Field Notes

As We May Think Again

Fri, 18 Sep 2026 12:49:53 GMT

Last week, I decided to audit an edX course on Human-Computer Interaction (HCI) to supplement my knowledge on the field. The specific course I audited is Georgia Tech’s HCIxI (Human-Computer Interaction I: Fundamentals & Design Principles). When I came upon the first set of readings it happened to include an essay titled As We May Think written by Vannevar Bush, who was an American engineer who headed the US Office of Scientific Research and Development during the second world war.

In the essay, Vannevar Bush encourages scientists and researchers to turn their attention towards inventing and innovating on tools that help people accomplish every day tasks or to improve quality of life. It’s an interesting essay, having read it 80 years later, and I would recommend everyone to read it.

Couple things I’m interested on:

In the article, Bush mentions that we are getting bogged down by the abundance of knowledge work from thousands of other scientists and researchers, but acknowledges that specialization is necessary for progress. There is a huge difference between the time it takes writing scholarly works and reading and understanding them.

We haven’t really slowed down in that account. New tools and technologies are born almost everyday, and new research gets published more often than that. Meanwhile, most of the things I learned in my undergraduate journey were introductory courses on the various fields of Computer Science. I doubt I would have learned even half of what’s out there if they gave me a couple more years, and it would probably be the case for other students too. Year after year students are expected to learn and digest decades of knowledge into four- and five-year courses. Currently, there’s not really a way to speed up this process (despite generative AI—in fact, AI has a tendency to exacerbate the issue).

In this aspect, internet technology and databases have helped immensely in curating and organizing our collective knowledge base across a wide array of subjects and specializations. Personal computers, storage devices, online documents, and the cloud can help us store and distribute this knowledge.

However, time and again paywalled knowledge remains one of the biggest blocker for knowledge sharing in that regard.

Memex: the first Personal Knowledge Management System?

In the essay, Bush also describes the idea of what he called the memex which is a physical desk—very similar to a desktop table—that contains drawers of information stored in microfilm reels with cameras to view each document. He also describes a mechanism to link these documents to each other, thus constructing a “trail” of information on any specific topic.

While I was reading it, it felt very much like what Personal Knowledge Management Systems (PKMS) software are like today. In fact, I believe most (if not all) of the functionalities he described for the memex are available in an app like Obsidian, with the only differences being the physical manifestation of the tool.

It is amazing how technology has evolved to allow for different interpretations of a tool addressing the same problems. And indeed knowledge acquisition and sharing must have been one of the greater problems in science and progress that it still boggles us today.

Today

Taking all this into current context, I wonder if generative AI really is the next step towards innovation, or have we missed a crucial step? We know that generative AI stands on the backs of our scientists, researchers, creatives, writers, technologists, and all other professionals who have ever published their work publicly (some not publicly) on the internet. It very much has the ability to comb through centuries of human knowledge and spit out an answer to any query regarding said knowledge.

Generative AI also has the ability to build on that knowledge and potentially create something novel. We already know it can exacerbate the issue of knowledge acquisition given the rate at which new knowledge becomes available to the point that humans can no longer keep up. But how do we properly attribute the new inventions it creates, and what should be the process for knowledge transmission among humans?

Regardless of my opinion about it, AI is here to stay; but like Terence Tao and other Fields Medalists points out in their Math and AI declaration I am also worried about what AI would do to human knowledge acquisition and sharing and whether we can keep up with it or not.

The future feels bleak

Thu, 10 Sep 2026 12:53:37 GMT

I find it incredibly hard to keep up with recent skill requirements in the Software Engineering space. Most recently it is leveraging AI in software development.

While I do use the assistance of AI in my day-to-day development tasks (whether professionally or just for fun), I can never really reconcile with the idea of letting AI do most of the work while I read and review its output.

I’m sure as developers and engineers we all have our reasons and personal stories on how we fell in love with programming and building software. For me, it is more learning how to build the software and the act of building the software myself that give me joy.

Much of that gets taken away with AI. I understand the advantage. It can churn out full websites and applications in mere seconds, while I would take days or weeks to produce the same. It is an unfortunate side effect of capitalistic greed for things to be done extremely fast.

