I'm a software professional with a love for end to end product building. I focus on reliability, security, and simplicity in web applications and services. This site is a living time capsule, where I archive work, learnings, experiments, and ideas.

Right now, I build high-volume network traffic routing, management, and classification systems during the day, and my mini software studio in the evenings. I'm also getting back into shooting more 35mm film.

Activity

Loading... Getting track from Spotify

Posts

Tags
Between the outputs Flaws are cheap and forgeable too, so we aren’t special because we’re imperfect. When the act itself is the point, we reveal ourselves with the care we put in. When cheap proxies no longer signal much, we don’t lose the ability to signal care. We’re mourning the loss of easy proofs of humanity and mistaking those forgeable outputs as all that humanity is. The integrity of the human touch reveals itself over time. You can forge her all the heartfelt love letters you want, but it shows in every missing act. Takes on personal software There’s this idea circulating that LLMs make software so easy to write that everyone will just build what they need. It has some terms like: Disposable software Highly bespoke software Personal software Vibe coded apps I think it’s interesting, and still forming my opinions about it, but as of now I don’t feel that it will take off. There are many reasons why I feel that way, but I’ll just focus on two for this post. Software is becoming a restaurant Everyone loves a good analogy to put something complex into simpler terms. Right now, software industry analogies are everywhere as a way to simplify the tech job market and where software is headed as a career. I think that some of them kind of work, but maybe not for the intended reasons. Software has been an anomaly as a career for much of its existence. The idea was that you’d get a college degree or bootcamp certificate, no licenses, and get a 100-200k starting salary for knowing how to do this thing that others find mysterious and hard. Kids were sold this dream for at least 15 years now, that they just “learn to code,” and get a job with a fun office and free food.
Does the attack surface of bragging matter? These are a few common quotes right now: “I’ve not written any code in months now” “I don’t even read the code anymore. It’s that good” “We’re a tiny team that now ships full stack apps end to end for our clients instead of just designing them” “I’m a PM and feel more empowered than ever. I made { some big project } in just 48 hours without writing a line of code” “I built an entire { massive project } in 48 hours. The future is here” In a lot of cases, it doesn’t matter. At this point, a lot of the demos remain in a prototype stage on local, and are semi-functional just for the demo. The tech fleece looked fake Today, I woke up to the shocking news that the US captured Maduro in Caracas. I won’t go into the loads of political takes from across the US and Europe, and instead wanted to talk about why it was among the most interesting days of geopolitics of my life so far. I happened to be in the right place at the right time during a trip to La Guajira, and was able to hear perspectives from Colombians, Venezuelans living in Colombia, and Venezuelans that left from Maracaibo this morning. Human in the loop In my opinion, 2025 really was the year of agents. Definitely many crappy ones, but also some good ones. I’ve been developing Wasy, an ambitious AI-augmented CRM for Latam based businesses every night for over a year now, and seen the cycle from every angle. There’s of course a lot that I couldn’t go deep into, but in general, I feel like working on this project got me extremely close to LLMs, research, techniques, and some of their shortcomings. Organization is our problem I’ve changed my views over the past few years on some controversial topics due to realizations while working at a big company. Very little is ever optimal when it’s something abstract and deals with humans. These things are inherently lossy and need some kind of “organization.” In a relatively small organization of only 1000 people, changes made just a few years ago got completely lost. Migrations from platform A to platform B with authors leaving the company in the mix made decisions practically disappear. Team A uses team B’s platform incorrectly for months without anyone realizing, until eventually it breaks. Uh oh, but Team A’s component was load-bearing and set up by a dev on vacation right now. Team’s gotta figure it out immediately and service is interrupted for over 3 hours. Various teams building for problems that don’t exist. Doing hard work every day on the wrong problems, but the organization is not configured to speak honestly. XY problems everywhere you look. Speeding up a tool, service, app, or layer that never should have been made. That’s just 1000 people, and in a highly successful organization of smart folks that seems like it’s flourishing from an outsider’s perspective. On the inside, it sometimes feels like everything’s on fire. Issues all over the place with not enough time in the day to even log it all in a coherent way. This is a detail oriented person speaking though, the truth is, the organization succeeding while being on fire is kind of just how they work. It’s not optimal, but it’s optimal enough to continue kicking. First year of coffee Last year, I was buying a latte or two every day. For something so routine, it started adding up. Between the cost and the time spent going to cafes, I was averaging around $8 per drink. I decided to look around for a few hours at the espresso subreddit, asked Claude a few things, and found a setup that was budget friendly, accessible, and easy to understand. I went out and got my setup together a year ago today. I went with: Balancing the worst choice I went too deep into utility maxing for too long. I lost myself really, because my life became a spreadsheet and I missed the things that make me who I really am. I noticed this most recently after I got into cycling, purely for utility purposes of avoiding traffic after Seattle opened the Alaskan Way bike lane. I was like “what’s the single bike that can do it all, isn’t expensive, gets me point A to B, and can do fun rides on weekends?” Too many doomers Over the past few years I’ve actually gotten more optimistic about the future of humanity. Seen how fast we’re able to move when enough interest and resources are on board in important moments. Especially in open-source. Noticed the upsides and downsides of EU-cookie-banner-causing styles of regulation. Online however, I keep seeing viral arguments appear that turn deeply doomer, like these two: Shut down or highly limit all large scale LLMs over water use concerns. AI is a slop machine without real value. Cool new innovations in the space are useless and are burning the planet. One example about using Genie 3 to “walk around” still photos or art said, “If you can walk around in photos then the whole reason for photography’s existence is laid moot.” Like, seriously? There are so many useful cases and loads of people pile on agreeing that it’s worthless And then in person, I started hearing them too. Then I saw this post, which made it make more sense: Confidence in software One of the best signals of quality software is user patterns showing high trust. In middle school, we had this website where we’d keep a quarterly journal of our achievements and progress. Every few months, we’d answer questions and write essays and attach photos and such. I remember seeing classmates crying a few different times because the content didn’t save after hours of writing it up. Usually it’d be something like them uploading a file a little bit too large, an alert popping up, and then the page reloading from some unhandled error. Jony Ive at Stripe Sessions A couple of days ago, Jony Ive gave a talk at Stripe Sessions. He dropped some gems, especially for designers and founders focusing on a lot at once. To me, it feels like everything’s marketed and shipped focusing on speed speed and more speed. It’s about getting “more” for less, with “more” being focused highly on monetary outcomes. Quality’s always been really important to me in products I consume and has been my goal for everything I create. But really, it’s hard to pursue top quality in a world focused on speed and clear metrics first. More voice assistants In Seattle, I still have yet to hear anybody talking to an AI assistant in public. It doesn’t seem like it at least, since AI chats are really assistant-coded: one input and wait for a response output with no active listening cues really. I’m a big fan of the newly launched Raycast iOS app which comes with a great UX for speech to text transcription, even directly from a widget. It’s a hold to transcribe + send interface, and as you release, the LLM immediately starts responding in text. Also, major shoutout to their amazing team who has fixed the only two nits I’ve ever had with their products in under a week both times! Shorter and shorter thoughts Today, Hieu Pham posted a realization that resonated with me: I noticed an alarming change within myself because of LLM usage: I have become lazy of reading. These days, if I ask grok a question and its answer is long, I would get annoyed and tell it to give me a TDLR. Sometimes, I even become impatient enough to tell it to answer concicsely in advance, which I just stop because that usually leads to worse answers. I am afraid that as Grok becomes infinitely better, and its short answers are all informative and correct, I will become addicted to consuming short chunks of text. Pretty much like the way some people are addicted to the short videos on TikTok. That is scary. Losing the enthusiasm to read is the key to falling behind. And if it is happening to me, I suspect it is happening to many who are growing up in this age of AI. Changing how I use this site I really enjoy the weblogs/microblog/TIL style of sites that people publish to frequently. I already set myself up a super easy process to publish photos or posts in a few seconds, so I realized I might as well use it more. I’m going to start posting anything, regardless of length or depth, like transcriptions from my driving thinking sessions, things I like, etc. It’s fun to have people stumble on your content and reach out, so this kind of maximizes those chances! Our newest computer-using users A few weeks ago, I finished up my post on AI ethics. I talked about 2025 being the year of AI agents, and how I am specifically interested in the future of autonomous computer-using agents. Basically, instead of crawling and scraping webpages to take an action and hoping that those pages never get changed, we will see more approaches using vision and contextual understanding that can practically spoof the usage patterns of a human. Questions about AI ethics: what's next? January 2026 updates As this site is simply a time-capsule to revisit and see how my thoughts shift over time, I’m fine saying that I think I was badly wrong multiple times over the past few years regarding LLMs, the AI boom, and the future of it all. I’m done speculating. Last years’ updates on this post were garbage, and all of the answers to the questions were too. I wrote a follow up post called “Human in the loop” where I actually go into specifics on this matter with concrete experience. Combatting helplessness I’ve noticed a growing subconscious sense of helplessness in a lot of Gen Z and Gen Alpha, with myself included. My hypothesis is not that the youth is any less intelligent; it’s that they have exclusively lived situations with reliance on powerful tools. Right now, this includes LLM reliance for homework, needing maps to go anywhere, generally shutting the brain off during day-to-day tasks, and feeling helpless when things don’t go as expected. Reach IPv4 on an IPv6 host When I began my personal site transition earlier this year, I decided to save some money and take on the challenge of an IPv6-only host. I thought it would be a great way to learn about the current state of IP and pick up new knowledge on IPv6. I was totally right in my assumption; I’ve learned a lot of new things and gotten frustrated a few times in the process! Most things I’ve ever learned have been centered around IPv4. From university classes to tasks as a software engineer, rarely does an IPv6-only network surface. Planned out systems As a kid in Hilo, I vaguely remember a story about some eucalyptus forest planters planting their trees much denser than they intended. I don’t remember all of the details, but I know that generally for a healthy forest that is suited for wood harvesting, the seeds shouldn’t be too densely placed that they will fight to overtop one another. That is, unless the harvesters also can run a thinning process. Using the minimum complexity A lot of my life relates to complexity. My job, hobbies, and levels of motivation are all closely tied to this idea of intricacy. Specifically, the problem of optimizing for functionality without sacrificing simplicity. Tennis is an intricate sport. As a true beginner, chances are, you will spend 90% of the time picking up balls, unable to keep a ball in play for more than two shots. After learning the basics and drilling for months, you’ll improve and be able to hit much more consistently. Then, you’ll begin worrying about adding more power and angle behind your shots to be able to compete at a higher level. It’s at this point especially that you have to remember to keep it simple. Finding my mental peace We all have our unique idea of mental peace. I think that we know about it subconsciously and move our way towards it, even if we don’t explicitly know what or where it is. Now that the sun is out and the temperature is nice, I’ve been going on a walk through the neighborhood every day. Sometimes with my phone to take some quick photos, but usually just with a watch to track my steps. I’ve done just about every permutation of turns on the streets and seen every home multiple times. Reflection on AI Over a year ago, I wrote about my main use case for LLMs. At that point in time, the general public was finally going wild with ChatGPT. It seemed like the entire world was impressed at once. After using ChatGPT for a few weeks, I wrote that I thought, “my reliance on these tools will grow substantially.” 15 months later, I can confirm. These tools changed my life. Right now, my Mac has ChatGPT, Perplexity, Claude, and even locally running llama3:70b in the dock. I use all of them for different purposes and have an entire workflow mastered, knowing which to ask certain types of questions to. I usually know quickly when an answer is going to be garbage and have good ways of combining as many sources as possible. Despite all of this, I am tired. About this space Every year since 2019, I’ve dramatically changed the design or functionality of my personal site. Usually, I make some type of change because I want to try out different stacks or get inspired by someone else’s setup. This year Although the design is going to stay just about the same, I’m going to be making the following changes over the next few weeks. Moving away from Next.js and Vercel for my personal site. Self hosting and automating its infra management myself. Styling the majority with the TailwindCSS executable. with plain css. Rendering the markdown pages to a static folder with Hugo. Minimizing and containerizing the built site. Serving the pages on a tiny container with a small reverse proxy file server. Gin based API for backend tasks such as signing up to the newsletter using a variation of my go-web-server. Writing my posts out on any of my devices on my iCloud Obsidian vault, then building and deploying in one step. Privacy-first statistics with GoatCounter (to do). Why? In my most recent job, I’ve gotten interested in infrastructure and reliability. With heavy abstractions, it’s not easy to manage the entire system yourself and get the satisfaction of knowing how every piece works. Additionally, I have grown fond of the manifesto of Obsidian, specifically in the aspect of durability and reducing dependencies in every place possible. Making art in 2024 Update from January 2026 Humanity is not necessarily what makes art feel good. So far, the idea of “good” art has not changed that much to me specifically. For others, it clearly has. I am seeing frequent misclassifications of real art as “AI” online. For example, actual human editing mistakes misclassified as a hallucinated artifact or a beautiful sunset called “too perfect” and generated. It’s the same thing with other fields: a master steering the tool well can result in a great finished product. Sloppy steering is what feels lifeless. Url as state Retention of state through url is a lost art, yet with most modern stacks it is not overly complex to set up. What is it? Maintaining state with the url is just ensuring that important actions are reflected in the url. This means that if you fully refresh the page or send the link to someone, it should open the exact page you wanted within all of the right menus. LLM reliance Something I’ve noticed recently is how over time, a lot of my Google queries have gotten much more complex. Beyond just complex, they’ve become abstract and sometimes have unverifiable answers. In high school and college, the queries slowly shifted from basic knowledge questions to more of a “how to” style. This meant that a lot of times, videos on YouTube were helpful for learning. I used to think that the creators of these videos were geniuses. Like how in the world does this guy know how to solve all of these integrals? The decentralization problem A few days ago, I presented to my class about the thing everyone just loves to hear about: NFTs. Specifically how things can go wrong with them. I gave a bit of background on blockchain technology and then went into what an NFT really is. When I was writing my presentation the first time around, I practiced it once and realized I had left out a crucial portion. What is the point of this technology?

Other

Contact

Get in touch by direct message or by email at [email protected]. Subscribe for occasional updates with your email below.