Stream

If you use Chrome and envy the vertical tabs that Arc has, I just modified a few lines in Polychrome to support side panel:

GitHub - nerdymomocat/polychrome-sidepanel-crx

Contribute to nerdymomocat/polychrome-sidepanel-crx development by creating an account on GitHub.

https://github.com/nerdymomocat/polychrome-sidepanel-crx/tree/main

Supports tab groups, multiple windows, search, drag and drop, pinned tabs, and duplicate tabs removal!

It has been so darn irritating realizing that I click on, consume, understand, think about, create, and recall information in different ways.

This is possibly the most irritating realization I have had about myself. It contextualizes so much, but I wish I was not this way.

AI Email Labeller is a nifty Google Apps Script that allows you to automatically label your emails based on your prompts. It’s hosted on Google Apps Script, enabling users to leverage the Gmail API directly, even those in a managed workspace like .edu domains. The only cost involved is the OpenAI API.

Check it out here:

GitHub - nerdymomocat/gas-openai-gmail-labeller

Contribute to nerdymomocat/gas-openai-gmail-labeller development by creating an account on GitHub.

https://github.com/nerdymomocat/gas-openai-gmail-labeller

and watch a preview here:

The one major thing GPT has enabled me to do is be able to create UIs for scripts that I would have just relied on code/spreadsheet for. Makes it much more likely that I will actually end up using something too!

Sketchnotes catch my eye but don't hold my attention

I've always been a fan of the concept of sketchnoting, but I have realized I don't consume information that way. And it really makes it much more unlikely for me to sketchnote, which honestly is sad. What I actually like consuming information from are infographics or supporting images, not icons.But, sketchnotes serve a different purpose for me. I don't know what part of sketchnotes makes me want to click on the link so much, but I do not actually want to learn or read the illustration? (Or maybe it is just that I have run out of ADHD meds at the moment, and have 0 visual focus, but that's unlikely).They are so irresistible for my mind. It's very much as if my mind goes Oooh, a pretty well thought-out illustration, that must have required so much effort. Let's click the link and read the article. They are the definition of possibly me judging a book by its cover.

Ever so often, I come across an app, I think I need, and it works perfectly, but I end up not needing it, at least not in the way I thought I'd need it. The latest app in the list is browser.horse, thanks to PascalPixel who very kindly allowed me to test the app a few months back. Browser horse claims to do one thing: create trail-like navigation structure to your tabs. And it does it perfectly.

On the worst of days, I have had 200+ tabs open in Horse; on the best, at least a 100. And it has been pretty stable for all of it. It is especially useful for situations where I open multiple references from a paper; if I'm deep diving into that paper. It feels like creating a flowchart, but in your browser.

So, why don't I use it? Tree-based navigation is not a new paradigm. SigmaOS and Stack_HQ are two other main browser apps that do something similar. And my lack of usage is not due to the shortcomings of browser horse; but rather due to my browsing pattern.

• Anything I browse does not get added to Google browsing history (and I want that - Yes, I know the privacy concerns; and I choose convenience and recall)
• Lack of extensions, so I cannot quick add to readwise, copy all links, quick add to todoist etc.
• But most importantly, I realized that I'm not really looking for an origin → tree structure when I want tab organization (aka trails).

Unfortunately, I am often deep diving into multiple parent papers/objects with overlapping concepts; and that is where the pattern breaks. I am looking for auto-organize and auto-deduplicate based on set of NLP kind filters.

If paper on A; ≤3 months, group it. I have multiple parent saplings and have overlapping leaves; with weird usage constraints - and if you don't have those; you'll enjoy Horse. It of course has some drawbacks - I wish it had full support for extensions + customizable navigation.

I, personally, won't pay for a browser, but if that is something that you're open to; and you often find yourself dead in 200 tabs, you should see if Horse is up your alley.

I have three reading models:
1. Skimming (50%)
2. Reading for pleasure (40%)
3. Understanding and dissecting a research paper (10%)

Contrary to popular advice, I don't take any notes for the first two categories, which make up 90% of my reading. I simply highlight and quote. Oh, and I also don't do booknotes.

When reading on my Kindle, I sometimes highlight stuff and sync it through Readwise. I don't do podcast notes either, opting for podcast snips instead. For the 3rd model, most of my 'note-taking' is actually replaced by AI extraction. For example, I ask what the Research Questions (RQs) are in a particular paper.

The only notes I do write are focused on processes (how to do something), understanding (what's involved), or questions/ideas related to my work. For me, any other written notes just serve as search pointers. I do write a lot outside of note-taking, like blogs, journals, letters, and academic papers.

I decided in May to start micro-blogging on my own website here. But as I've mentioned before, I'm not a fan of writing in markdown. Quarto offers a combination of md, html, and yaml, which works for me. So, now I write my initial thoughts on Twitter and then clean them up into YAML format.

It is heavily inspired by stream.thesephist.com.

It serves as a timeline of my thoughts. If you're curious about how I convert Twitter posts to YAML, you can find the output here. The exact prompt to achieve this can be found here.

Anyone has suggestions for in-background AI-based document editing apps? I don't need the app to tell me what to write next. What I'm looking for is an app that suggests changes like, 'This needs rephrasing to connect well to another section,' or 'Merge these two repeated sections.' 

I think my gripe with AI writing tools is that they're very generative rather than being aids for editing.

It's like someone misunderstood the term generative AI and decided that content generation is its best use. But writing a 5k+ token paper in my voice? That's a whole other ball game. It would need a strong mix of vectorization, concept mapping, segmentation, and context addition to be truly helpful as an editor.

I understand that achieving this level of sophistication is extremely difficult. Still, I find myself continually trying out these tools, secretly hoping for a miracle because, let's face it, I hate editing.

I have to say, WatchOS completely demolishes WearOS in the range and kinds of apps it provides. I have not been able to find a single useful app (for my purposes) on the WearOS store and that is just disappointing (and making me question my choices).

Something I have always done when I want to draw but I do not have something specific in my mind to draw. This is my favorite mode of relaxation - Alphabet designs.