerscene provides LM-63 data for your artificial skylights. Is there any chance you'd be open to providing Spectral Similarity Index scores for your skylight? I think it would really differentiate your product and show how much your company cares about quality.