Vinícius Figueiredo

-

[Vinícius Figueiredo @vinobebs]

https://t.co/xxVxnKdyd3

static pages on the go with 5.6 is really great

[Vinícius Figueiredo @vinobebs]

@coproduto @gustavokov https://t.co/thO9iH9F95

[Vinícius Figueiredo @vinobebs]

@plainyogurt21 @TheLAPurchaser mp3 url access is completely unpaywalled, even from data center ips. Then you just transcribe using your tts of choice.

[Vinícius Figueiredo @vinobebs]

@TheLAPurchaser earnings only, for other conferences you can use seeking alpha. The mp3s are on public unpaywalled urls, let codex explore

[Vinícius Figueiredo @vinobebs]

@TheLAPurchaser check dm

[Vinícius Figueiredo @vinobebs]

@simonw @lmstudio Would love to hear your opinion on this one. Maybe as a new blog post on Agentic Engineering Patterns

https://t.co/2hNt11GMuF

[Andrej Karpathy @karpathy]

I'm not very happy with the code quality and I think agents bloat abstractions, have poor code aesthetics, are very prone to copy pasting code blocks and it's a mess, but at this point I stopped fighting it too hard and just moved on. The agents do not listen to my instructions in the AGENTS.md files. E.g. just as one example, no matter how many times I say something like:

"Every line of code should do exactly one thing and use intermediate variables as a form of documentation"

They will still "multitask" and create complex constructs where one line of code calls 2 functions and then indexes an array with the result. I think in principle I could use hooks or slash commands to clean this up but at some point just a shrug is easier.

Yes I think LLM as a judge for soft rewards is in principle and long term slightly problematic (due to goodharting concerns), but in practice and for now I don't think we've picked the low hanging fruit yet here.

[Vinícius Figueiredo @vinobebs]

@bianco_____ @steipete @OpenAICodexCli but codex --yolo works just fine

[Vinícius Figueiredo @vinobebs]

@rationalexpec Codex no Plus (20$) tem limites muito generosos também. Mesmo em uso profissional, raramente chega perto do limite.

[Vinícius Figueiredo @vinobebs]

@rationalexpec @ArthurMotaSP Honestly, it might just do it on its own if you ask.

[Vinícius Figueiredo @vinobebs]

@rationalexpec Got it. What helps for data is installing the right skills. Maybe if it's a data issue then connecting to FRED as @ArthurMotaSP said might help. If you have a Bloomberg terminal you can use blpapi. I'd look into econometrics skills in the web, you might find something useful.

[Vinícius Figueiredo @vinobebs]

@rationalexpec How have you been using it? Just give it a repo with your research pdfs, or ask it to web search the papers you want it to know about. It should handle pretty well. I'm happy to chat about it too

[Vinícius Figueiredo @vinobebs]

@JordanNanos Will this be Livestreamed?

[Vinícius Figueiredo @vinobebs]

@ai_for_success Missing best STT and TTS categories

[Vinícius Figueiredo @vinobebs]

@HotAisle but they make that disclaimer tho ("nobody uses ATOM yet"). Isn't the dynamo builds industry standard for real world use cases?

[Vinícius Figueiredo @vinobebs]

@MrPowerGamerBR Se importa em compartilhar que tipo de problemas você tem tido dificuldade em usá-lo com eficiência?

[Vinícius Figueiredo @vinobebs]

@jerryjliu0 but 2026 solution doesnt scale well enough for tens of thousands of pdfs

[Vinícius Figueiredo @vinobebs]

@JFPuget @theodormarcu Yeah sure. But after that simple vector search is still the main way of integrating into an AI app? agentic search seems more like a solution to coding base questions than to PDFs chunks retrieval

[Vinícius Figueiredo @vinobebs]

@JFPuget @theodormarcu What are people doing for PDFs nowadays? still vector search?