thegeomaster

8 ms·

1.
Cerebras-powered, instant design canvas using Qwen3.8-27B [video](youtube.com)
1 points by thegeomaster 1d ago|0 comments
2.
by thegeomaster 24d ago
For what it's worth, before I hurl such an accusation I always check the post in Pangram ( https://pangram.com ). It always detects the text at 90+% AI generated. Notably, Pangram is very conservative, and it's not diffi
3.
by thegeomaster 1mo ago
Probably has something to do with the fact that the blog post is not human-written.
4.
by thegeomaster 2mo ago
Pangram is a very reliable tool and it does clock the text as AI generated.
5.
by thegeomaster 2mo ago
If even Earendil is publishing AI generated blog posts...
6.
by thegeomaster 4mo ago
Looks great, even has Linux support!
7.
by thegeomaster 4mo ago
> `from opentele while import trace` FYI, this to me points to an inference bug, bad sampling, or a non-native quant. OpenRouter is known to route requests to absolutely terrible, borked implementations. A model like DeepSeek V4 Flash sh
8.
by thegeomaster 5mo ago
Alexandr Wang on Twitter [0] mentioned open source plans: "this is step one. bigger models are already in development with infrastructure scaling to match. private api preview open to select partners today, with plans to open-source fu
9.
by thegeomaster 5mo ago
What's the "attention window"? Are you alleging these frontier models use something like SWA? Seems highly unlikely.
10.
by thegeomaster 6mo ago
And it seems they've decided to go closed-source for their largest, best models.
11.
by thegeomaster 6mo ago
Tried on a few of our production prompts and got comparable speeds to what we normally get with Fireworks Serverless (Kimi K2.5), but at a better price. Rooting for you!
12.
by thegeomaster 6mo ago
Thank you so much for the kind words and for the feedback! 1. Duly noted on USDC and other payment options - I have to see how easy this is do to as we're using stripe. 2. Teams and orgs are very high on the priority list and we hope t
13.
by thegeomaster 7mo ago
Thanks for the feedback! I'm trying to fix that. The trouble is actually that changing the src of an iframe on a page pushes an entry into the history implicitly. Since we use iframes to display the contents of designs, and they can up
14.
Show HN: Mowgli – Figma for the agent era, with Claude Code and design export(mowgli.ai)
6 points by thegeomaster 7mo ago|4 comments
15.
by thegeomaster 7mo ago
Not the parent commenter, but in my testing, all recent Claudes (4.5 onward) and the Gemini 3 series have been pretty much flawless in custom tool call formats.
16.
We're Still Underestimating What AI Means(tinyclouds.org)
3 points by thegeomaster 8mo ago|0 comments
17.
by thegeomaster 8mo ago
I actually ran this one. It measures some 700k lines of code, and seems to contain things like a full VBA implementation, complex currency and date parsing, etc. But the UI is extremely basic, doesn't seem to expose any of this advance
18.
by thegeomaster 10mo ago
Article talks about all of this and references DeepSeek R1 paper[0], section 4.2 (first bullet point on PRM) on why this is much trickier to do than it appears. [0]: https://arxiv.org/abs/2501.12948
19.
by thegeomaster 10mo ago
You could think of supervised learning as learning against a known ground truth, which pretraining certainly is.
20.
by thegeomaster 11mo ago
It's interesting to also compare this to getting a bare metal instance and provisioning microVMs on it using Firecracker. (Obviously something you shouldn't roll yourself in most cases.) You can get a bare metal AX162 from Hetzner
21.
by thegeomaster 1y ago
You didn't include the amortized cost of a Blackwell GPU, which is an order of magnitude larger expense than electricity.
22.
by thegeomaster 1y ago
Warning: LLM-generated article, terribly difficult to follow and full of irrelevant details.
23.
by thegeomaster 1y ago
Well this was a trip down the memory lane. I built a small game on Irrlicht at the time and I remember these discussions also. Irrlicht had its editor (irrEdit), a sound system (irrKlang), and some basic collision detection and FPS controll
24.
by thegeomaster 1y ago
What an astounding achievement. In 6 years, this person has written not only a very well-designed microkernel, but a build system, UEFI bootloader, graphical shell, UI framework, and a browser engine. The story of 10x developers among us is
25.
by thegeomaster 1y ago
Common sense: - The compute requirements would be massive compared to the rest of the industry - Not a single large open source lab has trained anything over 32B dense in the recent past - There is considerable crosstalk between researchers
26.
by thegeomaster 1y ago
tok/s cannot in any way be used to estimate parameters. It's a tradeoff made at inference time. You can adjust your batch size to serve 1 user at a huge tok/s or many users at a slow tok/s.
27.
by thegeomaster 1y ago
There's no way Sonnet 4 or Opus 4 are dense models.
28.
by thegeomaster 1y ago
Are you saying that you think Sonnet 4 has 100B-200B _active_ params? And that Opus has 2T active? What data are you basing these outlandish assumptions on?
29.
by thegeomaster 1y ago
Seems heavily vibe coded, down to the Claude-generated README and a lot of the LLM prompts themselves (which I have found works very poorly compared to human-written prompts). While none of this is necessarily bad, it requires a higher burd
30.
by thegeomaster 1y ago
Thanks for sharing this! It's difficult to find good examples of useful codebases where coding agents have done most of the work. I'm always actively looking at how I can push these agents to do more for me and it's very inst

More ›