coder543

12 ms·

181.
by coder543 1y ago
Historically, the term you're looking for might have been "patronage". Wealthy individuals supported artists, scientists, or explorers not purely for financial return, but because they believed in the person, the cause, value
182.
by coder543 1y ago
ChatGPT is the number one free iPhone app on the US App Store, and I'm pretty sure it has been the number one app for a long time. I googled to see if I could find an App Store ranking chart over time... this one[0] shows that it has b
183.
by coder543 1y ago
Another thing you’re running into is the context window. Ollama sets a low context window by default, like 4096 tokens IIRC. The reasoning process can easily take more than that, at which point it is forgetting most of its reasoning and any
184.
by coder543 1y ago
Huge demand for copper hasn’t brought its price down to the price of stainless steel, has it? Most definitely not, so it seems like Econ 101 was incomplete. Not all goods are perfectly elastic. Inelastic goods do not get cheaper with more d
185.
by coder543 1y ago
Nope... this stuff is 96.5% copper, and copper is ~3x as expensive as stainless steel. Even if tantalum and lithium were free, it would be substantially more expensive. Tantalum is not free, though. It's a very expensive material at ab
186.
by coder543 1y ago
"When you have thinking turned on, all output tokens (including thoughts) are charged at the $3.50 / 1M rate"[0] [0]: https://x.com/OfficialLoganK/status/1912981986085323231
187.
by coder543 1y ago
I downloaded a sample RAF file from the internet. It was 16 megapixels, and 34MB in size. I used Adobe DNG Converter, and it created a file that was 25MB in size. It was actually smaller. Claiming that DNG takes up 4x space doesn't ali
188.
by coder543 2y ago
No... OpenAI said it was "rolling out". Not that it was "already rolled out to all users and all servers". Some people have access already, some people don't. Even people who have access don't have it consisten
189.
by coder543 2y ago
On the web version, click on the image to make it larger. In the upper right corner, there is an (i) icon, which you can click to reveal the DALL-E prompt that GPT-4o generated.
190.
by coder543 2y ago
Is it drawing the image from top to bottom very slowly over the course of at least 30 seconds? If not, then you're using DALL-E, not 4o image generation.
191.
by coder543 2y ago
If it tried and failed repeatedly, then it was prompting DALL-E, looking at the results, then prompting DALL-E again, not doing direct image generation.
192.
by coder543 2y ago
To be clear, that is DALL-E, not 4o image generation. (You can see the prompt that 4o generated to give to DALL-E.)
193.
by coder543 2y ago
"B" just means "billion". A 7B model has 7 billion parameters. Most models are trained in fp16, so each parameter takes two bytes at full precision. Therefore, 7B = 14GB of memory. You can easily quantize models to 8 b
194.
by coder543 2y ago
Also true. Tap on any text field. In the menu where the "paste" option lives, there is also a Scan Text option. I've used that for a number of things over the years.
195.
by coder543 2y ago
On iPhone, WiFi QR codes work just fine. You just open the camera app, and point the phone at the QR code. They're automatically detected and scanned, the same as any other normal QR code. (No, you can't open the camera app during
196.
by coder543 2y ago
DeepSeek-R1 only has 37B active parameters. A back of the napkin calculation: 819GB/s / 37GB/tok = 22 tokens/sec. Realistically, you’ll have to run quantized to fit inside of the 512GB limit, so it could be more like 22G
197.
by coder543 2y ago
https://en.wikipedia.org/wiki/String_interning It's an established term in software. I would speculate that the word "intern" is short for "internalizing" here.
198.
by coder543 2y ago
Location: Birmingham, AL Remote: Yes Willing to relocate: Yes Technologies: Go, Rust, TypeScript, React, Postgres, Kafka, AWS, GCP, etc. Résumé/CV: https://drive.google.com/file/d/1VNC272B3n7ZEfppM
199.
by coder543 2y ago
> I'm a bit surprised there was no mention of semiconductors (mostly capacitors?) going bad on 1990's hard drives. I don’t think any kind of capacitor is a semiconductor.
200.
by coder543 2y ago
Yes. Hyperspace is finding the identical files and then replacing all but one copy with a reflink copy using the filesystem's reflink functionality. When you asked about the filesystem, I assumed you were asking about which filesystem
201.
by coder543 2y ago
I think the term to search for is reflink. Btrfs is one example: https://btrfs.readthedocs.io/en/latest/Reflink.html Like with Hyperspace, you would need to use a tool that can identify which files are duplicates,
202.
by coder543 2y ago
https://support.apple.com/en-us/102651 I think this is the better link. Advanced Data Protection is end to end encrypted, without the key being backed up to Apple’s servers.
203.
by coder543 2y ago
BunnyCDN charges significantly less for data that they serve, for example. I didn’t say all CDNs are cheaper. Some CDNs see an opportunity to charge a premium, and they do! Fastly sees themselves as far more than just a CDN. They call the
204.
by coder543 2y ago
The bigger the service, the more financial incentive they have to be smart and not pay absurd prices for things, since they can give themselves higher profit margins by controlling their costs.
205.
by coder543 2y ago
Azure and the other mega clouds seem to enjoy massive profit margins on bandwidth… why would they willingly drop those prices when they can get away with high prices? If bandwidth costs are important, there are plenty of options that will l
206.
by coder543 2y ago
How much do you think bandwidth costs? I’m trying to understand what 10x cheaper would look like to you, as an actual $/TB number. I think a lot of people have misconceptions about how much bandwidth really costs.
207.
by coder543 2y ago
That wasn’t the question… they asked if any multimodal models had been reasoning trained. o1 fits that criteria precisely, and it can reason about the image input. They didn’t ask about a model that can create images while thinking. That’s
208.
by coder543 2y ago
https://platform.openai.com/docs/models/#o1 > The latest o1 model supports both text and image inputs
209.
by coder543 2y ago
Or, if you really hate duplication, you could optionally do something like this, where you extract the common error handling into a closure: handleError := func(origErr error, context string) error { c, err := h() if e
210.
by coder543 2y ago
I recommend ignoring the other reply you just got. They are clearly building a bad faith argument to try to make Go look terrible while claiming to sing its praises. That is not at all how that would look in Go. The point being made was tha

More ›