HNHacker News
TopNewBestAskShowJobs

sdan

4,299 karma · joined April 23, 2018

https://sdan.io
submissionscomments
sdan··on Contrastive Language Models
I tried running this on a H100 and got 190ms compared to Jev's 170ms. Maybe I set it up wrong?
sdan··on Show HN: Discover Wikipedia articles popular on Hacker News
nice! kind of similar but re-made a scrollable wikipedia a few weeks ago: https://quack.sdan.io/
sdan··on Show HN: Scroll Wikipedia like TikTok
sorry not the UI but specifically the content. every new post on the feeds that are not "Following" are created in realtime. just a demo showing whats possible but I could just generate them before hand in bulk too.
sdan··on Show HN: Scroll Wikipedia like TikTok
thanks! i spent last weekend building out the engine to create content but agree UI needs more polish. this is all great feedback will try getting out v2 of quickquack rebranded to wikwok :)
sdan··on Show HN: Scroll Wikipedia like TikTok
Wikipedia articles are randomly sampled, theres no algorithm per-se yet but hoping to collaborate with others to see that through
sdan··on Show HN: Scroll Wikipedia like TikTok
how could i make it more useful? different formats? would you want to keep you engaged for longer? entertained for longer?
sdan··on Show HN: Scroll Wikipedia like TikTok
great suggestion ! i might change it to wikwok, so good
sdan··on Show HN: Scroll Wikipedia like TikTok
Note there's a rate limit on generated content for the time being, why you're unable to see any new content on the rest of the feeds. LLMs are expensive!
sdan··on Show HN: Scroll Wikipedia like TikTok
A few months ago I made a (theoretically) infinitely learning geo-guessing model that updated the policy with each user guess: https://geospot.sdan.io/

Hoping to implement a simple RL loop here and optimize whats generated by the LLM to create the perfect slop machine :)

sdan··on Show HN: Scroll Wikipedia like TikTok
https://marketplace.visualstudio.com/items?itemName=SuryaDan...

for clarification :)

sdan··on Show HN: Scroll Wikipedia like TikTok
thanks
sdan··on Show HN: Scroll Wikipedia like TikTok
nice did you make this just for the comment? haha
sdan··on Show HN: Scroll Wikipedia like TikTok
Yes I have base css/js that I inject on top of whatever codegen gemini 3 comes back with -- It runs via ai-sdk so the specific function is streamObject which is prompted to generate inner HTML elements
sdan··on Show HN: Scroll Wikipedia like TikTok
Thanks! Half of the work here was getting gemini to generate reasonable canvas animations within the window. I'm still experimenting which "style" I should keep/add
sdan··on Show HN: Scroll Wikipedia like TikTok
Huge- good to hear will see if I can add a way to type in what you want or to go dig further into just certain hashtags.
sdan··on Show HN: Scroll Wikipedia like TikTok
Yes just fixed timer, and using browser TTS nothing fancy here on purpose - when I did some research on tiktok videos generally simpler/worse quality seemed to be better XD
sdan··on Show HN: I embedded 10M StreetView images
To prevent overfitting
sdan··on Show HN: I embedded 10M StreetView images
Getting there! training a new model, on thinking machines' tinker api which uses a considerably larger and different architecture.

In the meanwhile I tried at least 100+ variations of trying to train this model to be SOTA most of which led me down getting better data which I believe I do, just needs more tuning still.

sdan··on Launch HN: Skope (YC S25) – Outcome-based pricing for software products
this is great, outcome based pricing is definitely a competitive advantage and more aligns you with your customer!
sdan··on A look at Cloudflare's AI-coded OAuth library
> Another hint that this is not written by people familiar with OAuth is that they have implemented Basic auth support incorrectly.

so tldr most of the issue the author has is against the person who made the library is the design not the implementation?

sdan··on Run DeepSeek R1 in the browser using WebGPU
thats great -- what TPS were you getting? github: https://github.com/sdan/r1-web
sdan··on Suno has raised $125M to build a future where anyone can make music
the filter between slop and good music increases
sdan··on Exact binary vector search for RAG in 100 lines of Julia
cool to binary vector search -- I made something similar with numpy in python: https://github.com/sdan/vlite
sdan··on Show HN: Fastest vector database made in NumPy
will do now
sdan··on X.org Server Clears Out Remnants for Supporting Old Compilers
Unrelated but that looks almost like Twitter took this logo and flipped it horizontally
sdan··on Netflix prepares to send its final red envelope
Possible they mixed up Los Gatos and Los Altos? Los Altos is much wealthier.
sdan··on Bay Area 2023 Maker Faire
im so glad maker faire is coming back. i got my first job there in high school and pushed me into getting into robotics and ee early on.
sdan··on Show HN: Serverless GPT Function Calling
yes! the js functions are run serverlessly on val.town so you can make your own that do little functions and drop them in as GPT functions immediately
sdan··on Show HN: Serverless GPT Function Calling
sort of — you can create js functions and create a simple API to use those functions via openai function calling
sdan··on PdfGptIndexer: Indexing and searching PDF text data using GPT-2 and FAISS
I am working on a simple vector db just with numpy: https://github.com/sdan/vlite

I think milvus, quickwit, and pinecone are geared more towards enterprise and are hard to use.

Page 1 of 24Next →