I like how the headline says 2/3 of Americans “think” tariffs will lead to higher prices. The other 1/3 aren’t thinking at all.

Tariffs are passed on straight to the consumer. Demand for products from other countries isn’t going to suddenly disappear. Some things can’t be easily made in the USA, and sometimes the items made overseas are a higher quality product.

It’s like rebates - they very rarely benefit the consumer. Having a 30% rebate for solar panels is fine, but it means the prices are at least 30% higher than they should be, which is obvious if you compare US prices to European and Australian prices for identical systems.

Do you think the EU would let the USA in? Even Australia isn’t part of the EU even though they compete in the Eurovision song contest.

You’re talking about direct to consumer from an overseas store, whereas I think the person you’re replying to is talking about the pricing at a US store?

deleted by creator

Source: https://www.reddit.com/r/dataisbeautiful/comments/10wt86h/leonardo_dicaprio_still_refuses_to_date_a_woman/ (sorry for Reddit link)

I like that the 140Wh is the part you decided to question, not the “consumes 1 x 500ml bottle of water”

deleted by creator

Why do people need an app for wallpapers? Just find some nice photos on Flickr, DeviantArt, whatever, save them all to a folder, and configure the OS to change it once per week.

Reminds me of the “free smileys” and “free mouse cursors” apps from the 2000s. I thought we had evolved past that.

Twilio’s TTS isn’t as good as ElevenLabs, and their transcription isn’t as good as AssemblyAI. AssemblyAI can pull key details out of the message (eg people’s names, company names, callback numbers, etc) and IIRC it’s quite a bit cheaper than Twilio’s transcription. AssemblyAI provide $50 free credit to try their service, which should last me a very long time assuming it doesn’t expire.

Plus now I can put “AI engineer” on my resume, lol. A lot of “AI” is all about gluing other people’s work together, and that’s exactly what I’m doing.

or just cut the exposed part of the nails using an oscillating tool? No need to be petty.

Voicemail’s definitely not dead.

I can’t find any voicemail services that work the way I want them to though, so I started building my own using Twilio to handle the incoming phone call + ElevenLabs for text-to-speech + AssemblyAI for speech-to-text + Trestle Smart CNAM API for identifying the caller. I’ll open-source the code once it’s ready.

A lot of apps don’t follow this guidance. There’s plenty of apps whose per-user installations aren’t installed into %LocalAppData%\Programs. I’ve even seen some that install into the roaming AppData directory!

You need to update a bunch of separate things on Linux too, though. For example, apt or dnf, rpms and debs that aren’t in a repo (although Deb-get handles some of those), Flatpak, Snap, fwupd for firmware, plus language-specific things (npm, dotnet, cargo, Python, etc). At least the UIs handle a lot of it now.

installs everything without needing admin rights.

I hate installers that do this because they don’t install the apps in the right place. Apps should be in Program Files.

lists way too much I did not install through Winget

That’s one of the features though. You can update apps via Winget even if you didn’t originally install them via Winget.

I sometimes just give up and use Docker or a Flatpak (depending on if it’s a CLI or GUI app)

Even with kernel updates, you can use something like ksplice or kpatch to update it without rebooting. It’s usually only used on servers though.

WinGet: Am I a joke to you?

I didn’t say I like centralized sites though. Web 2.0 didn’t necessarily bring centralized sites; it brought user contributions and user-to-user communication. Forums and wikis were big for example. It also popularized interoperability with things like RSS and Atom.

I don’t know much about AI models, but that’s still more than other vendors are giving away, right? Especially "Open"AI. A lot of people just care if they can use the model for free.

How useful would the training data be? Training of the largest Llama model was done on a cluster of over 100,000 Nvidia H100s so I’m not sure how many people would want to repeat that.