Hacker Newsnew | past | comments | ask | show | jobs | submit | prophesi's commentslogin

I was also frustrated by the short context length for summarizing and found you can increase it in about:config via:

browser.ml.chat.maxLength


I googled* it and dod the same. But that merely moves the problem forward. Now the API returns errors that it's too large.

A smarter client would use a tool chain in which the first step is a model that's good at taking large contexts/data and extracting actual content from it. Many sites have a very low S/N ratio (readable content / dom).

Then pass that content, eg markdown, along to a model that's optimized at getting relevant parts out of content for the task at hand.

And only then onto the generic model to "do stuff" with it.

But many clients, including afaiks the Firefox one, just send the entire dom or html along to a generic model.

*Well DuckDuckGo


I'm reminded of the infamous HN Dropbox comment.


Is it FOSS? Tried visiting the repo link at the bottom of the page and there's only a readme, a config file, and the license.

[0] https://github.com/tinkersec/writerdeckOS


I also resorted to using archive.is when I first visited the page so that I wouldn't need to agree to their data collection for personalization.

You only need to pass the app review once, then you're free to deploy over-the-air updates for as long as you'd like. Though you'd need to use a framework like React Native, Ionic, Flutter, etc which supports it. Essentially anything where you can change app code without making any changes to the underlying native code (as that would require going through the app review process again to publish those changes).

Opus 4.1 is better, but imo not 5 to 6 times the price better.

This makes me miss Google Cardboard and Daydream.

https://en.wikipedia.org/wiki/Google_Daydream


Cardboard still exists. Daydream was shelved. I might boot up my Lenovo Mirage to see if it still has any functionality that works. (I bought and never finished Virtual Virtual Reality which is actually well regarded)


I miss Cardboard too. That little magnet "button" was one of the most creative ways to create an input method I've ever seen.


I think RAG is out of favor because models have a much larger context these days, so the loss of information density from vectorization isn't worth it, and doesn't fetch the information surrounding what's retrieved.


That's true if you use RAG to mean "extra context found via vector search".

I think vector search has shown to be a whole lot more expensive than regular FTS or even grep, so these days a search tool for the model which uses FTS or grep/rg or vectors or a combination of those is the way to go.


On its own, nothing seems out of the ordinary. It's the extremes that La Liga takes to ensure they're getting that 200€/m that makes it insane.


If your Youtube video is 8 minutes or longer (and your channel is monetized), you're able to place midroll ads every minute or so to maximize ad revenue. Typically Youtube only serves a very small fraction of these midroll ads to each user; usually every 10 - 15 minutes. So 16min+ has been the sweet spot.

It's this ad incentive that has made long-form videos more popular on Youtube.


Yeah, I have unsubscribed/stopped viewing many specific creators because of this.

They start cycling content and using innovative ways to make videos artificially longer. Some videos of have "what this video is about" and "Summary" sections which can be even half of the video length in total. Sponsored sections are getting longer. There are longer pauses and less editing. The list goes on.


every 10 - 15 minutes? Thats cute


Yea, I've seen some YouTube videos that display ~30 seconds worth of ads every 3.5 minutes of a hour long video. It's gotten quite ridiculous.


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: