<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:content="http://purl.org/rss/1.0/modules/content/"><channel><title>Ai on Hanh Pham's Blog</title><link>https://hanhpham.vercel.app/tags/ai/</link><description>Recent content in Ai on Hanh Pham's Blog</description><generator>Hugo</generator><language>en-us</language><lastBuildDate>Fri, 02 Oct 2026 10:00:00 +0700</lastBuildDate><atom:link href="https://hanhpham.vercel.app/tags/ai/index.xml" rel="self" type="application/rss+xml"/><item><title>The Renaissance Developer: Why AI Makes Systems Thinking the Ultimate Moat</title><link>https://hanhpham.vercel.app/posts/the-renaissance-developer/</link><pubDate>Fri, 02 Oct 2026 10:00:00 +0700</pubDate><guid>https://hanhpham.vercel.app/posts/the-renaissance-developer/</guid><description>When generative AI commoditizes syntax into zero-cost token emissions, the developer&amp;rsquo;s moat shifts from typing code to systems synthesis: understanding how a single concurrent change ripples across database row locks, cache lines, and cloud bills.</description></item><item><title>Deleting Language from the LLM: Inside TypeSafe's Jev and Decision Models</title><link>https://hanhpham.vercel.app/posts/typesafe-jev-decision-models/</link><pubDate>Tue, 22 Sep 2026 09:00:00 +0700</pubDate><guid>https://hanhpham.vercel.app/posts/typesafe-jev-decision-models/</guid><description>When you force a 70-billion-parameter language model to output &amp;rsquo;true&amp;rsquo; or &amp;lsquo;false&amp;rsquo;, you pay for sequential token decoding, uncalibrated probabilities, and broken JSON. Here is why TypeSafe stripped natural language out of the loop.</description></item><item><title>The Four-Layer AI Coding Stack: Why Your Assistant Fails (and It's Not the Model)</title><link>https://hanhpham.vercel.app/posts/the-four-layer-ai-coding-stack/</link><pubDate>Sat, 19 Sep 2026 09:30:00 +0700</pubDate><guid>https://hanhpham.vercel.app/posts/the-four-layer-ai-coding-stack/</guid><description>Most developer frustration with AI coding assistants comes from conflating the editor, the agent loop, the gateway, and the model. Here is how the four layers actually fit together, where failures originate, and how to build a setup that survives production work.</description></item><item><title>Apple M4 vs Huawei Ascend 950: Two Theories of the AI Chip</title><link>https://hanhpham.vercel.app/posts/m4-vs-ascend-950/</link><pubDate>Wed, 09 Sep 2026 09:00:00 +0700</pubDate><guid>https://hanhpham.vercel.app/posts/m4-vs-ascend-950/</guid><description>One is a fixed-function FP16 block hidden inside a laptop chip Apple won&amp;rsquo;t document. The other is a programmable FP4 datacenter engine Huawei ships in 8,192-chip fabrics. Same goal, opposite theses: here&amp;rsquo;s what the M4 family actually is, and what the numbers do and don&amp;rsquo;t mean against the Ascend 950.</description></item><item><title>Huawei's Ascend 950 &amp; DaVinci v3: How China Built a Competitive AI Chip Without EUV</title><link>https://hanhpham.vercel.app/posts/huawei-ascend-950-davinci-v3/</link><pubDate>Mon, 07 Sep 2026 09:00:00 +0700</pubDate><guid>https://hanhpham.vercel.app/posts/huawei-ascend-950-davinci-v3/</guid><description>DeepSeek just ordered 160,000 Huawei Ascend 950DT chips: a chip built on a DUV-only process, with self-developed HBM and a homegrown interconnect, because export controls left no other choice. Here&amp;rsquo;s how the thing actually works, and where it still falls short.</description></item><item><title>Agentic Coding Assistants, and Why I Don't Let Them Grade Their Own Work</title><link>https://hanhpham.vercel.app/posts/how-ai-actually-fits-into-my-dev-workflow/</link><pubDate>Wed, 02 Sep 2026 10:00:00 +0700</pubDate><guid>https://hanhpham.vercel.app/posts/how-ai-actually-fits-into-my-dev-workflow/</guid><description>&amp;ldquo;Agentic&amp;rdquo; gets thrown around loosely, but it means something specific: an assistant that decides its own next step instead of waiting for yours. That&amp;rsquo;s also exactly why you can&amp;rsquo;t trust it to tell you when it&amp;rsquo;s done a good job; here&amp;rsquo;s the verification habit that fixes that.</description></item></channel></rss>