Hacker Newsnew | past | comments | ask | show | jobs | submit | phonon's commentslogin

> Acetylcysteine

That's NAC (N-acetylcysteine, C5H9NO3S), mentioned in the article many times.


my bad

4 DIMMS =/= 4 channels

I knew that, but I still thought most desktops with 4 dimm slots supported quad-channel memory. I guess I was wrong.

Because Cerebras handles large models poorly due to latency/bandwidth issues to main memory. See https://openai.com/index/introducing-gpt-5-3-codex-spark/ where its performance is significantly below that of the regular Codex 5.3, and can only handle a 128k text context window. For some use cases its great, but most would rather use a better, slower model.

In the future, they plan hybrid implementations, to be able to serve large models better, e.g.

"AWS. We signed a binding term sheet with Amazon Web Services for AWS to become the first hyperscaler to deploy Cerebras systems in its data centers. Deployment in AWS data centers will require us to meet strict standards for performance, scale, and reliability.Pursuant to the term sheet, we will create a co-designed, disaggregated inference-serving solution that will integrate AWS Trainium3 chips with Cerebras CS-3 systems, connected via high-bandwidth networking, to partition inference workloads across Trainium3 and CS-3. Each system will perform the type of computation at which it most excels. The approach is expected to deliver 5 times more token throughput in the same hardware footprint, at up to 15 times faster speeds compared to leading GPU-based solutions as benchmarked on leading open-source models."


Pat has been on record many times that he was against dividends but the board declared them anyway.

Isn't that just 2x supersampling? If you want "perfect" antialiasing that's the minimum you need, no?


Yes, it is supersampling but historically almost no one runs that way.



It's fascinating to me that judges are elected in Texas, and what's more, run as members of a political party.


So that is not correct workaround at all for AGPL licenses. By moving the MuPDF logic into a Web Worker, you are still providing a "modified version" of the program to the user to interact with. The "separation" via a Web Worker does not change the fact that the user is interacting with a system that includes AGPL-licensed code.


Exactly.

It is a clear AGPL license violation without contacting the MuPDF sales team.

If the OP doesn't want to contact the sales team, it should be open source.


Are you kidding? He had extremely sensitive roles as Devin Nunes' House committee aide from 2017–2019 in the House Permanent Select Committee on Intelligence, National Security Council aide and deputy director of national intelligence (2019–2020), and then Chief of staff to the secretary of defense (2020–2021).


I wonder how much of 2021. Two FBI agents reported that he was the bag man for payments to alter Jan 6 cases.

https://www.nytimes.com/2023/03/02/us/politics/house-weaponi...



SpaceX originally partnered with https://en.wikipedia.org/wiki/Greg_Wyler and https://en.wikipedia.org/wiki/Eutelsat_OneWeb in 2014, then they eventually went their separate ways.

https://x.com/greg_wyler/status/1116101020675977218


https://en.wikipedia.org/wiki/Surrey_Satellite_Technology

You can clearly see the tech had an older history at SpaceX pre acquisition

2004

I believe they also signed up a teledesic exec Larry Williams around the same time


Consider applying for YC's Summer 2026 batch! Applications are open till May 4

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

Search: