Stream

I finally synced the block list and muted word list between my main and alt accounts on twitter, and I feel so relieved 😅. You might be wondering how I did it?

At the moment, it's a completely non-automated process because API costs money. I downloaded the archived data through X, extracted the block list and mute list from block.js and mute.js using GPT4 code analysis, and then copied and pasted the curl commands into a for loop for all the darn accounts with userid as a variable. I even took a scrolling screenshot of muted words, converted it into a list using GPT4-V, and muted those through a bash script with cookie tokens.

You can sense my desperation from this. I was tired of seeing the same nonsense across multiple accounts for over a year, and I just had to do something about it.

Mar 3, 2024, 08:52 PM: I got tired of trying to find userids from usernames for new blocks, because of course I wasn’t going to download the archive every time. There are websites you can do it from, but I wanted an option to not leave the website, and hence this comment from github helped. Paste it into console and voila.

id = JSON.parse(
document.evaluate(
    '//script[@type="application/ld+json"]', 
    document.lastChild, 
    null,
    XPathResult.ANY_TYPE, null).iterateNext().textContent
).author.identifier;

Someday, I will use these libraries but today is not that day

GitHub - trevorhobenshield/twitter-api-client: Implementation of X/Twitter v1, v2, and GraphQL APIs

Implementation of X/Twitter v1, v2, and GraphQL APIs - trevorhobenshield/twitter-api-client

https://github.com/trevorhobenshield/twitter-api-client

rettiwt-api

An API for fetching data from TwitterAPI, without any rate limits!. Latest version: 4.1.4, last published: 2 months ago. Start using rettiwt-api in your project by running `npm i rettiwt-api`. There are 2 other projects in the npm registry using rettiwt-api.

https://www.npmjs.com/package/rettiwt-api

Just a tiny Shottr appreciation post, for everything it has. When I started using Shottr, I didn’t have Cleanshot X. The only reason I got Cleanshot X was because (a) it had a deal on AppSumo; and; (b) I can record videos with my face in them easily.

But for everything screenshot related (and now you can paste in multiple screenshots too!); you should definitely use Shottr.

I recently discovered this app called Clop. And man, it is so good. I often take these 3MB full screen screenshots and directly paste them into Notion (including on these pages rendered as a website), and Clop basically auto-optimizes any images on clipboard, whether shot directly, or if you use ⌘+C in your finder on an image file.

The free version works for all images (though doesn’t have an option to auto-convert to webp, I so wish it did!). The paid version (or with limits on the free version) can also work with videos and pdfs, though, I haven’t tested that part much.

Clop - Image, video, PDF and clipboard optimiser

Image, video, PDF and clipboard optimiser. Copy large, paste small, send fast

https://lowtechguys.com/clop/

I have been wanting to explore paper reading apps that work on a spatial canvas like structure. I do want them to work with Apple Pencil because I feel the most comfortable writing with papers, so heptabase is not on the list. I am not looking for paper annotation apps like Highlights, more for excerpting regions of paper into a canvas I can drag those “components” around in. The main thing I am looking for that isn’t a usual part of most canvas apps like tldraw or concepts, is the ability to locate the “snippet” or excerpt back in the pdf.

The main app that does this is Muse, but (a) it is an expensive (fairly priced subscription) for just paper reading, and (b) I would pay that much if you could share the boards on web, but you need the Muse app, unfortunately.

So, I tried other apps, and I am going to try BoardNotes for a bit, because that kinda checks the boxes of snippets, canvas, Apple Pencil, and pdf locator. GUGA on the other hand is collaborative, has snippets, has a great canvas, works well with Apple Pencil, but does not have pdf locator unless the specific pdf is already open. Neither of them have a macOS app like Muse does, or have publicly shareable boards, but they both are pretty cheap, a one-time purchase. Kosmik 2.0 seems to have public universes as a thing even for free plans, and there is an extract option, that doesn’t remember pdf location, but remembers the pdf the extraction came from. Or maybe I should stop trying to find something and just use Muse or Liquid Text (though I am unsure how it compares with MarginNote).

Some apps that I tested

Val town script here (which you can import as a module if you want!)

