siwakotisaurav

5 ms·

1.
by siwakotisaurav 2mo ago
Thanks for that, looks really good. I can see why they were constantly pushing back fable going out of the max sub with these benchmarks
2.
by siwakotisaurav 3mo ago
Yea this one specifically is mostly fable, remotion skills and allowing it to use Gemini 3.1 pro to understand the video The Claude one was a 3-prompt to get it to recreate the original “Agent view” announcement video. But wanted to challen
3.
by siwakotisaurav 3mo ago
AI videos as in remotion based videos look much better imo since it can code much better than it can prompt for videos with a coherent narrative https://youtu.be/uDAeAuYyl0E (parody of Claude announcements) https:/&#x
4.
by siwakotisaurav 3mo ago
Now the goalpost is “all” of code? I don’t think anything I say is likely to change your mind. You might need to get through this valley of depression by yourself in a few years and join the rest of devs using ai
5.
by siwakotisaurav 3mo ago
If you take them at their word, the people working on the current browsers, compilers and databases are all using it Google self reports 70% ai usage in code, bun was fully ai rewritten to be rust
6.
by siwakotisaurav 4mo ago
Was about to split my $200 max plan into $100 Claude and $100 codex, let’s see if I still need to
7.
by siwakotisaurav 5mo ago
Yea and also with AI I treat throwaway side projects as a way to develop my stack more so that for the next project I can just point Claude to it and say use this as reference instead of having to really work hard on thinking about architec
8.
by siwakotisaurav 6mo ago
Yea some opinions are just widely shared I guess hn shouldn’t be this political mr … politician
9.
by siwakotisaurav 8mo ago
Nice work. Tried remotion when the tweet came out but the results I got were pretty mediocre, some of the results on your homepage are pretty good though.
10.
Voice Coding on iPad with Claude(youtube.com)
2 points by siwakotisaurav 9mo ago|1 comments
11.
by siwakotisaurav 9mo ago
Saw a recent post about coding on the go and wanted to share a somewhat recent video about doing it on an iPad. Is a big help for me cause of having to walk a lot due to sciatica
12.
by siwakotisaurav 1y ago
Antitrust in US seems more and more of a joke for every new one. First few literally broke apart big companies, then we have Microsoft with “just don’t make IE default browser” and now basically nothing for google, forget selling off chrome
13.
by siwakotisaurav 1y ago
Usually don’t believe the benchmarks but first in web dev arena specifically is crazy. That one has been Claude for so long, which tracks in my experience
14.
by siwakotisaurav 2y ago
You’re using slightly improved qwen and llama not r1. R1 only has 600b model
15.
by siwakotisaurav 2y ago
Groq doesn’t have r1, only a llama 70b distilled with r1 outputs. Kinda crazy how they just advertise it as actual r1
16.
by siwakotisaurav 2y ago
Would like to see how much of this is also the case with r1 zero which I’ve heard is less censored than r1 itself, ie how many questions are still censored R1 has a lot of the censorship baked in the model itself
17.
by siwakotisaurav 2y ago
Models other than the 600b one are not R1. It’s crazy how many people are conflating distilled qwen and llama 1 to 70b models as r1 when saying they’re hosting them locally The point does stand if you’re talking about using deepseek r1 zero
18.
by siwakotisaurav 2y ago
None of the models other than the 600b one are R1. They’re just prev gen models like llama or qwen trained on r1 output making them slightly better
19.
by siwakotisaurav 2y ago
Dude you still have lorem ipsum on your site
20.
by siwakotisaurav 2y ago
RealvisXL is much better, probably the best model for photorealism. Example with same prompt: https://gencdn.aieasypic.com/original/ed7c2742-4a76-4e22-9bb... It’s currently the best open weights model for prompt adhere
21.
by siwakotisaurav 3y ago
Does this also apply to online transactions or are they free to keep boiling the frog?
22.
by siwakotisaurav 3y ago
https://x.com/dan_abramov2/status/1763938392654405985?s=46 I remember seeing this on my timeline earlier, he seems fine with it
23.
by siwakotisaurav 3y ago
HN hug of death
24.
by siwakotisaurav 3y ago
They’re probably in the business of being the hardware provider. Best thing would be if Microsoft buys a lot of their chips and that way chatgpt is actually sped up. It’s basically model independent
25.
by siwakotisaurav 3y ago
They’re competing with the lowest cost competitors for mistral atm, which afaik is currently deepinfra at the same pricing
26.
by siwakotisaurav 3y ago
The biggest one missing is stable diffusion.
27.
by siwakotisaurav 3y ago
Getting this on my second fb account, good to know it’s not just me
28.
by siwakotisaurav 3y ago
FYI there’s also this fork for faster inference: https://github.com/152334H/tortoise-tts-fast
29.
by siwakotisaurav 3y ago
Have you tried tortoisetts? I believe eleven labs basically forked that and made improvements on voice quality and speed there
30.
by siwakotisaurav 3y ago
What did I just read? Taxes that a computer program pays?

More ›