Hacker Newsnew | past | comments | ask | show | jobs | submit | Pikamander2's favoriteslogin

Maybe we need to spread the word about the optimal stopping rule [1]

[1] https://en.wikipedia.org/wiki/Secretary_problem


I've started using Tiny11 on all my setups, then running DeBloat afterwards. This removes every piece of crapware* and none of them seem to try to reinstall on update (so far, fingers crossed).

https://ntdotdev.wordpress.com/2024/01/08/the-complete-tiny1...

https://github.com/Raphire/Win11Debloat

* be warned, this also removes Edge, so you have to grab your own browser installer of choice


Never thought I'd run an LLM released in 2025, on my phone, in full BF16. With ~80tps on an iPhone 16 pro btw.

When you use PDF.js from Mozilla to render a PDF file in DOM, I think you might actually get something pretty close. For example I suppose each Tj becomes a <span> and each TJ becomes a collection of <span>s. (I'm fairly certain it doesn't use <canvas>.) And I suppose it must be very faithful to the original document to make it work.

I distinctly remember someone making an experiment by asking ChatGPT to write jokes (?) about different groups and calculating the likelihood of it refusing, to produce a ranking. I think it was a medium article, but now I cannot find it anymore. Does anyone have a link?

EDIT: At least here is a paper aiming to predict ChatGPT prompt refusal https://arxiv.org/pdf/2306.03423 with an associated dataset https://github.com/maxwellreuter/chatgpt-refusals

EDIT2: Aha, found it! https://davidrozado.substack.com/p/openaicms An interesting graph is about 3/4 down the page, showing what ChatGPT moderation considers to be hateful.


If this is just a matter of motivation or lack of time I can understand, but if cost is an issue, why not just move to hetzner? A dedicated server there can be had for around 40 bucks, e.g.:

  €42.48 max. per month.  
  CPU: Intel Core i7-7700
  RAM: 64 GB
  Drives: 2 x 4.0 TB Enterprise HDD
Thats with unlimited traffic, but no ddos protection or similar, so I don't know how essential that was at DO. Also you're on physical hw which is always more annoying if you have to call in because of a failing disk, but from my years of experience this is as smooth as it gets; shut down the server, open a ticket requesting replacement ASAP and give the drive's SN, and the server will be up again within 20 minutes. Absolutely acceptable for a side-project that doesn't offer anything mission critical. But I'd really be curious what the bill currently is at DO, and maybe you have some monster HW there that can't be matched here. Genuinely curious.


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

Search: