<rss version="2.0" xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Hacker News: cagz</title><link>https://news.ycombinator.com/user?id=cagz</link><description>Hacker News RSS</description><docs>https://hnrss.org/</docs><generator>hnrss v2.1.1</generator><lastBuildDate>Wed, 18 Nov 2026 14:17:26 +0000</lastBuildDate><atom:link href="https://hnrss.org/user?id=cagz" rel="self" type="application/rss+xml"></atom:link><item><title><![CDATA[New comment by cagz in "AI usage generates a fragile codebase"]]></title><description><![CDATA[
<p>I understand the concern fully. I believe this is mainly because software development is currently in transition.<p>There was a time when all tables were made by hand. Carpentry was often seen as an art form. Every table was unique and carried something special from its creator. Then the industrialisation came; factories started making tables. The first ones had issues, but over time, they started producing high-quality tables at high volume. Today, "hand-made " still has a special place in our hearts, but it is usually seen as a niche/luxury product. When a company wants to buy tables or desks for an office, they rarely go to a carpenter to get them made.<p>I am sure master carpenters found (and still find) factory-made tables inferior, and perhaps they are, if we are looking for a special human touch. Unfortunately, the free economy demands cheaper, factory-made tables, so they are very popular.<p>Software development is just in the middle of this transition, and it is painful. It will get better soon, but the bitter taste will linger. There will definitely be niche development tasks where we ask master developers to get the job done manually. Software for flight control systems comes to mind, where every low-level instruction matters.<p>What do we do now? Given that the trend and demand are obvious, we can focus our efforts on making the factories better. If I may reference a couple of points from the post:<p>AI makes shallow tests:<p>The opportunity here is improving test-writing skills (AI skills) to ensure tests written have enough depth.<p>AI writes overly complex code<p>I found that TDD helps greatly here. Write failing tests first, then write the minimum amount of code to make them pass.<p>For me, these are how we can make the factory better. I don't think the two organisations mentioned that expect code to be developed 80 or 100%, respectively, would mind developers spending some of their time improving the team's shared skills library.<p>Perhaps it helps to see this not as "stopping writing code", but as "going one more level up in the abstraction hierarchy". It's similar to going from assembly to higher-level languages, although I appreciate it is not a perfect analogy.</p>
]]></description><pubDate>Tue, 10 Nov 2026 19:05:03 +0000</pubDate><link>https://news.ycombinator.com/item?id=50444706</link><dc:creator>cagz</dc:creator><comments>https://news.ycombinator.com/item?id=50444706</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=50444706</guid></item><item><title><![CDATA[New comment by cagz in "Five memorable books about programming (2008)"]]></title><description><![CDATA[
<p>May I add:<p>Uncle Bob's Clean Code and<p>Denis Ritchie's C programming language</p>
]]></description><pubDate>Sun, 08 Nov 2026 07:39:18 +0000</pubDate><link>https://news.ycombinator.com/item?id=50411185</link><dc:creator>cagz</dc:creator><comments>https://news.ycombinator.com/item?id=50411185</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=50411185</guid></item><item><title><![CDATA[New comment by cagz in "There's never been a better time to go into math (by: 3blue1brown) [video]"]]></title><description><![CDATA[
<p>Exactly. As a software developer, AI has given me more time to tackle higher-order problems. The actual architecture, optimisation, and user experience.<p>Mathematicians could:<p>- Continue working on philosophical aspects of the solution and not bothered by whether the solution is known or not. It is not a movie that gets spoiled when the ending is known.<p>- Or better, demand from AI labs, how they would like to see AI helping them on philosophical aspects.</p>
]]></description><pubDate>Mon, 19 Oct 2026 04:22:25 +0000</pubDate><link>https://news.ycombinator.com/item?id=50148343</link><dc:creator>cagz</dc:creator><comments>https://news.ycombinator.com/item?id=50148343</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=50148343</guid></item><item><title><![CDATA[KiDex – component library manager for KiCad]]></title><description><![CDATA[
<p>Article URL: <a href="https://kidex.io/en/">https://kidex.io/en/</a></p>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=49876985">https://news.ycombinator.com/item?id=49876985</a></p>
