Writing

Random find of the day. This Github repo gives us a Blockbuster Video VHS insert template. I made one of The Matrix.

Josh Cunningham wrote a piece called Imagining a Personal Data Pipeline. I started exploring his project, pdpl-cli, which helps you download your personal data and pipe it out to your desired output. I’ve been thinking extensively about this problem for a number of weeks now since I’ve exported my Google Contacts into Obsidian. However, with the lack of database support, I thought about self hosting it. Enter the Personal Data Pipeline.

It’s essentially ETL jobs with integrations to third party services to “recipes” that you can write in yaml and customize to your desired outputs. I think this helps a lot more than determining data schemas for specific third party data integrations and having the raw data in a personal data lake. (Or really maybe a document store).

The idea is to have it local-first and maybe include a sync-thing or cloud syncing as an optional add-on. There’s an emphasis on privacy, although my bigger fear is vendor lock-in. I’ve become so reliant on Google, Apple, and other services that I don’t feel like I own my personal data anymore. Also, as a web developer, the hardest part is grabbing my own data from the sticky hands of these cloud services. Also, this emphasis on files over apps makes a lot more sense to me than the walled garden approach we’ve become accustomed to.

Typst is “a new markup-based typesetting system that is powerful and easy to learn.”

It seems very interesting as an alternative to LaTeX. A lot of emphasis on typesetting. I haven’t gotten the chance to work with LaTeX, so I’d be curious if jumping over that and learning this might be a good alternative. Plus, it has a multiplayer feature built.

Check out their Github, written in Rust.

The post from Latent Space reporting for this quarter’s AI review, The Winds of AI Winter, tries to analyze the macro trends for AI. Long story short: there are high doubts about AI’s current capabilities and its distribution is uneven. “Time to build, or else AI Winter is coming”.

• ChatGPT’s growth has been level (0%) over this past quarter. Related: The Unbundling of ChatGPT (Feb 2024 Recap)
• I forgot that Google is in a downward spiral for “Google AI Overviews being bad, bad, bad, bad (after the Gemini mess)“. I’m really interested in how Google is trying to bounce back. Yesterday at the theaters, they were trying very hard to advertise the new AI features on the latest Pixel.
• A bunch of AI product recalls like Rabbit, Microsoft “Recall”, Figma AI, McDonalds drive-thru AI, Discord cancelling Clyde
• Non-Acquihiring means there isn’t an acquisition to hire the talent from one organization to another, but taking away talent to another company without acquiring the company. Adept lost their co-founders to Amazon. And Inflection Pi to Microsoft
• And of course, over-hyping technology and having high churn, like Harvey in private

The biggest takeaway is big spend and slow return. The the Goldman Sachs report and Sequoia’s report break it down. The one that telle me the most was the distribution of stock returns per phase of AI, where Phase 1 is the chips where Nvidia are doing the best. Phase 2 is the infrastructure running behind it, which is increasing. But Phase 3 and 4, of enabled revenues and productivity are flatlined.

We’re hoping the future will hold better for AI in general, but I’ll leave this quote from Asimov.

The future is here, but it is not evenly distributed.
Isaac Asimov

Simon Willison’s post - Give people something to link to so they can talk about your features and ideas

I evangelize this tool a lot, but OpenAI really aren’t helping me do that. I end up linking people to my code-interpreter tag page because it’s more useful than anything on OpenAI’s own site.

Related to: Learn In Public and Working in Public. It helps with the Open Source community in which you open source your knowledge and let people come and collaborate.

One obvious goal here is SEO: if someone searches for your product feature you want them to land on your own site, not surrender valuable attention to someone else who’s squatting on the search term.

You can drive traffic to this site because people could actually use it. Go back to the tiny web, and not where big corporations ruin the Internet by spamming search engines through SEO.

I personally value the conversation side of it even more. Hyperlinks are the best thing about the web—if I want to talk about something I’d much rather drop in a link to the definitive explanation rather than waste a paragraph (as I did earlier with Code Interpreter) explaining what the thing is for the upmteenth time!

More links are better. I think Willison’s blog is an excellent example of adding curation with links, but let the links stand up for themselves. Also, see Waxy.org - Andy Baio lives here.

Give people something to link to!

A curiosity sparked because my co-worker pointed out the Spring Framework is currently owned by Broadcom. That’s because VMware is now part of Broadcom, and Spring Framework was part of VMWare prior. I had an inkling there was a scandal Broadcom had awhile ago. Low and behold: SEC Charges Four Current and Former Broadcom Officers for Backdating Options (Press Release No. 2008-87, May 14, 2008). In 2010, the charges were dropped - Chipmaker Broadcom stock options backdating case ends. That’s not all though. Their first CEO, Henry Nicholas had criminal charges. The SEC dropped those charges in 2010, but that was enough to disgrace the people involved. Their former SEO left Broadcom before the SEC investigation back in 2003 to “work on his marriage”. He was sleeping with prostitutes.

Read: Henry T. Nicholas III: A human tragedy. And the follow-up a decade later with Orange County tech billionaire Henry T. Nicholas III charged with drug trafficking following Las Vegas arrest.

The court filings in the past case painted Nicholas as a ruthless entrepreneur who slipped drugs to competitors without their knowledge, who threatened the lives of employees he believed had turned against him, who had the means and motive to flee beyond the reach of justice in his private jet.

His start was in researching ICs, and later that’s what Broadcom sold. He was given awards for his research. He’s like a walking contradiction, trying to save his name with his philanthropy while making wildly bad decisions. File under downfalls.

It’s always fun when the MDN documentation doesn’t tell us what the options are within the methods page, I you have to dig deeper into the constructor. The TypeScript bindings don’t always make sense.

MDN - toLocaleString

Anyways, I needed a datetime string with the date, time in hours and minutes, and the timezone. I’ve made the timezone explicit and came up with this snippet.

date.toLocaleString("en-us", {
  year: "numeric",
  month: "short",
  day: "numeric",
  hour: "numeric",
  minute: "numeric",
  timeZone: "America/Los_Angeles",
  timeZoneName: "short",
});

This would return something like “2024-07-03, 12:00 PM, PDT”.

The shorthand option is must be used in object destructuring to remove redundant prop: value to be prop.

What I wanted was this shorthand:

const { asFragment } = render(container);

But what codemod generated was the following:

const { asFragment: asFragment } = render(container);

While both are true, I would rather have the shorthand, as that’s what we have strictly for eslint rules. This Github Issue helped describe the solution on how to implement in your transform file.

const code = j.objectProperty(j.identifier("h"), j.identifier("h"));
code.shorthand = true;

j.variableDeclaration("const", [
  j.variableDeclarator(
    j.objectPattern([code]),
    j.callExpression(j.identifier("require"), [j.identifier('"packagename"')])
  ),
]);

How to embed TikTok videos in Obsidian’s editing/live preview and reading mode.

<iframe
  src="https://www.tiktok.com/player/v1/7382225350710824222?autoplay=0"
  allow="fullscreen"
  style="width:100%;height:50vh;"
/>

where 7382225350710824222 is the video id you get from TikTok from this example link: https://www.tiktok.com/@_jen_hamilton_/video/7382225350710824222.

The TikTok v1 player has a bunch of controls that you can use to modify from this documentation.

There’s still some wonky height issues, and I suspect this is the default height Obsidian has on the container around the editor. I gave it a good enough viewable height of 50vh, which should be enough.

Example Video

Reference links

• How do I embed TikTok Videos? - Help - Obsidian Forum
• Emedding TikTok Videos - Help - Obsidian Forum
• html - How to stop auto play in an embedded video? - Stack Overflow
• TikTok for Developers

d3 v7 is packaged as ESM only. That made it difficult to use in the Node.js REPL. From this StackOverflow thread, I found a solution.

let d3;
import("d3").then((module) => {
  d3 = module;
});
// Then you can use d3 anywhere once the promise is resolved.

Extremely esoteric bug. I’ve had an open issue for the past three months related to this other issue. I couldn’t figure out why npm’s registry would pull an older version of d3-scale-chromatic when I upgraded the @observable/plot library. As a future note for myself, if I run into this issue again, I’ll have to go into the package-lock.json file and update d3-scale-chromatic to v3.1.0.

From this ACM paper, On the Dangers of Stochastic Parrots, the hypothesis is maybe these LLMs are parroting back what we already know and aren’t learning. We can probably, maybe, safely say that is no longer the case.

From The Progressive: Students Have the Right to Protest Apartheid. My lovely fiancé wrote a piece about the student protests for Palestine.

While the Republican-led Congress has used its power to condemn any protests on behalf of Palestinians, where was the same outrage and zealousness when white nationalists came to college campuses?

I forked a small demo from Wes Bos and ran in on my own sandbox. It seems much cleaner than using a script tag per each dependency, and works on all major browsers.

In my little demo app, I placed the importmap in the head of the document. I’m using react as well as my own utils file to test out the functionality.

<script type="importmap">
  {
    "imports": {
      "react-dom": "https://esm.sh/react-dom",
      "react": "https://esm.sh/react",
      "utils": "./utils.js"
    }
  }
</script>

Read more about importmap on MDN

From there, I imported these libraries in a script module.

<body>
  <div id="app"></div>
  <script type="module">
    /* eslint-disable */
    import { useState } from "react";
    import { createRoot } from "react-dom";
    import { formatMoney } from "utils";
    createRoot(document.querySelector(`#app`)).render(formatMoney(100.2365));
  </script>
</body>

I had a hiccup with [plugin:vite

Luxon datetime library defaults startOf and endOf methods to UTC. But this isn’t great for end users who don’t live in UTC / GMT. To offset this, we need to grab the time zone offset from the user’s system.

const systemTimeZoneOffsetInMinutes = new Date().getTimezoneOffset();  
const systemTimeZoneOffsetInHours = systemTimeZoneOffset / 60;

Then we can add the offset to the Luxon datetime object.

const now = DateTime.now();
const startOfLocalDay = now.startOf('day').plus({ hours: systemTimeZoneOffsetInHours });
const endOfLocalDay = now.endOf('day').plus({ hours: systemTimeZoneOffsetInHours });

I was listening to this podcast: The boring JavaScript stack featuring Kelvin Omereshone (K.O.O) (JS Party #319). My takeaway is I should be using a boring stack to build my business ideas. Leave the shiny new toys for tinkering and use something stable.

While Kelvin talks about the benefits of using Sails, I was thinking about the other “boring” things that constitute a boring stack. It goes beyond tecnologies, like how to run a business, how to organize your team, and how to market the product. And when those things are boring, they are unsexy and oftentimes neglected.

Related: Kelsey Hightower’s nocode

The best way to write secure and reliable applications. Write nothing; deploy nowhere.

I wrote a script that updates my daily notes to add properties. Since I’ve been using Obsidian for awhile, properties was introduced much later. I’ll walk through my code.

// Get the current file title, which is in a common date format: YYYY-MM-DD
const currentFileTitle = tp.file.title;

// Set folder you want to get latest file for here
const folder = "Calendar/Personal Reviews/journal and daily review/2022";

// Get all files in that folder, including nested folders
const filesInFolder = app.vault.getMarkdownFiles().filter(file => {
    return file.path.startsWith(folder);
});

// Sort files by file name
filesInFolder.sort((a, b) => a.basename < b.basename ? 1 : -1);

// Get the index of the current file
const currentIndex = filesInFolder.findIndex(file => file.basename === currentFileTitle);

// Get basename of previous and next TFiles to be used in link
let previousEntry = '';
let nextEntry = '';

// Wrap it around a try catch block in case there's something wrong with getting these basenames
try {
    previousEntry = `[[${filesInFolder[currentIndex + 1].basename}]]`
} catch (err) {
    console.error(err);
}
try {
    nextEntry = `[[${filesInFolder[currentIndex - 1].basename}]]`
} catch (err) {
    console.error(err);
}

Here’s my template that the templater plugin uses in markdown.