I’ve read a few articles saying our new role as software engineers is to become AI orchestrators, managing and directing AI agents and then reviewing its output. And although that is quite different with managing people, the transition from primarily writing code to only reading code (and lots of it) is also quite a hurdle. (Is there research somewhere about people losing the skill when they don’t actively practice it?)

In creative spaces, I imagine it’s much worse. Art is a different discipline. I don’t consider myself an artist but I believe the process of creating art describes art more than the output does. And AI completely obliterates that process from weeks to mere seconds. (Luckily, I think a lot of people agree that AI art sucks.)

It’s a very weird time in history where businesses are having massive layoffs and we’re all just supposed to be fine with that. I think radical changes in our working habits and our wages need to accompany the radical change that AI brings. Do we still need to be working 8 hours a day, and do I still need to work more than 3 months to afford an iPhone? (On a Filipino minimum wage, a little over 5 months of work provided they don’t spend a single cent of their salary towards anything else. Hey, what value does AI bring again?)

I don’t have enough confidence in companies and businesses to put more value in humans than on AI. And I think they’re all just racing towards some goal post just to say they’re the first to do it. Companies don’t even have the confidence in other companies to trust that they would do the right thing, so why would I?

But maybe it’s not all bad.

Maybe AI will become a public good, and maybe (just maybe?) those big tech companies won’t hoard all the resources to sell it at a premium so that only businesses and individuals with abundant capital can take advantage of the radical change that AI brings, and maybe it will instead be available to all (like electricity, like they say; where the only setting is on and off—there’s no premium electricity, is there?).

Maybe if human labor loses most of its value there will be some other way to earn a living.

Maybe.

I have a website

Mon, 31 Aug 2026 00:00:00 GMT

Due to recent changes in my employment (I was retrenched), I have a lot of time on my hands to upskill and learn new software and tech. What better way to do that than to create my own personal website?

I always wanted a personal website, and back in my early days as a software developer I attempted to make one with GitHub Pages. It is heavily inspired by Twitter (now X) and has not been updated in five years. It also uses the free .github.io domain. Instead of updating the website, I decided to grab the opportunity to learn new things by starting from scratch.

Designing with Figma

During my 5+ years as a software engineer, I have often used Figma in Dev Mode to implement designs for various websites but have never experienced designing a website from scratch. I tried my hand at it with the free tier, and I can say it is difficult work especially as someone who has no eye for design (self-proclaimed) and have little to no idea about the design process.

Of course as a frontend-leaning software engineer there are many familiar concepts (like design system, theming, design tokens, components, etc.) but I find that being quite methodical, rigid, and structural about these concepts is way more beneficial for the design phase than actually implementing them in code. There was a lot of boilerplating to be done before I could confidently start with designing pages and views—which, now that I think about it, taught me why there is a lot of research that happens in this phase of software development. Changing one small thing like a brand color is not as easy as it would be in code.

I gave it up after finishing the light and dark mode for the mobile view. I figured I can learn it more deeply another time, and I could no longer wait to have something—anything—deployed to call my own.

Developing the site with Astro

I learned and used Astro for the first time. The documentation was very helpful and I have quite an expansive experience with Next.js so I was not a stranger to SSG and SSR. Currently, the whole site is fully statically generated.

At the start, I had opted for vanilla CSS for my styling but that went south pretty quickly. I am just a tiny bit smitten by the extra tooling that Sass provides, so pretty soon after my first few deployments I changed to using Sass. I also could have opted for Tailwind, but I am more used to “speaking” in CSS and Tailwind is exactly like not writing CSS.

Along the way I learned about CUBE CSS. In many ways I had already subscribed to this methodology without knowing it had a name. It is a good mental model to have, but I find it is much harder to stick to on a team setting (esp. for larger teams) unless there are tools that can enforce rules or everyone is aligned with everything all the time (impossible). I am currently in the process of refactoring my styles to use this methodology loosely.