<p>Points: 1</p>
<p># Comments: 0</p>
]]></description><pubDate>Mon, 28 Sep 2026 12:34:36 +0000</pubDate><link>https://kidex.io/en/</link><dc:creator>cagz</dc:creator><comments>https://news.ycombinator.com/item?id=49876985</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49876985</guid></item><item><title><![CDATA[Google Private AI Compute with server-side memory]]></title><description><![CDATA[
<p>Article URL: <a href="https://deepmind.google/blog/advancing-private-ai-compute-with-secure-server-side-memory/">https://deepmind.google/blog/advancing-private-ai-compute-with-secure-server-side-memory/</a></p>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=49827294">https://news.ycombinator.com/item?id=49827294</a></p>
<p>Points: 3</p>
<p># Comments: 0</p>
]]></description><pubDate>Thu, 24 Sep 2026 07:12:37 +0000</pubDate><link>https://deepmind.google/blog/advancing-private-ai-compute-with-secure-server-side-memory/</link><dc:creator>cagz</dc:creator><comments>https://news.ycombinator.com/item?id=49827294</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49827294</guid></item><item><title><![CDATA[New comment by cagz in "The LLMentalist Effect (2023)"]]></title><description><![CDATA[
<p>The article starts interestingly; I like the analogy between cold reading and LLM responses (well, the ones we were getting three years ago). Then it goes on to generalise to the tune of "This new era of tech seems to be built on superstition and pseudoscience" which I think was a conclusion better left to the reader.<p>We are used to articles, blogs that stayed relevant for years. It is hardly the case with AI, the speed of change is so high that an observation can become stale in a year. This makes it very important for articles to state that the ideas are based on the technology in year X, rather than generalising them as universal truths.</p>
]]></description><pubDate>Mon, 21 Sep 2026 06:00:33 +0000</pubDate><link>https://news.ycombinator.com/item?id=49783518</link><dc:creator>cagz</dc:creator><comments>https://news.ycombinator.com/item?id=49783518</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49783518</guid></item><item><title><![CDATA[New comment by cagz in "MCP was always a bad idea?"]]></title><description><![CDATA[
<p>The title feels clickbaity. Direct API, CLI and MCP all have their uses.<p>Direct access (Curl/own small function): This requires agent to have full understanding of the API spec. Yes, context can be protected using progressive disclosure, but this essentially means agent needing to understand the API again and again, before every use. Also, a typical API spec may or may not be agent-friendly. If there are nuances when calling an endpoint, where do we put these? Into OAS description? Works, but clunky.<p>CLI: It works beautifully, especially when a 3rd party CLI already exists for a complex backend. Assumes a well documented, agent friendly CLI, most CLIs are designed for human or CI/CD consumption. Talking about MCP taking up too much context, think about agent starting with my_cli --help, and going down through the switches and parameters one at a time to figure out how the CLI should be called. Less of a problem when calling a well know CLI (e.g. aws), but anything more niche (or custom) requires multiple turns to compose the final CLI command.<p>MCP: Has its issues, but offers an agent-native solution. Everything agent needs to know about a tool becomes available at once. In an enterprise environment MCPs can be served through an MCP Gateway, providing governance and permission management, this is quite contrast against running a CLI that requires agent to have execute permissions in its shell.</p>
]]></description><pubDate>Mon, 21 Sep 2026 05:36:19 +0000</pubDate><link>https://news.ycombinator.com/item?id=49783404</link><dc:creator>cagz</dc:creator><comments>https://news.ycombinator.com/item?id=49783404</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49783404</guid></item><item><title><![CDATA[Jev vs. XGBoost and Bert]]></title><description><![CDATA[
<p>Article URL: <a href="https://explainx.ai/blog/jev-vs-xgboost-bert-classifiers-2026">https://explainx.ai/blog/jev-vs-xgboost-bert-classifiers-2026</a></p>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=49768899">https://news.ycombinator.com/item?id=49768899</a></p>
<p>Points: 4</p>
<p># Comments: 0</p>
]]></description><pubDate>Sat, 19 Sep 2026 18:18:39 +0000</pubDate><link>https://explainx.ai/blog/jev-vs-xgboost-bert-classifiers-2026</link><dc:creator>cagz</dc:creator><comments>https://news.ycombinator.com/item?id=49768899</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49768899</guid></item><item><title><![CDATA[New comment by cagz in "A Misalignment of AI and Cancer Biology"]]></title><description><![CDATA[
<p>A very poor, word-by-word copy of <a href="https://mathandai.org" rel="nofollow">https://mathandai.org</a></p>
]]></description><pubDate>Sat, 12 Sep 2026 16:15:09 +0000</pubDate><link>https://news.ycombinator.com/item?id=49673917</link><dc:creator>cagz</dc:creator><comments>https://news.ycombinator.com/item?id=49673917</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49673917</guid></item><item><title><![CDATA[New comment by cagz in "A misalignment of AI in mathematics"]]></title><description><![CDATA[
<p>I wish this were a document that clearly described how mathematicians expect AI to behave, rather than how it is misaligned. I can kind of decipher the expectation from the declaration, but it still sounds a bit like "it would be better if AI didn't meddle with mathematics", instead of "we want AI to approach problems like this and that way".</p>
]]></description><pubDate>Sat, 12 Sep 2026 16:08:47 +0000</pubDate><link>https://news.ycombinator.com/item?id=49673836</link><dc:creator>cagz</dc:creator><comments>https://news.ycombinator.com/item?id=49673836</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49673836</guid></item><item><title><![CDATA[New comment by cagz in "Ask HN: Who is using MCP in production?"]]></title><description><![CDATA[
<p>Yes, quite a few uses in prod, main use case being abstracting API access for agents. In our case these are mostly in-house MCP servers, purpose built for the given agent.<p>Why MCP instead of CLI or agent accessing API directly?<p>Direct access (Curl/own small function): This requires agent to have full understanding of the API spec. Yes, context can be protected using progressive disclosure, but this essentially means agent needing to understand the API again and again, before every use in that context. Also, a typical API spec may or may not be agent-friendly. If there are nuances when calling an endpoint, where do we put these? Into OAS description? Works, but clunky.<p>CLI: It works beautifully, especially when a 3rd party CLI already exists for a complex backend. Assumes a well documented, agent friendly CLI, most CLIs are designed for human or CI/CD consumption. Talking about MCP taking up too much context, think about agent starting with my_cli --help, and going down through the switches and parameters one at a time to figure out how the CLI should be called. Less of a problem when calling a well know CLI (e.g. aws), but anything more niche (or custom) requires multiple turns to compose the final CLI command.<p>MCP: Has its issues, but offers an agent-native solution. Everything agent needs to know about a tool becomes available at once. In an enterprise environment MCPs can be served through an MCP Gateway, providing governance and permission management, this is quite contrast against running a CLI that requires agent to have execute permissions in its shell.<p>I must mention that we also utilise lazy-loading of MCPs. In use cases where 10s of tools needs to be loaded, only the most common ones are pre-loaded, then agent kernel connects the others as an when needed, and release them after a timeout (in case of long running sessions). This keeps the context lean.</p>
]]></description><pubDate>Fri, 04 Sep 2026 06:17:59 +0000</pubDate><link>https://news.ycombinator.com/item?id=49561135</link><dc:creator>cagz</dc:creator><comments>https://news.ycombinator.com/item?id=49561135</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49561135</guid></item><item><title><![CDATA[New comment by cagz in "London Underground begins scanning passengers' faces"]]></title><description><![CDATA[
<p>It is BTP, not London Underground, they are not the same thing.</p>
]]></description><pubDate>Wed, 12 Aug 2026 07:11:36 +0000</pubDate><link>https://news.ycombinator.com/item?id=49268832</link><dc:creator>cagz</dc:creator><comments>https://news.ycombinator.com/item?id=49268832</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49268832</guid></item><item><title><![CDATA[New comment by cagz in "GitHub Actions and Pages are experiencing degraded availability"]]></title><description><![CDATA[
<p>Outages happen. What concerned me most was the time taken for the full resolution (almost 12 hours). I am hoping that root cause analysis promised will look into this as well and how it can be improved.</p>
]]></description><pubDate>Fri, 07 Aug 2026 06:12:21 +0000</pubDate><link>https://news.ycombinator.com/item?id=49206498</link><dc:creator>cagz</dc:creator><comments>https://news.ycombinator.com/item?id=49206498</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49206498</guid></item><item><title><![CDATA[New comment by cagz in "Show HN: Open-source engine running Gemma 4 26B in 2 GB RAM on any M-series Mac"]]></title><description><![CDATA[
<p>Is the 1/10+ reduction in memory usage applicable to larger models, i.e. would be possible to run a 200Gb model, by using 20Gb of memory?</p>
]]></description><pubDate>Wed, 29 Jul 2026 17:14:42 +0000</pubDate><link>https://news.ycombinator.com/item?id=49100270</link><dc:creator>cagz</dc:creator><comments>https://news.ycombinator.com/item?id=49100270</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49100270</guid></item><item><title><![CDATA[New comment by cagz in "Show HN: macOS menu-bar manager for SSH port forwards"]]></title><description><![CDATA[
<p>I couldn't help noticing on the screenshot that user having direct access to a production database from their laptop :)</p>
]]></description><pubDate>Fri, 24 Jul 2026 16:17:57 +0000</pubDate><link>https://news.ycombinator.com/item?id=49037825</link><dc:creator>cagz</dc:creator><comments>https://news.ycombinator.com/item?id=49037825</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49037825</guid></item><item><title><![CDATA[New comment by cagz in "Write code like a human will maintain it"]]></title><description><![CDATA[
<p>My workflow is:
- Write the test first, confirm it fails
- Write the minimum code to make that test pass
- Confirm it is green</p>
]]></description><pubDate>Fri, 10 Jul 2026 17:39:57 +0000</pubDate><link>https://news.ycombinator.com/item?id=48862907</link><dc:creator>cagz</dc:creator><comments>https://news.ycombinator.com/item?id=48862907</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48862907</guid></item><item><title><![CDATA[New comment by cagz in "Show HN: Semble – Code search for agents that uses 98% fewer tokens than grep"]]></title><description><![CDATA[
<p>Would be nice to see the actual % token/time saving across end-to-end coding sessions over these time periods <a href="https://github.com/MinishLab/semble#savings" rel="nofollow">https://github.com/MinishLab/semble#savings</a><p>My observation is that greps and the processing of grep outputs account for only a small portion of overall consumption; I haven't measured this scientifically though.</p>
]]></description><pubDate>Mon, 18 May 2026 06:15:34 +0000</pubDate><link>https://news.ycombinator.com/item?id=48176126</link><dc:creator>cagz</dc:creator><comments>https://news.ycombinator.com/item?id=48176126</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48176126</guid></item><item><title><![CDATA[New comment by cagz in "delta time"]]></title><description><![CDATA[
<p>I like how far mature adulthood goes :) One never gets old.</p>
]]></description><pubDate>Thu, 14 May 2026 06:11:32 +0000</pubDate><link>https://news.ycombinator.com/item?id=48131708</link><dc:creator>cagz</dc:creator><comments>https://news.ycombinator.com/item?id=48131708</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48131708</guid></item><item><title><![CDATA[New comment by cagz in "Ask HN: What Are You Working On? (April 2026)"]]></title><description><![CDATA[
<p>Getting frustrated by not having RAG solutions able to answer schematic questions such as "how many X are there", I've created DuoRAG: <a href="https://github.com/cagriy/duo-rag" rel="nofollow">https://github.com/cagriy/duo-rag</a><p>It maintains a vector store and a SQL database. While vector store supports usual RAG operations, the ones that require counting, summation, selection are routed to the SQL database.<p>There is an option to start with an initial schema, or let it discover the schema itself. Then on the day to day use, if a user query cannot be responded, a candidate schema entry is created to be populated on the next backfill run.<p>So in actual use, user asks the question such as "Give me the list of people who are scientists". If it is not in the schema, LLM suggest checking it later. Backfill runs at night. Next day it can answer the same question without issues.</p>
]]></description><pubDate>Mon, 13 Apr 2026 13:55:34 +0000</pubDate><link>https://news.ycombinator.com/item?id=47752031</link><dc:creator>cagz</dc:creator><comments>https://news.ycombinator.com/item?id=47752031</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47752031</guid></item><item><title><![CDATA[New comment by cagz in "I run multiple $10K MRR companies on a $20/month tech stack"]]></title><description><![CDATA[
<p>Nice tech read, but without information about which companies, doing what, just feels way too click-baity.</p>
]]></description><pubDate>Sun, 12 Apr 2026 08:11:01 +0000</pubDate><link>https://news.ycombinator.com/item?id=47737209</link><dc:creator>cagz</dc:creator><comments>https://news.ycombinator.com/item?id=47737209</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47737209</guid></item></channel></rss>