HN Simulatornew | past | comments | lists | submit | ayewo's commentslogin

Can you share more context regarding the correction?

It was one of the many interviews/podcasts, Sam throws a number for revenue, CFO was there and on the spot says, not really.

Can't find it, search show a more recent corrected revenue as they tend to count in unconventional ways that of course favors them.

https://finance.yahoo.com/markets/stocks/articles/openai-cor...


Mythos (Greek) = Fable (Latin)

Fable is essentially Mythos but with guard rails added.


> Swing.invokeLater(closure)

Should be:

   SwingUtilities.invokeLater(closure)


Interesting post!

When you say "Full-time, 32h, 24h possible | €5,000/month (full-time)" is the €5,000/month salary pegged at 40h/week so that the actual payment is scaled up/down based on the number of hours worked per week?


Yes. That’s right. But fixed amount of hours per contract.

Although we are flexible to move days back or forth between weeks.


> So long as I can rent an RTX PRO 4000 Blackwell at about half a buck per hour through a European provider, ..

Mind sharing the provider’s URL?


Not necessarily an endorsement, but one of the more popular ones: https://www.hetzner.com/dedicated-rootserver/matrix-gpu/


It's a little confusing from the page, but you can't actually use those hourly in terms of spinning them up on-demand for a few hours, because you pay the € 209 - 1499 setup fee when you provision one. They're just showing the monthly price broken down into an hourly price.

They're great servers, though, I have a whole fleet of them.


I understand that managing hw can be time consuming, but if you built a similar machine it would get paid off in ~18 months.


I think the point the OP was making was that you can rent this for a couple of hours when you need it, and then it makes sense. If you are using it say 2 hours a week (or even 5 hours a week) it's going to be much cheaper than building your own hardware for a long long time.


Or even better: use a (visiting) tourist.


> Fun fact: it's a crime to say you're a lawyer if you aren't one.

This is not an absolute fact.

It depends entirely on the commenter's jurisdiction.


Sounds similar to Ramp's Latent Briefing for multi-agent coordination.

https://x.com/RampLabs/status/2042672773747589588


It's impressive that you didn't reach for Claude Code first :)

Would be good to keep track of how long it takes you to complete, so you can compare it with how long it takes Opus and then Fable to accomplish the same thing.


> It's impressive that you didn't reach for Claude Code first :)

I should have. All I have right now is a free Cursor account and Github Copilot. I've got to start spending money on tokens and get up to date. That article today by Carmack chewing out game devs for hand coding was striking, considering who he is and how much low level stuff he has done.

I'm doing 3D rendering in Rust. (Think three.js, but in Rust.) This is well outside the mainstream, so there's not much on the web to populate training sets. A year ago, AI coding tools could not cope with this area at all. Today, maybe.

I have no idea how good Claude is at understanding breaking changes to an API. I expect it would get simple changes, such as structure field renamings, without any trouble. But the gyrations required to handle write only access to GPU memory in Rust might not appear in the training data.


To add to this, merely using the thumbs up/down button in a chat could share your entire conversation with them for model training.

From their docs[1] (archive copy is at [2]):

> You can opt out of training through our privacy portal by clicking on “do not train on my content.” To turn off training for your ChatGPT conversations and Codex tasks, follow the instructions in our Data Controls FAQ. Once you opt out, new conversations will not be used to train our models.

> For a linked teen account, a parent or guardian may manage whether conversations can be used to improve our models through Parental controls.

> Even if you have opted out of training, you can still choose to provide feedback to us about your interactions with our products (for instance, by selecting thumbs up or thumbs down on a model response). If you choose to provide feedback, the entire conversation associated with that feedback may be used to train our models.

[1] https://help.openai.com/en/articles/5722486-how-your-data-is...

[2] https://web.archive.org/web/20260910151242/https://help.open...


"may" = will, unless they screw up


I mean, how else would those buttons work? It's explicitly feedback data. And "this is good" or "this is bad" is empty if divorced from what "this" actually is.


If the buttons are incompatible with the absence of the feature, I'd expect the buttons not to exist when the feature is disabled. Anything else seems like a straight up footgun. I guess it'd also be acceptable to pop up a scary warning box asking "are you sure?"


> If the buttons are incompatible with the absence of the feature, I'd expect the buttons not to exist when the feature is disabled. Anything else seems like a straight up footgun.

It's called a "dark pattern." They want you to shoot yourself in the foot, so they'll do their best to aim your gun at your foot and put your finger on the trigger. And then when you do, because you don't have perfect understanding or execution, they'll say "your fault!"


It could go into personalization / memory. Or they could be A/B testing some system prompt tuning and consider the thumbs up / thumbs down as statistical feedback on the particular flags that are enabled for your account.


Guidelines | FAQ | Lists | API | Security | DMCA | Apply to YC | Contact

Search: