erjiang

I think there’s evidence found for the origin of “彁” as the result of a poor scan of a newspaper article. Look up “彁 新聞” to find some japanese sources about this.

The API lets you directly choose the model you want. Automatic thinking is a ChatGPT feature since ChatGPT has always been a “GPT wrapper” in that sense.

The list of models to be retired is about ChatGPT. Those models are still in the API.

On the ChatGPT website, there should be an option to enable the legacy models in your user settings.

I don't think the dialog as described in the article is accurate and I can't find a screenshot of Windows NT that says "Windows has been shut down." The only screenshots I can find say, "It is now safe to turn off your computer." The confusion around the Restart button is understandable, but the framing of the story seems to imply that the old dialog led to the later phrase, "It is now safe to turn off your computer."

Looking at the photos, that calculator wheel on the back looks wonderful. Certainly better than the contemporary paper calculator wheel that I got with my Argus.

In my case, I wanted to edit some HDR environment maps. They are a 360deg image of the environment used for 3D work. They need to be floating-point so that arbitrary brightnesses can be captured and used to calculate the lighting in the scene correctly.

Totally, and Racket is cool because you can see it as a framework for creating languages.

But Racket is built on top of Chez Scheme so you can also use that directly if you just want a Scheme.

This is interesting... when I visited recently I realized that Suica on Apple Wallet was more convenient than the physical card. The top reason is that you can use Apple Pay to top up your Suica whenever and wherever you are, without downloading any special app or needing to login to something.

However, one of my credit cards didn't work for that with no clear reason given, but a different one worked almost every time.

In my past life I built and sold dispatch software for microtransit / on-demand rides. (UberPool as a service, more or less.)

What this article doesn’t say:

* Many, many cities have something similar already, but only for riders with disabilities. (“paratransit”) You need to schedule your ride the day before, but they will take you from door to door.

* The cost per ride is quite high: more than $20 per ride, often. This cost is borne by the city, while riders pay little to nothing. In very few cases does it make financial sense - most places aren’t replacing their buses with microtransit.

* The best utilization I’ve seen is on campuses, where there are a fixed set of stops in a small region, and a large population of people who can’t or don’t want to drive (maybe due to limited parking).

It's only a "failure" if you look at your net worth, right? It sounds like you have the experience, knowledge, and even resume of someone much older than 28. If you had instead spent ages 22-28 on a PhD, you'd still have no net worth, but people wouldn't see that as a failure.

Have you tried Factorio multiplayer? I tried it with some other people and it really is like real life.

Somebody put in an ugly hack to deal with scaling rather than rearchitect? Well, I'd like to fix it but it'll mean downtime that'll interrupt player 2's OKR of getting the rocket control units made.

Plus, we didn't allocate enough space, so I need to talk to player 3 who owns the adjacent area (and whose functionality I don't understand). Player 3 doesn't like the idea of moving or redesigning that area, so I'm going to try asking Player 2 or Player 4 to help me rebuild it in an undeveloped area.

Building the new section while the old section is still running has overdrawn on our electricity capacity, which has been oversubscribed for a while but we just kept accumulating tech debt on electrical with ugly hacks. Which brings me back to step 1.

We may not have forgotten that Heroku exists, but we may have forgotten that Heroku was an IDE. At least, I wasn't aware of it. Are there old screenshots of what it looked like?

I built a no-code web-app creator that also has a concept of blocks and I can relate to this. Each paragraph is a different block, and I still haven't quite figured out the best way of handling cross-block interactions like your example of cross-block text selection.

I think I would need to redo the browser-based editor, because even if I solved e.g. cross-block selection across blocks of the same type, it's unclear how it should work across, say, a text block and a video-embed block.

It makes sense to me that gamers as a market are more likely to be brand-loyal and gaming is also more moat-able than hashing.

For miners, they just want whatever does more hashes per second. There's no loyalty there and the hardware is competing solely on hash/power/cost. For gaming, you have proprietary software and APIs like "RTX"-branded stuff, game-specific driver optimizations, etc. where you can better defend yourself against AMD.

So for Nvidia, if they're going to sell X GPUs either way, they'd rather sell to gamers than miners to help preserve their gaming market share which plausibly has more long-term value. Just my guess, at least.

The numbers that they are providing are pretty reasonable and nobody is really disputing that removing the mining requirement and just staking your Ethereum is enormously more efficient. Even if you assume each staker is running 1 server, then 140k computers NOT running at full tilt all the time is still a tiny consumer of electricity compared to the current situation.

A lot of the debate seems to be around whether or not it's as secure or viable, or whether the existing Ethereum miners will try to stage a coup or something.

Disclaimer: I don't hold cryptocurrency and I think proof-of-work cryptocurrencies are a tragic waste.

