devidw
Given the fact that we are limited to seeing only one output at a time, it's kind of hard to compare outputs from different models, adjustments made to the prompting, and sampler settings.
But even when keeping the generation parameters the same (e.g., to test for reliability in the output) and just going for multiple passes, there is no easy way to have a side-by-side comparison to keep track of the outputs from the multiple "rounds".
I really like excalidraw and thinking visually, so that got me thinking: why not build a sort of "open-world" playground to just place nodes that represent system, user, and assistant messages on a big canvas? Since they all fit on the screen, we have this easy side-by-side comparison that is basically unlimited in the number of direct side-by-side threads at any level. Meaning one user input could lead to two assistant outputs that in turn introduce new sub-branches that can be tracked visually.
https://github.com/devidw/tabgod/issues/1#issuecomment-19336...
Execute any JavaScript from any browser tab on any browser tabs
It doesn't feel like a very nice fit in the context of a digital reading experience. Forcing the user to leave the current reading position to get the additional information at the bottom of the page, and struggling to find back into the text.
But I guess this only applies to the classic implementation of footnotes at the bottom of the page. I assume there are better implementations with a more digital character available.
I think the only way to opt-out is to alter appearances and interactivity on our side with something like a browser extension.
I would love to see your conclusion about the usage of parentheses respected by authors more often, it makes content so much more accessible and would make a solution like this obsolete.
Quoting Wikipedia:
> In most English-speaking countries, an unqualified word "bracket" refers to the parenthesis (round bracket); in the United States, the square bracket.
https://en.wikipedia.org/wiki/Bracket
---
Thanks sharing your thoughts.
What makes the technique useless for your reading/browsing experience?
The concept of collapsing any text between brackets "(" and ")". Like code folding in IDEs but for the web, mainly for reading long articles blown up by tons of brackets. Wikipedia is a good example.
> tried to interact with the “Try it here” section
Make sure to enable the checkbox on the right side of the demo saying "Collapse brackets", toggling it will turn every pair of brackets into an interactive element, collapsed by default, expandable as needed.
Realization: A web without brackets is the better web – folding all brackets
Actually, you can go ahead to such a spam page and view the index.html page of the current folder and what you will get is an HTML file with PHP code, which is not interpreted by the PHP interpreter, because the file has a dot HTML extension.
In the script, you can see from where the script fetches the displayed data and from which IPs the data is requested and received.
I copied such an index.html file into a gist: https://gist.github.com/devidw/ce2bdb78bb2e30a8e8437acc2c587...
I also wrote a blog post about the details of what is happening in the PHP script: https://david.wolf.gdn/google-untitled-links-i-found-the-sou...