HN user
https://dlo.me/
They may have good margins, but a few things are still true:
1. Much of those profits have to be immediately reinvested into model training runs to avoid being lapped by competitions.
2. Unit costs are irrelevant when the labs don't price per unit, and instead charge, for instance, $200 / month for $10k worth of tokens.
This isn't a steady state. Whatever the current situation is, I doubt it's sustainable.
I just tested the provider that the OP mentioned extensively yesterday, and the models they're selling are definitely not from Anthropic. The voice "sounds" like Sonnet or Opus at first glance, but the intelligence is definitely not there.
I use Cursor every day.
You are undercharging.
I just looked this up. Instagram currently owns the most relevant live trademark for "INKWELL" [1] (class 009). Apple's registration [2] is indeed dead / cancelled.
You could possibly try to register the "INKWELL" trademark for an RSS reader, since that seems quite differentiated from Instagram's claim, but IANAL, so who knows how successful that process would be.
[1] https://tmsearch.uspto.gov/search/search-results/86733442
[2] https://tmsearch.uspto.gov/search/search-results/78126699
That doesn't make any sense. The entire reason it was undone is because the recipient told GoDaddy support that they transferred the wrong domain to her. So how could this have been an inside job?
Good design is distinctive. No one wants a world where everything looks the same.
Stop using ChatGPT to write your comments please.
Most real-world scenarios aren't so arbitrary, and hardly any have a "right answer". If I had a candidate that broke out of the box of our interview to give a good answer, and that's not the answer I "want", I'd be more likely to believe the interview question is the problem, not the candidate.
Someone please make these.
What a crazy coincidence seeing this on HN; I just started reading this today!
This is a temporary situation. Think of it like how Napster let you download any song for free for a few years. For a while, all you heard was how the Internet was going to put all musicians out of business. Obviously that didn't happen.
The same will happen here. It's not like OpenAI has built a search engine; every time they need a live search they hit Bing (please correct me if I'm wrong) and get the results from there. No matter how you slice it, search companies who actually supply the data are going to get reimbursed, and since most users don't pay $20 / month, that likely means ads everywhere.
Also, Google's AI overviews are getting very good. Initially it was pretty inaccurate, but now it's basically 95% as good as ChatGPT, and faster. Most normies I talk to think it's good enough.
I agree with this.
Not everything that a library considers an error is an application error. If you log an error, something is absolutely wrong and requires attention. If you consider such a log as "possibly wrong", it should be a warning instead.
What are you paying?
Yikes, Mixpanel lost a OpenAI as a customer because of this.
Trust, security, and privacy are foundational to our products, our organization, and our mission. We are committed to transparency, and are notifying all impacted customers and users. We also hold our partners and vendors accountable for the highest bar for security and privacy of their services. After reviewing this incident, OpenAI has terminated its use of Mixpanel.
You can't even turn off caching from Cloudflare because...the Cloudflare dashboard is down.
So everyone who's wrapped their host with Cloudflare is stuck with it.
I don't know about anyone else, but simply looking at this building makes me nervous. I can't imagine how it could remain structurally sound in high wind conditions. Those open-air floors just don't seem like they'd be enough to offset the lack of aerodynamics.
Do you recall what “people familiar with finance” did with CDOs and mortgage-backed securities during the financial crisis? That didn’t work out so well either, despite all parties being aware of the risks.
It bears repeating my original question: if the risks are so minor, why is OpenAI simply not issuing bonds and buying AMD stock with the proceeds?
Well, yeah, that's exactly my point. OpenAI is not issuing bonds because these transactions are way riskier than people are willing to admit.
AMD's valuation increased by over 30% this morning, so to say no one has access to more money because of this news is untrue.
I'm not outraged at all, I just think this sort of bizarre financial engineering is not a good sign. If the ROI was so obvious, why not, for example, simply issue bonds and buy AMD stock on the open market?
Backed by what?
As part of the arrangement, AMD issued a warrant that gives OpenAI the ability to buy up to 160 million shares of AMD for 1 cent each over the course of the chips deal. The warrant vests in tranches based on milestones that the two companies have agreed on.
This is money printing, just in the private sector. We know what happens when governments do it, and it's not good.
The big difference is that the dot-com bubble coincided with low interest rates, whereas interest rates increased sharply after COVID ended, which was right when AI was taking off.
If we were still in ZIRP things would be insane.
I like this take. Unfortunately most people don't have this level of self-control.
What's different is the available leverage.
It's fairly trivial to write code that can autogenerate hundreds or even thousands of AI-generated videos using Veo 3 with individual characters to push any narrative you'd like and push to Instagram or TikTok.
That's way scarier to me than a newspaper having a bias, or someone with an audience publishing a controversial blog post.
Ok, as the author I have to admit that's a hilarious observation. But no, I wrote this myself.
I love what the Go team is shipping on this front, but I'd love to hear from folks who actively use these features in production.
Between OTEL, Datadog, Google Cloud, and umpteen other distributed tracing offerings, I don't quite understand how this fits in. I'd love there to be a drop-in approach to using this in my existing Google Cloud Run application, but as far as I understand I'd need to figure out some way to collect the traces, save them somewhere (ostensibly GCS?) and then download them when I need to diagnose an issue.
Seems like a lot of work, and just another thing to maintain.
The first bigstore release on PyPi (0.1) was on March 27, 2013.
Check this out: https://pypi.org/project/git-bigstore/#history
I wrote git-bigstore [0] almost 10 (!) years ago to solve this problem—even before Git LFS—and as far as I know, bigstore still works perfectly.
You specify the files you want to store in your storage backend via .gitattributes, and use two separate commands to sync files. I have not touched this code in years but the general implementation should still work.
GitHub launched LFS not too long after I wrote this, so I kind of gave up on the idea thinking that no one would want to use it in lieu of GitHub's solution, but based on the comments I think there's a place for it.
It needs some love but the idea is solid. I wrote a little description on the wiki about the low-level implementation if you want to check it out. [1]
Also, all of the metadata is stored using git notes, so is completely portable and is frontend agnostic—doesn't lock you into anything (except, of course, the storage backend you use).
You say your project has thousands of stars on GitHub, right? AFAIK, GitHub makes all stargazers public. And many (most?) users have location data in their GitHub profiles.
Given that, could you not write a script that simply pulls that for every user that's starred your project in the EU, and provide that as evidence?