unpopular opinion, yet very true. also, the quality of information is often lower than in written formats / planned radio broadcasts. I still listen to long audio formats on Radio France. but in the end, what differentiates a podcast than any recorded discussion on TV/Youtube?

number of product launches per day (on producthunt.com). all facts are made up by o1. anyone with accurate data? i am trying to figure out how much the ai wave helps makers shipping new products.

accounting for SMEs: the whole issue could be solved if companies would send an email with the invoice. most don't. for some companies, it's even impossible to get a proper invoice without making a support request. Qonto has this neat feature where you forward your invoice

no users, no downtime

gemini 30x less expensive than 4o for the same perf? am I missing something or openai just took a big hit?

simple is very hard.

public release coming soon 👀 - classify with natural language - "Cursor like" interface to interact with your mails (auto suggest attachments, contacts, other emails) - Drive 100% implemented in your email, no more going back and forth

You can deploy open source model via Digital Ocean, and drag and drop files into their knowledge base (they do the embedding for you). not gonna use it because I need to partition embedding between my users, still cool

stop putting filthy things in our croissants. americans, spaniards, this is directed at you—just stop.

Je viens de finir ça : Zest résume et extrait automatiquement l'information de vos .pdf et on met aussi tous vos mails de paiement / facture dans un dossier "facture" ❤️

any advice on dark mode?

RAG is not embedding. RAG is enriching an AI response with related documents. fetch these documents with a simple keyword matching search, it's still RAG. don't start your RAG pipeline with semantic search, start it with simple metadata/keyword matching