You may be thinking of proof-of-SPACE cryptocurrency, where the more hard drive space you dedicate, the more rewards you get.

Proof of STAKE is where you dedicate your cryptocurrency and are rewarded based on that. Since owning cryptocurrency doesn't take any physical resource, the only expenditure is keeping your staking node (a computer) online so that it can participate in validation.

When you write types in Python, the interpretation of your type annotations are up to the type checker. Your annotations that you write for e.g. Pyre may not type check in mypy. The end result is that your type annotations are not "Python" type annotations, but "Pyre"/"mypy"/etc. type annotations.

In C, you ultimately care about what the compiler says. And this has also led to dialect-specific C code that works fine in one compiler but doesn't compile or runs incorrectly in another compiler.

Yayagram was submitted nearly a dozen times before it got attention, but when it did get attention it was extremely well received.

Is it just that the ratio of submitted stories to people browsing new is too high, causing a low chance of any particular story to be looked at? If I had novel ideas on improving this I'd write it out, but I think others have already suggested things like randomly mixing new stories into the hot pages.

On a split keyboard with two halves, you can independently move the two halves around. This is great if you don't like the Microsoft-designed angle or want to change the positioning.

Other than that, probably just customizability in terms of switches, keycaps, etc.

I'm currently using a Freestyle 2 but very tempted to get one of the keyboards featured in the linked page.

The main gripe I have with the Freestyle 2 is that I'd like to be able to separate the two keyboard halves much farther. I wish the cord between the two halves wasn't so short and hard to replace.

The other gripes are minor: don't really love the switches, the volume buttons don't work on Linux, and the tilt kit is a little annoying when I move the keyboard around (which I do a lot).

Range without an antenna is around 10-20cm. I was able to hold my RPi next to my ceiling fan and control it, but obviously that’s not very practical.

Theoretically you could exfiltrate data with it, but if you have root access to the RPi already then I don’t see why you wouldn’t just use Wi-Fi.

If anyone has a Raspberry Pi and is interested, check out rpitx[0]. It bit-bangs a GPIO pin to transmit arbitrary radio, turning your Pi into a messy SDR transmitter. By connecting a wire to the GPIO pin you can transmit for tens of meters.

I attached a low-pass filter to the GPIO pin and was able to control my ceiling fan with my Raspberry Pi.

[0] https://github.com/F5OEO/rpitx

Edit: This generates square waves which have harmonics. You will transmit signals on unintended frequencies if you don't use a filter.

I think most if not all of the non-Wacom brands use UC-Logic digitizers. XP-Pen and GAOMON tablets should be pretty similar in price and capabilities to Huion-brand tablets.

I agree that drivers are definitely a big differentiator. Most of the non-Wacom tablets have OK support on Linux thanks to the DIGImend project but it may be a pain to customize things like frame buttons. Win/Mac you are stuck with whatever driver the manufacturer last made for your specific model.

Your question is hard to answer if you don't give some examples of what you consider "expensive".

Drawing tablets are now cheaper and higher quality than they have ever been. A quick of Amazon shows an "Amazon's choice" of a 6.3x3.9" tablet for 34 USD. That's a bit on the small side, but very usable.

If you are looking at flagship professional-level Cintiqs, sure, they are expensive. But that's like looking at an Apple XDR display and complaining that monitors are so expensive. I'll add that display tablets have multiple drawbacks compared with non-display tablets - nothing major for the occasional user though.

Somewhat unrelated, but does anyone know how Inkscape is doing? I remember 10+ years ago it was really promising as a vector companion to GIMP (the "Illustrator" to GIMP's "Photoshop"), but using it again recently it doesn't seem to have gotten enough attention in the last decade or so.

With a UI revamp and stability improvements, it could really be a shining star in FOSS creative software.

Curious to know what issues you have with deploying Python codebases. Out of all of the minor and major gripes I have with Python, deployment is not one of them.

It was fun to read this because I happened to study at Indiana University where most of the work on *Kanren happened.

If anyone is interested in practical applications, Will Byrd (mentioned in the story) went on to work with Matt Might (HN's favorite CS-prof-turned-doctor). One of the products of that was mediKanren[0] which operates over medical research. There was some publicity around mediKanren last year.[1]

[0] https://github.com/webyrd/mediKanren [1] https://www.uab.edu/news/health/item/10703-diagnosis-in-2-12...

Awesome! I've been wanting to add a proper parser to my Scheme interpreter for a long time now, and this might lower the barrier enough to motivate me to actually do it.

It turns out that's because ICU uses a combined Chinese/Japanese dictionary instead of separate dictionaries for each language. Which probably is a little more robust if you misdetect some Chinese text as Japanese and vice-versa.