Shoutout to Simon Dann whose website very loosely inspired the very early-stage design of this website. And also shoutout to Henry from Online for just existing—as a reminder to myself that beautiful websites still exist. I hope one day I can create something as pretty.

AT Protocol and the Atmosphere

This might sound like a digression but I swear it’s relevant.

When I first signed up for Bluesky back when all hell broke loose on X (okay, X was already a hellsite way before that), I did not know what the Atmosphere was. I recently started becoming more active on the platform after deciding I could use it as my primary source of tech news, and pretty quickly I started seeing more and more people talk about AT Protocol.

I read more about it, what it is and what problems it was trying to solve, and I’m interested enough in the tech that I decided to dip my toes a little and publish my write-ups on the Atmosphere. At the moment, I’m sure my process is still a little janky and immature but my current system is as described:

I have multiple publications. Currently: Field Notes, Changelog, and Against the Dying Light
Each publication is a content collection in Astro.
Blog posts in each publication is written in either Markdown or MDX.
Every time I publish a new blog post, I can decide to also publish it on the Atmosphere using Standard.site lexicons via a small CLI app that I wrote myself.

The CLI app is written in Node/TypeScript and only does three things: publish a site.standard.publication, publish a site.standard.document, and update a site.standard.document. I will improve it as time goes by and as needed, but right now it is enough for my process.

I have heard of Sequoia and astro-standard-site but I wanted to create my own because I wanted to customize the process to the way I like, and it helped me learn more about AT Protocol and the Atmosphere.

I admit it was a very confusing process, sometimes I wish the documentation was a little more informative and expansive—but I’m chalking that up to the simple fact that I don’t know what I’m doing. ChatGPT helped by answering a lot of my (probably stupid) questions.

Domain and deployment with Cloudflare

Also for the first time ever, I bought a domain. It was a fairly easy and straightforward process but I did have the urge to overthink my domain. In the end I went with riyana.dev, my first name with a .dev TLD.

The site is deployed in Cloudflare Pages - perfect for static sites and also free.

Plans for the future

There is a lot to be done. My website is very bare bones as of the moment, with only a homepage and my 3 publications. Some plans that are in line:

Publishing my Against the Dying Light blog posts to AT Proto
Adding a /tags/[my-tag] route to list my blog posts by tag
Adding a /bookshelf - I’m not sure how this will look like yet

And of course, the design will probably be ever-evolving because I can never be satisfied with anything that I create.

Hello world

Tue, 25 Aug 2026 00:00:00 GMT

My first post! This is really a test post.

Lorem ipsum dolor sit amet, consectetur adipiscing elit. Pellentesque risus libero, aliquet quis arcu id, ultricies rutrum velit. Mauris at lacus sit amet dolor pharetra convallis. Pellentesque ornare eget justo ut euismod. Suspendisse tristique at sem ut faucibus. Vivamus facilisis felis porta tempor pulvinar.

“The brown fox jumped over the lazy dog. Nam leo lacus, viverra eu orci sit amet, sagittis posuere orci. Morbi finibus ipsum dui, et mollis augue eleifend non.”

Nam leo lacus, viverra eu orci sit amet, sagittis posuere orci. Morbi finibus ipsum dui, et mollis augue eleifend non.

Section One

Aliquam erat volutpat. Vestibulum pretium vestibulum dui nec volutpat. Vestibulum vitae nibh malesuada, imperdiet lorem ut, elementum magna. Sed at ornare lorem. Aenean porta urna sit amet laoreet rutrum. Fusce id justo erat. Praesent dictum molestie ex nec gravida.

This is an update!

Subsection

const myVariable = "hello world!";

const sayHello = () => {
    return myVariable;
}

Sed aliquet velit blandit sapien scelerisque interdum. Integer eu rhoncus magna, et convallis est. Cras pulvinar et lacus vel tempus. Suspendisse potenti. Donec ornare turpis sed lacus tempus fermentum. Nam ut magna id lorem porta gravida. Quisque pellentesque ex sit amet elit eleifend ullamcorper. Donec vulputate volutpat nunc, ac sodales sem condimentum quis. Cras ut fringilla leo.