Wehrdo

It's largely an artifact of classifier-free guidance used in diffusion models. It makes the image generation more closely follow the prompt but also makes everything look more saturated and extreme.

Sounds like the Lishen batteries might have a safety issue when charged fully. Not sure why else they would limit the charge on them. At least they're offering free replacement or cash payment.

Here is an article that explains more about the outliers that emerge in large transformer models, which is what this modified softmax is being proposed to fix:

https://timdettmers.com/2022/08/17/llm-int8-and-emergent-fea...

The fact that these only emerge in larger models is likely one reason the author hasn't actually tried it.

Very cool. This is exactly how I would expect ChatGPT to be integrated into a search engine - removing the burden of of reading multiple results to find the complete answer to your question. Let ChatGPT read the results and provide a summary tailored to the wording of your question.

I hope developers still default to making their personal repos public after this change. One of the fringe benefits of GitHub is the ability to search across the entire site for uses of obscure, poorly-documented APIs. Defaulting to most repos becoming private would greatly hinder this.

I think another factor for the marathon finishing time discontinuities is the use of pacers --- people who run carrying signs at a pre-defined speed (usually targeting a specific finishing time, e.g. 3:15). Lots of people who have finishing goals will run in a pack with this person so they know they're going fast enough, then will all finish just under the target time.

I'm curious - was that TV edge-lit? I've been on the search for a suitable TV for this project for a while now. Most of the cheap/broken TVs you can find are backlit with fluorescent tubes. I disassembled one and discovered it did not have the lens. And from searching for TVs, it looks like most large screens (beyond about 30") directly backlight with LEDs rather than using LEDs around the sides. Was wondering if the direct-lit LED TVs still had the lens.

The article mentions this, but ads on the lock screen are tied to enabling the auto-changing lock screen image.

For what it's worth, I'm still able to open and use Doxygen HTML docs without running a web server. Looks like it loads both CSS and JS. So I'm not sure what the changenote actually means.

teamblind.com

I used Blendle regularly for a while, and noticed that their painless refund option really mitigated this "decision stress" for me. There were a few articles that just turned out to be garbage click-bait, so at the end of the article I could simply click "refund me", and immediately get refunded for that article. It didn't happen that frequently, but knowing it was an option made the decision to buy seem less final.

I got my Dad set up using GnuCash several years ago when I transitioned him from using an ancient version of Quicken on Windows 3.1. He's in his 70s and uses it to manage his small business finances. Although the dense menus can make it hard to find what he's looking for (and for me to help him), he's been pretty happy using it for just basic accounting and creating monthly/yearly reports. Good to see that it's still being developed.

Beyond the readability aspect, I really dislike that the programmer is required to make sure that the inline datatype matches the type passed to sizeof. If the API ever changes, e.g. to using a long, it has to be changed in two places, and there will be no compiler error if it isn't. Using a separate variable takes care of that automatically.

Are you describing Blendle[0]? Access to paywalled content on a pay-per-article basis.

[0] https://launch.blendle.com/

Why don't projects let you sign away your copyright so your contribution becomes public domain? Is that even possible?

Not sure why you were downvoted; this is my understanding also. Could someone with deeper GPS knowledge chime in?

Looks like a great project!

The webpage mentions the app size, but no mention of RAM usage, which is a bigger concern with Electron apps to me. Can you comment on how it compares?

When I was first learning git, I found an online visualizer like this [0] that really helped make make concrete the ideas of git history being a graph, and what different operations did on that graph.

There was still obviously the issue of memorizing the commands, but at least I knew what the commands were doing on a deeper level.

[0] https://learngitbranching.js.org/

Never had that experience, but I presume to prove it was delivered?

A capacitor shouldn't allow electron flow between its (dielectric separated) plates. The charges should build up on the plates, then exit through the terminals when discharging. If charges cross the dielectric, then you have leakage.

As a soon-to-be college grad, I can attest to this. I applied and interviewed at a number of big tech companies, but explicitly avoided Facebook because of ethical concerns.

On the other hand, I know of only a few classmates whom I believe would not consider them as an employer.

If you want a serious answer to "What would it take to not be terrified of this fact?", I suppose it would be: Evidence indicating that populations are better served by ignoring reasoning and obtaining their beliefs through some other means, be it faith or authority.

In fairness, that opinion is not unprecedented in history.

> 45 per cent of participants explicitly denied, at least once, that anything could change their mind on a particular topic

This is the most troubling data from the study for me. It's natural that people will hold opinions for which they don't have evidence, since we can't be experts on everything (e.g. "I trust my friend, who thinks that"), but the fact that half the population hold opinions that they would not change, given any amount of evidence, is actually somewhat terrifying.

Looks like it's only available to some countries in Asia [0]. I've been using Firefox Focus on my phone (uses Webkit under the hood), since regular Firefox is too slow.

[0]: https://support.mozilla.org/en-US/kb/get-started-firefox-lit...

Glad that Ubuntu works well for the author, but I've used Ubuntu on and off for a good 5+ years, and could fill a book with small things that just don't work quite right, and cumulatively makes it feel like I'm fighting the computer instead of working with it.

As a brief taste:

* No trackpad swipe gestures (plus many more subtle trackpad issues)

* Pre-Wayland the only way to attach a second screen at a different DPI was with xrandr that caused random flickering

* Accidentally bricked it once because I restarted while updates were happening in the background (no indication they were, and no prompt to let me know before restarting)

I could go on for awhile...

> charging at < 20 MPH.

I've never heard charging speeds referenced in this way, but it's a really clever way to take into account the charging rate and vehicle efficiency, while being very intuitive! Is this a common "unit" in the EV space?

Personally, I feel like if I calculate an exact 15% to the penny, I'm being petty. So I'll typically calculate the exact tip, then round up.

I used Oboe (and its predecessor, Howie) to make a self-balancing robot on Android. We were able to use the headphone jack (RIP) as a low-latency output to send serial data to an Arduino[0].

Android Studio has also finally improved NDK support enough that writing large chunks of your app in C++ isn't painful anymore.

[0]: https://davidawehr.com/blog/audioserial/

That was my first thought, too. It seems quite plausible that the exact skills suited to doing well in sales (competitiveness, focus on short-term gains, etc.) may be the exact opposite of skills suited for managing.

The article's generalization from a specific employee type to all employees is unfounded.

Each journal is different in what they allow to be shared. There are primarily three "versions" of the paper:

1. Pre-review, pre-print: i.e. the journal had no part in the paper up to this point

2. Post-print: Changes have been made to satisfy the reviewers' comments. Contains all the final content that will be published

3. Publisher's version: The publisher has done copy-editing (layout changes), and probably adjusted it to fit whatever style their journal is.

You can check the rules for most publications here: http://www.sherpa.ac.uk/romeo/index.php