flakiness
flakiness on Hacker News
Totally off topic but still made the day.
They seem to go with Android-style alphabetic order code names instead of pointed numbers (4.1, 4.2...) and drop "B" for some reason (which is creating this confusion.)
What would be the computing cost to go through the book (or the course)? Do you need Cloud GPUs, or can this be run locally, if so how powerful your computer has to be?
I've been wondering about getting a PC to poke around LLM stuff (not a practical use, but more for learning purposes) but I don't have any intuition about how much is enough and whether it is worth it to me.
That's why I mentioned "non-model" quality. Although it's getting harder to tell harness quality from model quality, my impression was Muse is just lacking to deliver their ambition.
I worry how buggy their app is. I love their model but Meta/Muse has the edge on non-model implementation. Hope this turn out to be false alarm.
One explanation I heard was that it is to avoid slow government certification-like process. New version == another run of certification.
Although I'm totally on board of the high-level points,
> This remains yet another shameful example of large AI companies appropriating the commons and presenting others’ work as their own.
If you work at Meta which is, you know, one of large (aspiring) AI companies, a bit of self awareness might help boost your credibility a bit more.
That sounds bad. I just haven't encountered it personally (or didn't notice).
Despite of the shortcomings, I'm still sticking to Ubuntu. It's boring, and occasionally break things. But at least it's not hostile.
If you use the words like "still" and "at least", there is little enthusiasm. But that's how a long-term relationship looks like I guess.
Can anyone explain the difference between sampling vs mononoke?
Edit: I was too lazy. From the README: "Mononoke is the server built for the Sapling Source Control System"
They want you to vibecode this using Claude.
Android phones don't need (and often don't have) upstreamed "drivers". They use qualcomm's closed-source code. Also, Android user land driver (HAL) isn't useful for non-Android Linux distros.
If the upstream is properly done (that is a big if), this can be a go-to choice for the Linux laptop because it has much smaller config variations compared to Intel-based laptops: Everything from wifi (and even modem?), to audio, to camera ISP is included in the single SoC. If it works it works.
Posted by
flakiness on Friday September 18, 2026 @08:03PM
from the machinegods.fm dept.
Yeah I had a slight hope that they'd be a one of the first real Rapidus customers. but apparently not (for this one at least).
mind sharing a link or two on the native code generation bit? I'm far behind from the scene but still curious.
I'm not sure it's a bad thing.
If you spend a long time with C++ code base you'll be able to decipher the otherwise-unreadable compiler errors pretty quickly, and I'd consider it a skill.
https://duckdb.org/roadmapCurrent Google cache of this page states:
> "The DuckDB project is governed by the non-profit DuckDB Foundation . The Foundation and DuckLabs are not funded by external investors (e.g., venture capital)."
But "(Last updated: Aug 2026)" and I was not able to find this text anymore.
I wonder if they took vc money after all. I guess not, but still, it seem like to be difficult to live as an independent open source company. Getting a big co as a parent/sponsor is probably the next best thing.
If we program NPUs on Qualcomm-based chips using Mojo, the PC AI programming as a hobby would be getting much more interesting. I wish they don't limit this to the server chips.
Now Baseten's pricing is cheaper than the official one? Probably won't last, but still interesting.
https://www.baseten.co/pricing/
If anyone has tried Baseten versions of these Chinese frontier models, let me know what you found.
Tell me more about the ideas behind Cordis the plugin system. The paper is a bit too mathy to consume and I think it deserves a more accessible post or something.
The post in question:
https://openjdk.org/legal/aiIt is 'OpenJDK Interim Policy on Generative AI' and their lawyers are writing the final version, according to this page.
It sounds like a sensible action, given past scars around Java and copyright, plus this is from a big old corp.
That said I personally don't expect that final proposal will end up any better.
I felt the same before then realized that what I missed was my youth. Plenty of time to kill. No real privacy need. Being a night owl with little hangover, etc. etc.
Not saying that's you. Just cannot help going down the memory lane.
Fair point. Their page doesn't list any investors.