@nerdymomocat/add_to_notion_w_ai

Uses instructor and open ai (with gpt-4-turbo) to process any content into a notion database entry. Use addToNotion with any database id and content. await addToNotion( “DB_ID_GOES_HERE”, “CONTENT_GOES HERE”//“for example: $43.28 ordered malai kofta and kadhi (doordash) [me and mom] jan 3 2024” ); Prompts are created based on your database name, database description, property name, property type, property description, and if applicable, property options (and their descriptions). Supports: checkbox, date, multi_select, number, rich_text, select, status, title, url, email Uses NOTION_API_KEY , OPENAI_API_KEY stored in env variables and uses Valtown blob storage to store information about the database. Use get_notion_db_info to use the stored blob if exists or create one, use get_and_save_notion_db_info to create a new blob (and replace an existing one if exists).

https://www.val.town/v/nerdymomocat/add_to_notion_w_ai

Demo using the template Money Database

@nerdymomocat/add_to_notion_w_ai_webpage

Example usage of the add_to_notion_w_ai val Try with the money database . Read and watch the demo run here

https://www.val.town/v/nerdymomocat/add_to_notion_w_ai_webpage

Bookmark for https://www.val.town/v/nerdymomocat/add_to_notion_w_ai_webpage

A video demonstrating the use of function - with an http endpoint

• Uses instructor and open ai (with gpt-4-turbo) to process any content into a notion database entry.
• Use addToNotion with any database id and content.

await addToNotion(
  "DB_ID_GOES_HERE",
  "CONTENT_GOES HERE"//"for example: $43.28 ordered malai kofta and kadhi (doordash) [me and mom] jan 3 2024"
);

Prompts are created based on your database name, database description, property name, property type, property description, and if applicable, property options (and their descriptions).

Supports: checkbox, date, multi_select, number, rich_text, select, status, title, url, email. Filters properties where either type is title or description starts with ✨.

• Uses NOTION_API_KEY, OPENAI_API_KEY stored in env variables and uses Valtown blob storage to store information about the database.
• Use get_notion_db_info to use the stored blob if exists or create one, use get_and_save_notion_db_info to create a new blob (and replace an existing one if exists).

These two comments on reddit exactly explain why I don't watch productivity content. Whether it be youtube or twitter or anything else — I do not want to be productive running courses, I do not want to be productive running a productivity business - but the content creation grind requires so much input — that sooner than later, that becomes your job, either the content itself, or the course itself, or the business teaching about productivity. And I do not fault creators for that. But, that isn’t me. And now it isn’t relatable. Ali Abdaal used to stand out for his content being productive at being/or learning to be a doctor, but that is no longer the case.

That+sponsorships -- but the latter is not productivity sphere related.

As much as I like Scenario, and am on the fence about AI art, I really do not like terminology shift from training → pre-training, and fine-tuning → training. Companies can say things like "train on your art", which still means the model was pre-trained on scraped art.

And the math & opt-in/opt-out debate aside -- I do not like the "lying by omission" aspect of it.

Like, a platform isn't more ethical than midjourney just because it provides services to fine-tune a model on a set of art samples. 🤷🏽‍♀️

I know, I know. I decided not be critical in the new year and only reinforce positivity, but what is up with really poorly done AI images used as thread’s main points or OG images for links? You want to save money by not paying an illustrator but you also want to save time by not spending hours and hours actually getting the perfect image? Is that what it is?

Because otherwise, how do you explain these images at all?

One of my goals this year is to create animated illustrations. Do you all have recommendations for programmatic animation libraries that can animate illustrations (not divs/elements/simple shapes)? I'd use procreate dreams but you cannot configure those with trigger states, so I can't have them be interactive and procreate is not vector based so I can't export a svg 😟

One of the hard parts about this not being associated with a project is I have open ended questions about the scope.

• Do I need trigger animations for illustrations or is that need coming from data animations?
• Am I looking for video exports, runtimes (like rive) or svgs?

Honestly, val.town is cool and fun and I have never been able to get something up and running as quickly as I have on val.town. Once repl.it stopped allowing you to run flask apps without upgrading, I knew it was only a matter of time, when I would need to switch to a different cloud function app.

I was able to move my reference manager from repl to val.town Jan 4, 2024: Notion Reference Manager

Here is a quick friction log:

I tried to use it as an IDE (based on around 3 hours of exploration), but there were a few challenges:

• The all-scripts-on-one-page layout made it feel less like an IDE.
• There was a noticeable delay between clicking run and seeing the output.
• The auto-format code feature often didn't work as expected.
• Every save increased the version, which was a personal irritation as I prefer versions to represent actual changes.
• Sometimes, saving didn't work if I collapsed the tab group.
• For it to feel like a true script IDE, it might need code folding functionality.

On the positive side, I appreciated the TypeScript argument suggestions. So, I've been using it more as an executable pastebin. Overall, I like how compact and clean the interface is, considering how much you can do with it.