GaggiX

7 ms·

121.
by GaggiX 8mo ago
Wtok and Wpos should be 26-dim along one of the axis but it shows a 16x16 matrix be default, fc1 instead 16x64 with the default settings (not 16x16).
122.
MathArena: Evaluating LLMs on uncontaminated math questions(matharena.ai)
2 points by GaggiX 8mo ago|0 comments
123.
by GaggiX 8mo ago
I had fun with AI detectors in particular for images, even the best one (Hive in my opinion) was failing miserably with my tests, maybe the one trained on text are better but I find it hard to trust them, in particular if someone know how t
124.
by GaggiX 8mo ago
>These tools are not perfect, but they're right more often than they're wrong. Based on what in particular? The only time I have used them is to have a laugh.
125.
by GaggiX 8mo ago
I have generated my own using your prompt and post it in the previous comment. You haven't posted a z-image one of Atlantis. I'm not at home to try but I have trained lora for z-image (it's a relatively lightweight model), I
126.
by GaggiX 8mo ago
>Because it's too easy obviously. Still the vast majority of models fail at delivery an image that looks real, I want realism for a realistic settings, if it can't do that than what's the point. Of course you can always pa
127.
by GaggiX 8mo ago
>In your example, z-image and Nano Banana Pro look basically equally photorealistic to me I disagree, nano banana pro result is on a completely different league compare to flux.2 and z-image. >But anyway, realistic environments like a
128.
by GaggiX 8mo ago
Are you sure you are not confusing nano banana pro for nano banana, z-image still has a bit of AI look that I do not find with nano banana pro, example for a comparison: https://i.ibb.co/YFtxs4hv/594068364-2510105688951
129.
by GaggiX 8mo ago
For me the only model that can really generate realistic images is nano banana pro (also known as gemini-3-pro-image). Other models are closing the gap, this one is pretty meh in my opinion in realistic images.
130.
by GaggiX 8mo ago
Clang is not written in Rust tho
131.
by GaggiX 8mo ago
>So it's not just whisper v3 under the hood? Why it should be Whisper v3? They even released an open model: https://huggingface.co/mistralai/Voxtral-Mini-4B-Realtime-26...
132.
by GaggiX 8mo ago
Gpt4o mini transcribe is better than whisper, the context is the parent comment.
133.
by GaggiX 8mo ago
Gpt4o mini transcribe is better and actually realtime. Whisper is trained to encode the entire audio (or at least 30s chunks) and then decode it.
134.
by GaggiX 8mo ago
>Having something that can actually parse "yep lets do 4pm tomorrow" from texts and create calendar holds is the kind of thing that's always felt 5 years away. Isn't that just Google Assistant? Now with Gemini it seem
135.
by GaggiX 8mo ago
Even in the worst-case scenario, Anki is already perfect for me as is.
136.
Step 3.5 Flash(static.stepfun.com)
2 points by GaggiX 8mo ago|0 comments
137.
by GaggiX 9mo ago
This is just a Nano Banana wrapper I imagine.
138.
by GaggiX 9mo ago
Well the Taiwan flag doesn't appear on mainland china phones, I think the phone does fall back to the TW letters.
139.
by GaggiX 9mo ago
If people want an AI as a boyfriend at least they should use one that is open source. If you disagree on something you can also train a lora.
140.
by GaggiX 9mo ago
Maybe you can divide the task into verifiable environments like an HTML5 parser environment where an agent is going to build the parser and also check the progress against a test suites (the https://github.com/html5lib/
141.
by GaggiX 9mo ago
It would be really cool if it was able to render Wikipedia correctly, I really like the idea of a browser with minimal dependencies having the ability to navigate most static websites, this one for now compiles instantly and it's incre
142.
by GaggiX 9mo ago
I don't like promoting gore websites but in watchpeopledie and search Iran you can count yourself at least hundreds of body from all the videos, often in a single location, usually the aftermath.
143.
by GaggiX 9mo ago
A friend of mine (from Iran) managed to send me a few messages on January 18th via Telegram (Telegram is very popular in Iran) when the situation was though to be resolving and then nothing, blackout again. And even when the blackout was no
144.
by GaggiX 9mo ago
The vulnerability was found by Claude: >This is true by definition as the QuickJS vulnerability was previously unknown until I found it (or, more correctly: my Opus 4.5 vulnerability discovery agent found it).
145.
by GaggiX 9mo ago
The vulnerability was found by Opus: "This is true by definition as the QuickJS vulnerability was previously unknown until I found it (or, more correctly: my Opus 4.5 vulnerability discovery agent found it)."
146.
by GaggiX 9mo ago
After setting the environment and the verifier you can spawn as many agents as you want until the conditions are met, this is only possible because they run without human assistance, that's the "industrialisation".
147.
by GaggiX 9mo ago
The NSO Group going to spawn 10k Claude Code instances now.
148.
FLUX.2 [Klein]: Towards Interactive Visual Intelligence(bfl.ai)
227 points by GaggiX 9mo ago|58 comments
149.
by GaggiX 9mo ago
I love that everyone is making their own TTS model as they are not as expensive as many other models to train. Also there are plenty of different architecture. Another recent example: https://github.com/supertone-inc/su
150.
by GaggiX 9mo ago
For a comparison Iran reported "200 killed" in the Mahsa Amini protests.

More ›