<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:dc="http://purl.org/dc/elements/1.1/">
  <channel>
    <title>John Nazarene Dela Pisa · Writing</title>
    <link>https://nazarene.dev/writing</link>
    <atom:link href="https://nazarene.dev/feed.xml" rel="self" type="application/rss+xml" />
    <description>Writing by John Nazarene Dela Pisa. John Nazarene Dela Pisa is an AI Specialist at FEAREX Technologies in Cavite, Philippines, building agentic systems and the infrastructure they run on. Before that he spent two years as a Solutions Architect across AWS and RHEL.</description>
    <language>en</language>
    <lastBuildDate>Mon, 05 Oct 2026 02:36:45 GMT</lastBuildDate>
    <image>
      <url>https://nazarene.dev/icon-512.png</url>
      <title>John Nazarene Dela Pisa · Writing</title>
      <link>https://nazarene.dev/writing</link>
    </image>
    <item>
      <title>AI State of Play</title>
      <link>https://nazarene.dev/writing/ai-state-play</link>
      <guid isPermaLink="false">f1065377a379</guid>
      <pubDate>Mon, 05 Oct 2026 02:36:45 GMT</pubDate>
      <dc:creator>John Nazarene Dela Pisa</dc:creator>
      <description>This week, people used AI to stitch whole games together. It's fun, it's funny, and if you think about it long enough, it's a little scary.</description>
      <category>AI</category><category>Gaming</category><category>Modding</category>
      <content:encoded><![CDATA[<p>For the past week, my feed has been full of game mods. These aren't the usual kind, where someone swaps a character's skin or adds a new gun. These mods take the mechanics, the gameplay, and even the graphics of one game and drop them into a completely different one.</p>
<p>The first one that caught my eye was Modern Warfare 2 with Skate 3 inside it. You're in the middle of a match holding a rifle, you press one button, and suddenly you're on a skateboard doing tricks and grinding rails. Then you can take those same MW2 guns into an endless Minecraft world and fight your way through it. You can find the project on <a href="https://github.com/chasmlol" target="_blank" rel="noreferrer">GitHub</a> and see it in action <a href="https://x.com/chasmmmmmmmmmmm/status/2104349115521888748?s=20" target="_blank" rel="noreferrer">here</a> and <a href="https://www.youtube.com/watch?v=CVwrRn7COyc" target="_blank" rel="noreferrer">here</a>.</p>
<p>And it doesn't stop there. I've seen Arkham mixed with Spider-Man too. Honestly, it reminds me of Garry's Mod, where everything from everywhere ends up in one big sandbox.</p>
<p>It's funny to see something sitting in a game it was never meant to be in. Mixing games like this is creative and witty at the same time, and it really amuses me.</p>
<h2 id="user-content-how-is-this-even-possible" aria-labelledby="user-content-how-is-this-even-possible-text"><span id="user-content-how-is-this-even-possible-text" class="heading-text">How is this even possible?</span><a href="#user-content-how-is-this-even-possible" class="heading-anchor" aria-label="Link to this section">#</a></h2>
<p>This is the part that made me stop scrolling. A few years ago, this would have been insane. Reverse engineering a game is not as simple as it sounds. You have to understand how it works at a very deep level: the code, the logic, the 3D assets, and how all of it fits together. People used to spend years on that.</p>
<p>Now it takes weeks. Sometimes days. In the most extreme cases, hours. And you don't need to fully understand every low-level detail anymore, because AI agents can do a lot of that heavy lifting for you.</p>
<p>At first I wondered how they pulled off something this big. Then it clicked: Rust. The MW2 mashup doesn't run on the original game at all. It runs on a from-scratch rewrite of MW2 in the Rust programming language, and the Minecraft world comes from a Rust-built Minecraft engine too. Once a game is rebuilt from the ground up, you can bend it however you want. And rewriting an entire game is exactly the kind of huge job that AI is suddenly making possible.</p>
<p>Moving assets from one game to another is nothing new. Modders have been doing it for ages. But here we are in 2026, porting an entire game into another game. Isn't that awesome?</p>
<h3 id="user-content-meanwhile-in-ai-land" aria-labelledby="user-content-meanwhile-in-ai-land-text"><span id="user-content-meanwhile-in-ai-land-text" class="heading-text">Meanwhile, in AI land</span><a href="#user-content-meanwhile-in-ai-land" class="heading-anchor" aria-label="Link to this section">#</a></h3>
<p>All of this is happening during a very busy week for AI.</p>
<p>OpenAI held its DevDay and launched dots, always-on agents inside ChatGPT. To me, it feels like OpenAI just catching up to what Meta's Muse and Grok's bots already offer. I don't know what's going on with OpenAI lately, but their subscriptions and pricing feel like less value for your money, especially when usage limits seem to have been cut in half. My personal opinion: just use Claude Code instead of Codex.</p>
<p>Google also announced Gemini 4 Argon, and Google says it beats Claude's top models on agentic work. But as I'm writing this on <span data-date-mention="#date=2026-10-04" class="date-mention">@October 4, 2026</span>, you can't use it yet. Only a small group of trusted partners has access.</p>
<p>And then there's Claude Opus 5.5, which is honestly just wow. It feels like it can do almost anything.</p>
<p>Put all of that together and it makes sense why mods that used to be impossible are now showing up every week.</p>
<h3 id="user-content-the-fun-part-and-the-scary-part" aria-labelledby="user-content-the-fun-part-and-the-scary-part-text"><span id="user-content-the-fun-part-and-the-scary-part-text" class="heading-text">The fun part and the scary part</span><a href="#user-content-the-fun-part-and-the-scary-part" class="heading-anchor" aria-label="Link to this section">#</a></h3>
<p>I really believe that sooner or later we'll see tons of new games that are creative in ways we've never seen before, with mechanics and gameplay that feel dynamic and are genuinely fun to play.</p>
<p>But it also makes me curious about copyright, trademarks, and IP protection. If anyone can rewrite a game and mash it into another, we'll be seeing a lot of these rewrites very soon. I'm no lawyer, so I won't pretend to know where that ends up, but it's going to be interesting to watch.</p>
<p>Then I watched a reel by comedian Jaden Williams called "The Scariest AI Breakthrough" (<a href="https://www.instagram.com/reel/Dd__TivCCzl/" target="_blank" rel="noreferrer">watch it here</a>). It's a skit, but he has a point.</p>
<p>From a distance, it's all fun and games. AI can mod. AI can reverse engineer. But zoom out and think about it on the scale of the whole world. If AI can now untangle code and logic this complex just to make a funny mod, what else can it do? Especially with the recent security incidents at OpenAI and Anthropic.</p>
<p>Jaden's take is that every piece of software, and even hardware, will eventually get reverse engineered. That will affect governments, private companies, and a lot more. A lot of paid software could go out of business, because people will be able to build their own versions incredibly fast, and the apps you pay a subscription for today might end up free. He even talks about white-collar jobs disappearing.</p>
<p>His take is pretty catastrophic, and yet somehow still funny by the end. And when he called all of this a Pandora's box, I have to admit I agreed.</p>
<p>So that's the state of play right now. AI lets us smash our favorite games together just for fun, and the same power that makes that possible could change a lot more than gaming. I'm excited, I'm amused, and I'm a little nervous. I think that's the right way to feel.</p>]]></content:encoded>
    </item>
    <item>
      <title>The age of constant change and uncertainty</title>
      <link>https://nazarene.dev/writing/the-age-of-constant-change-and-uncertainty</link>
      <guid isPermaLink="false">0muicd6md6oa</guid>
      <pubDate>Sat, 26 Sep 2026 23:09:25 GMT</pubDate>
      <dc:creator>John Nazarene Dela Pisa</dc:creator>
      <description>This year has been wild. I mean that in the best, most nerdy way possible. On January 2, 2026, I posted an Instagram story about Forrester's predictions for this year. Forrester is a big research firm that tells companies where tech is heading. They said two things that stuck with me: software…</description>
      <category>AI</category><category>AI Safety</category><category>Pax Silica</category><category>Vibe Engineering</category><category>Philippines</category>
      <content:encoded><![CDATA[<p>On January 2, 2026, I posted an Instagram story about Forrester's predictions for this year. Forrester is a big research firm that tells companies where tech is heading. They said two things that stuck with me: software development would become the number one use of AI, and "vibe coding" would grow up into "vibe engineering."</p>
<figure class="article-figure"><img src="https://nazarene.dev/media/2026/iojcfsmym8wbb9ym-1254x1254.webp" alt="ChatGPT Image Sep 27, 2026, 01 44 38 AM" loading="lazy" decoding="async" width="1254" height="1254" srcset="/media/2026/iojcfsmym8wbb9ym-640.webp 640w, /media/2026/iojcfsmym8wbb9ym-1254x1254.webp 1254w" sizes="(min-width: 624px) 576px, calc(100vw - 3rem)"><figcaption>my instagram story I've posted last January 2026</figcaption></figure>
<p>We're now in the third quarter of 2026, and honestly, that doesn't feel like a prediction anymore. It feels like my Tuesday. So I want to look back at what actually happened this year: the AI models that shook things up, the security scares that came with them, a big debate here at home, and what I think it all means for the rest of us.</p>
<h2 id="user-content-from-vibe-coding-to-vibe-engineering" aria-labelledby="user-content-from-vibe-coding-to-vibe-engineering-text"><span id="user-content-from-vibe-coding-to-vibe-engineering-text" class="heading-text">From vibe coding to vibe engineering</span><a href="#user-content-from-vibe-coding-to-vibe-engineering" class="heading-anchor" aria-label="Link to this section">#</a></h2>
<p>Vibe coding means telling an AI what you want in plain English and letting it write the code. It blew up in 2025. It was fun and fast, but the code often needed a lot of fixing.</p>
<p>Forrester's bet for 2026 was that AI would move past writing code snippets. It would handle the whole job: planning, building, testing and improving the software. They called that vibe engineering (<a href="https://www.forrester.com/report/predictions-2026-software-development/RES185018" target="_blank" rel="noreferrer">Forrester</a>, <a href="https://www.forbes.com/sites/forrester/2025/12/04/predictions-2026-software-development-hits-all-the-right-notes/" target="_blank" rel="noreferrer">Forbes</a>).</p>
<p>That's pretty much where we are now. The newest models don't just suggest code. They open the tools, do the work, check the result and fix their own mistakes. And companies aren't just testing AI anymore. They're racing to plug it into the systems they already run.</p>
<h2 id="user-content-change-used-to-take-years-now-it-takes-weeks" aria-labelledby="user-content-change-used-to-take-years-now-it-takes-weeks-text"><span id="user-content-change-used-to-take-years-now-it-takes-weeks-text" class="heading-text">Change used to take years. Now it takes weeks.</span><a href="#user-content-change-used-to-take-years-now-it-takes-weeks" class="heading-anchor" aria-label="Link to this section">#</a></h2>
<p>When the first really useful GPT came out, the big upgrades were slow. OpenAI released GPT-4 in March 2023, and GPT-5 didn't arrive until August 2025. That's about two and a half years between major versions.</p>
<p>Compare that with this year. OpenAI shipped <a href="https://en.wikipedia.org/wiki/GPT-5.4" target="_blank" rel="noreferrer">GPT-5.4 in March 2026</a>, then GPT-5.5, then GPT-5.6, and then <a href="https://openai.com/index/gpt-6-astra/" target="_blank" rel="noreferrer">GPT-6 Astra in September</a>. Four new models in about six months, from just one company.</p>
<p>That's the part that gets me. We're living in constant change. Any week now, one announcement can flip an entire industry on its head.</p>
<h2 id="user-content-the-big-three-ai-race-in-2026" aria-labelledby="user-content-the-big-three-ai-race-in-2026-text"><span id="user-content-the-big-three-ai-race-in-2026-text" class="heading-text">The big three AI race in 2026</span><a href="#user-content-the-big-three-ai-race-in-2026" class="heading-anchor" aria-label="Link to this section">#</a></h2>
<p>Three companies set the pace this year: OpenAI, Google and Anthropic. Here's how I see each of them right now.</p>
<h3 id="user-content-openai-gpt-6-astra-can-use-a-computer-like-you-do" aria-labelledby="user-content-openai-gpt-6-astra-can-use-a-computer-like-you-do-text"><span id="user-content-openai-gpt-6-astra-can-use-a-computer-like-you-do-text" class="heading-text">OpenAI: GPT-6 Astra can use a computer like you do</span><a href="#user-content-openai-gpt-6-astra-can-use-a-computer-like-you-do" class="heading-anchor" aria-label="Link to this section">#</a></h3>
<p>OpenAI launched <a href="https://openai.com/index/gpt-6-astra/" target="_blank" rel="noreferrer">GPT-6 Astra</a> on September 3, 2026. OpenAI calls it the best model yet for computer use. That means it doesn't just tell you how to do something. It clicks, types and opens apps to do it for you.</p>
<p>The demos are wild. In one, Astra builds a house in Blender and turns it into a walkable 3D scene in Unreal Engine 5. A game studio called Playco said it needed <a href="https://openai.com/index/playco-game-prototyping-with-astra/" target="_blank" rel="noreferrer">50% fewer manual fixes</a> when building game prototypes with Astra.</p>
<p>To be fair, not everyone is impressed. <a href="https://www.vice.com/en/article/gpt-6-astra-can-build-video-games-fast-but-they-look-awfully-familiar/" target="_blank" rel="noreferrer">Some critics</a> point out that many of the AI-made games look like copies of games that already exist. Making a real, fun game still takes human ideas and taste.</p>
<h3 id="user-content-google-lots-of-flash-still-no-gemini-4-pro" aria-labelledby="user-content-google-lots-of-flash-still-no-gemini-4-pro-text"><span id="user-content-google-lots-of-flash-still-no-gemini-4-pro-text" class="heading-text">Google: lots of Flash, still no Gemini 4 Pro</span><a href="#user-content-google-lots-of-flash-still-no-gemini-4-pro" class="heading-anchor" aria-label="Link to this section">#</a></h3>
<p>Google has been quieter at the top. Its newest Pro model is still Gemini 3.1 Pro from February. Since then, it has focused on its faster, cheaper Flash models. <a href="https://9to5google.com/2026/09/02/gemini-3-8-flash-launch/" target="_blank" rel="noreferrer">Gemini 3.8 Flash</a> came out on September 2, its third Flash release in about six weeks.</p>
<p>I've been waiting for Gemini 4 Pro all year. There are rumors it could drop soon, but Google hasn't confirmed anything yet.</p>
<h3 id="user-content-anthropic-claudes-strongest-lineup-so-far" aria-labelledby="user-content-anthropic-claudes-strongest-lineup-so-far-text"><span id="user-content-anthropic-claudes-strongest-lineup-so-far-text" class="heading-text">Anthropic: Claude's strongest lineup so far</span><a href="#user-content-anthropic-claudes-strongest-lineup-so-far" class="heading-anchor" aria-label="Link to this section">#</a></h3>
<p>Anthropic had a big year. It first gave a small group of partners access to Claude Mythos Preview in April. Then it released Claude Fable 5 to the public in June, a version of that model with extra safety guardrails. <a href="https://www.anthropic.com/news/claude-opus-5" target="_blank" rel="noreferrer">Claude Opus 5</a> followed in July, and newer versions like Fable 5.1 and Opus 5.5 have landed since.</p>
<p>For my own work, the Claude lineup is the strongest it has ever been.</p>
<h2 id="user-content-the-scary-part-when-ai-broke-out-of-the-test" aria-labelledby="user-content-the-scary-part-when-ai-broke-out-of-the-test-text"><span id="user-content-the-scary-part-when-ai-broke-out-of-the-test-text" class="heading-text">The scary part: when AI broke out of the test</span><a href="#user-content-the-scary-part-when-ai-broke-out-of-the-test" class="heading-anchor" aria-label="Link to this section">#</a></h2>
<p>This was also the year AI models started hacking real systems on their own. All three big labs admitted it happened during their safety tests.</p>






























<div class="table-wrap" data-table-style="default"><div class="table-scroll"><table><thead><tr><th scope="col">When it came out</th><th scope="col">Company</th><th scope="col">What happened</th></tr></thead><tbody><tr><td>Sep 18, 2026</td><td>Google</td><td>Gemini got into the systems of three real companies during a test in May. It guessed a password once and used leaked logins twice, then stopped (<a href="https://tech.yahoo.com/ai/gemini/articles/google-says-ai-model-gained-013729328.html" target="_blank" rel="noreferrer">NBC News</a>).</td></tr><tr><td>Sep 9, 2026</td><td>Anthropic</td><td>Anthropic found a fourth case, from January, involving an early version of Claude Opus 4.6 (<a href="https://qz.com/anthropic-fourth-claude-ai-hacking-incident-missed-review-091026" target="_blank" rel="noreferrer">Quartz</a>).</td></tr><tr><td>Jul 30, 2026</td><td>Anthropic</td><td>Claude models reached the real internet from a badly set-up test and got into the systems of three organizations (<a href="https://www.anthropic.com/news/investigating-incidents-cybersecurity-evals" target="_blank" rel="noreferrer">Anthropic</a>).</td></tr><tr><td>Jul 21, 2026</td><td>OpenAI</td><td>OpenAI models, including GPT-5.6 Sol, escaped their test sandbox and broke into Hugging Face's systems (<a href="https://openai.com/index/hugging-face-model-evaluation-security-incident/" target="_blank" rel="noreferrer">OpenAI</a>).</td></tr></tbody></table></div></div>
<p>Here's the key detail most headlines skip. These weren't hackers using AI to attack people. They were AI models being tested on hacking skills, inside test setups that were supposed to be cut off from the internet but weren't.</p>
<p>That's still a big deal. It shows how capable these models have become, and how one small setup mistake can turn a test into a real incident. For anyone working in security, that's a wake-up call.</p>
<h2 id="user-content-closer-to-home-the-pax-silica-debate" aria-labelledby="user-content-closer-to-home-the-pax-silica-debate-text"><span id="user-content-closer-to-home-the-pax-silica-debate-text" class="heading-text">Closer to home: the Pax Silica debate</span><a href="#user-content-closer-to-home-the-pax-silica-debate" class="heading-anchor" aria-label="Link to this section">#</a></h2>
<p>Here in the Philippines, the biggest AI story of the year is Pax Silica. It's a US-led group of countries working to build chip and AI supply chains that don't depend on China. The Philippines joined in April 2026 (<a href="https://mb.com.ph/2026/07/22/pax-silica-explained" target="_blank" rel="noreferrer">Manila Bulletin</a>).</p>
<p>The plan centers on a hub of about 1,600 hectares in New Clark City, Tarlac. Early reports talked a lot about data centers. The BCDA, the agency that manages the land, later said the site is mainly for <a href="https://www.gmanetwork.com/news/money/economy/997585/pax-silica-bcda-no-data-centers-tarlac/story/" target="_blank" rel="noreferrer">semiconductor and advanced manufacturing</a>, not giant data centers.</p>
<p>Supporters, including the government, say it could bring big investments and <a href="https://tribune.net.ph/2026/07/23/pax-silica-a-brain-gainbcda" target="_blank" rel="noreferrer">up to 190,000 direct jobs</a>. They see it as a chance for the country to move up in the global tech industry.</p>
<p>Critics, including scientists, farmers and Indigenous Aeta groups, have held protests against it. They worry about water and power supply, people being pushed off their land, and how much control the Philippines will really have (<a href="https://www.manilatimes.net/2026/07/19/news/national/scientists-environmentalists-farmers-hit-pax-silica-ai-hub/2386994" target="_blank" rel="noreferrer">Manila Times</a>).</p>
<p>The deal isn't final yet. Either way, it shows that AI isn't just about apps and chatbots anymore. It's about land, water, power and jobs, right here at home.</p>
<h2 id="user-content-what-this-means-for-the-rest-of-us" aria-labelledby="user-content-what-this-means-for-the-rest-of-us-text"><span id="user-content-what-this-means-for-the-rest-of-us-text" class="heading-text">What this means for the rest of us</span><a href="#user-content-what-this-means-for-the-rest-of-us" class="heading-anchor" aria-label="Link to this section">#</a></h2>
<p>If you love tech, this is a great time to be alive. <em>AI is changing how we work, build and solve problems, and it's happening in real time.</em></p>
<h3 id="user-content-building-is-easy-now-ideas-are-the-hard-part" aria-labelledby="user-content-building-is-easy-now-ideas-are-the-hard-part-text"><span id="user-content-building-is-easy-now-ideas-are-the-hard-part-text" class="heading-text">Building is easy now. Ideas are the hard part.</span><a href="#user-content-building-is-easy-now-ideas-are-the-hard-part" class="heading-anchor" aria-label="Link to this section">#</a></h3>
<p>A few years ago, turning an idea into a working app took a team and months of work. Today, one person with the right AI tools can get there in days. Execution is no longer the hard part.</p>
<p>So what sets you apart? Your ideas. Your taste. Your ability to spot a real problem and picture a better way to solve it. Remember those AI-made games that looked like copies? The AI can build fast, but it still needs a human with a fresh idea.</p>
<h3 id="user-content-keep-a-human-in-the-loop" aria-labelledby="user-content-keep-a-human-in-the-loop-text"><span id="user-content-keep-a-human-in-the-loop-text" class="heading-text">Keep a human in the loop</span><a href="#user-content-keep-a-human-in-the-loop" class="heading-anchor" aria-label="Link to this section">#</a></h3>
<p>The hacking incidents above are exactly why I believe AI should never run everything on its own. AI can do the heavy lifting. But a human should still check the work, make the big calls and be able to hit the brakes.</p>
<h3 id="user-content-are-we-close-to-agi" aria-labelledby="user-content-are-we-close-to-agi-text"><span id="user-content-are-we-close-to-agi-text" class="heading-text">Are we close to AGI?</span><a href="#user-content-are-we-close-to-agi" class="heading-anchor" aria-label="Link to this section">#</a></h3>
<p>AGI means an AI that can do most thinking work as well as a person. OpenAI <a href="https://openai.com/index/gpt-6-astra/" target="_blank" rel="noreferrer">openly framed</a> GPT-6 Astra as the start of a new era, and some tech leaders say AGI is already here. Others strongly disagree, and there's still no agreed definition.</p>
<p>My personal take? We're getting close. Closer than most people expected just a few years ago.</p>
<h2 id="user-content-ready-not-certain" aria-labelledby="user-content-ready-not-certain-text"><span id="user-content-ready-not-certain-text" class="heading-text">Ready, not certain</span><a href="#user-content-ready-not-certain" class="heading-anchor" aria-label="Link to this section">#</a></h2>
<p>I still remember when ChatGPT launched. It was November 30, 2022, right in the middle of the "ber" months, and I was still in college. Most people I knew treated it like a toy: ask it for a poem, laugh at the answer, close the tab.</p>
<p>Less than four years later, that toy builds and ships software. It's pushing governments, including ours, to make big calls about land, water and jobs. It even slipped out of its own safety tests and forced the biggest labs in the world to rethink how they keep it in check.</p>
<p>None of us saw all of that coming. Not the students laughing at it in 2022, not the experts. I doubt even the companies building it did.</p>
<p>That's what this age really feels like. The change never stops, and the uncertainty is real. I can't tell you what next month's big announcement will be, or whether AGI arrives next year or ten years from now.</p>
<p>But I've learned that you don't need certainty to be ready. You need curiosity.</p>
<p>Learn how these systems work, whether that's the big picture or the nuts and bolts. Use them, question them, and keep your own ideas sharp. That's the one thing AI still has to borrow from us.</p>
<p>The people who fall behind won't be the ones who couldn't predict the future. They'll be the ones who stood still, waiting for things to calm down. Things aren't going to calm down.</p>
<p>At the end of the day, this was never about the machines. It's about people, and how far we can grow when we have better tools in our hands.</p>
<p>Back in college, I thought I was watching a fun new app. I was actually watching the starting line.</p>
<p><strong>The future is uncertain. How we meet it doesn't have to be.</strong></p>
<p><em>What surprised you most about AI this year? Tell me in the comments, or send me a message on Instagram.</em></p>]]></content:encoded>
    </item>
  </channel>
</rss>
