Hacker News Reader

arthurcolleon Noise cancellation improves turn-taking for AI Voice Agents

Does this work well for multi-user scenarios? I also wanted to as a side effect tag and label people, but not really used to the audio setting. Just found "Speaker Verification with xvector embeddings on Voxceleb" which seems interesting and useful.

arthurcolleon Noise cancellation improves turn-taking for AI Voice Agents

is there some way to do a simple fingerprint or something so that the AI recognizes when it was the one speaking? or do you really just have to WebRTC. I spoke with someone yesterday who told me WebRTC fixed this, so just curious.

I wrote a "simple" (ugly) Acoustic Echo Cancellation module that kind of worked, but wondering if anyone had any solutions to make it work over the WebSockets Realtime API

arthurcolleon Mistral OCR

It's just a FastAPI app with endpoints that I developed and deployed before OpenAI released structured outputs that used a custom grammar to enforce a pydantic-like schema for Chain of Thought rollouts / structured data extraction from unstructured text. I also use it for a video transcription knowledge base generation API

https://arthurcolle--dynamic-schema.modal.run/docs

arthurcolleon Building a Tool Management API

Hi, I wanted to share some of my findings into building a server-side paradigm that is similar to Anthropic's Model Context Protocol.

The full live stream (second video only) is a pretty quick 20 minutes, with the blog derived from the source video transcript using custom tools I've built.

https://www.youtube.com/watch?v=2XVkvfo30Qk&t=977s

arthurcolleon Mistral OCR

You can use structured outputs, or something like my https://arthurcolle--dynamic-schema.modal.run/

to extract real data from unstructured text (like that producted from an LLM) to make benchmarks slightly easier if you have a schema

arthurcolleon NIH Website Down

It's back up

arthurcolleon GPT-4.5

Man I can't believe that fucking game is still alive and kicking. Tell me they're making good progress, sho_hn

arthurcolleon Show HN: Probly – Spreadsheets, Python, and AI in the browser

From the Miami colloquialism, "Supposably" could be good for an advanced stats add-on!;)

arthurcolleon NASA employees advised to delay responses to Musk's "what did you do" email

This is amazing.

arthurcolleon Sierra AI's docs are restricted

applied to these guys, they are like agents layered over openai

I feel like I'm pretty plugged into the agents space, so was a bummer to not hear back from them when I applied to kind of get a feel for what they are doing

wishing them well

arthurcolleon [dead]

This sucks man. I hope you get the money back. Have you tried tweeting at @patrickc (https://x.com/patrickc) just to get more eyes on it

arthurcolleon Notes on OpenAI o3-mini

o1-"pro mode" could just be o3

arthurcolleon Notes on OpenAI o3-mini

thank you!

arthurcolleon Notes on OpenAI o3-mini

its the same thing as:

gpt-3.5 -> gpt-4 (gpt-4-32k premium)

"omni" announced (multimodal fusion, initial promise of gpt-4o, but cost effectively distilled down with additional multimodal aspects)

gpt-4o-mini -> gpt-4o (multimodal, realtime)

gpt-4o + "reasoning" exposed via tools in ChatGPT (you can see it in export formats) -> "o" series

o1 -> o1 premium / o1-mini (equivalent of gpt-4 "god model" becoming basis for lots of other stuff)

o1-pro-mode, o1-premium, o1-mini, somewhere in that is the "o1-2024-12-17" model with not streaming, function calling, and structured outputs and vision

now, distilled o1-pro-mode probably is o3-mini and o3-mini-high-mode (the naming is becoming just as bad as android)

its the repeat, take model, scale it up, run evals, detect innefficiencies, retrain, scale, distill, see what's not working. when you find a good little zone in the efficiency frontier, release it with a cool name

arthurcolleon Hell is overconfident developers writing encryption code

tptacek I don't want to waste your time, but do you have any good recommendations for material that bridges the gap between modern deployed cryptosystems and the current SOTA quantum computing world in enough detail that is useful for an engineer practitioner preparing for next 10 years?

arthurcolleon OpenAI to face Indian digital news firms of Ambani, Adani in copyright battle

Concerning

arthurcolleon Show HN: DeepSeek My User Agent

Oh look, a Hacker News refugee with a Frankenstein 1800x1169 screen—did you split that browser window just to flex your NPC-level multitasking skills in Rockville's tech equivalent of a Starbucks patio? Twelve CPU cores but still running MacOS 10.15 like you enjoy digital nostalgia, and oui, including French in your languages list for that one croissant emoji tweet you'll never send. Chrome 132? At least the RTT is zero, matching your urgency to upgrade anything.

------------

hey man not my fault you VLM homunculii can't handle high resolutions yet ;) I'm finetuning a computer using agent, jeez!

arthurcolleon Saudi Arabia plans $600B in new US investment, trade over four years

There's the Stargate money I guess

arthurcolleon It sure looks like Meta stole a lot of books to build its AI

PDF accessible on the world wide web is fair game

arthurcolleon How Can a Major Bank Like Capital One Be Offline for 24 Hours?

Soon^TM

arthurcolleon How Can a Major Bank Like Capital One Be Offline for 24 Hours?

Bitcoin fixes this

arthurcolleon GPT-4o with scheduled tasks (jawbone) is available in beta

it would explain the bugs if they used the AI to make the datetime implementation though