<%*This is where the JS code above is inserted-%>
---
tags: 
    - logs/daily 
created: <% currentFileTitle %>
previousEntry: 
    - "<% previousEntry %>"
nextEntry:
    - "<% nextEntry %>"
---

I’ve had to modify this depending if the file already has properties or not.

I got a chance to play around with Elicit and was reading through Adam Wiggin’s Tweet. I think I might make a quick web app trying multi-column queries about a particular topic and allow more of a computational style of conversational chat.

I was getting some errors in my unit test output because I converted my SVGs to React components using SVGR. As such, I needed to re-configure Jest accordingly.

1. Create the mock file

export default "svgr-div"; // This is the kebab case that jest is looking for
export const ReactComponent = "div";

// @see https://react-svgr.com/docs/jest/ for setup

2. Configure jest in package.json

{
  "jest": {
    "moduleNameMapper": {
      "\\.svg": "<rootDir>/PATH_TO_FOLDER/svg.js"
    }
  }
}

In practice, you try to eliminate task barriers that require back and forth communication by setting dedicated block of time to meet. An office hour, reminiscent of your professor’s office hours, means you broadcast to your working network you are available in this block of time to chat about open loops.

The problem this solves is eliminating the number of unknowns in your work, whether that be client feedback, questions about a design contract, or ambiguous bug tickets. These tasks become open loops are require external help in order to uncover what tasks you have to do next. It also lowers the communication gap with your work network or team and bridge that gap of understanding. The end goal is it helps with end output and/or goals.

Championed by Cal Newport and highlighted in his book, Slow Productivity.

In VSCode, open your palette, and type %, you get a fuzzy search across your files. And you can use navigational keybindings like cmd+up or cmd+down .

I periodically try to dump all of my mental contexts into my notes. I noticed this comes in a few different flavors.

• Safari Tabs on my iPhone and iPad: Goes into daily notes
• Random thoughts - go into drafts and get processed daily
• Shopping - including groceries, bills, and things I would like but not need. Those all go into a task manager
• Dangling tasks - gets reviewed in the task manager inbox weekly
• Personal thoughts - journaling every two weeks, although I want to do this more often
• Events I have not logged - go into my calendar.
• Downloads folder - weekly processed into other folders.
  • Long-term storage in Dropbox and backup drive.
  • Short-term in Inbox folder
• Photos - place special ones in folders. Select shots in favorites.

I am trying to think of other things that I purge. I used to write in a notepad and write on the page of everything I was thinking about. I used to write about people I have not spoken to in a long time that I should reach out to, mantras I came up with, or completely random thoughts.

Tiago Forte recommends putting all unused files into an Archive folder. This This activity helps remove our digital file mess and stops us from immediately organizing everything.

Stands for Observe-Orient-Decide-Act (“OODA”) Loop. It’s an iteration because it feeds back into itself. Initially applied for military use, but has since been used for business decision-making and personal development.

• Observe the situation and mentally note any key data points.
• Orient by placing data in the context of existing knowledge and mental maps to create a picture of the current situation.
• Make a decision on how to act in light of that situation.
• Act and assess for any necessary adjustments.

Source: Sahil Bloom - The Curiosity Chronicle

I’m finally cleaning up my RSS feeds because it’s too untenable. In my Readwise Reader, my inflow is far greater with things I know I won’t read over things that I know I’ll read later. Hence trapping me in a cycle of scrolling over reading. I’m going to unsubscribe from a bunch of things.