<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Tanveer Singh: Writing</title><link>https://tanveerhari.com/writing/</link><atom:link href="https://tanveerhari.com/feed.xml" rel="self" type="application/rss+xml"/><description>Essays by Tanveer Singh on software architecture, intelligent systems, and the craft.</description><language>en</language><item><title>Why AI should grow wiser</title><link>https://tanveerhari.com/writing/why-ai-should-grow-wiser.html</link><guid>https://tanveerhari.com/writing/why-ai-should-grow-wiser.html</guid><description>LLMs lowered the walls between ideas and their making. The harder question is what all that capability is for.</description><category>Essay</category></item><item><title>The story behind Maiaen</title><link>https://tanveerhari.com/writing/the-story-behind-maiaen.html</link><guid>https://tanveerhari.com/writing/the-story-behind-maiaen.html</guid><description>Fifteen years of separate threads, a music composer, a language for testers, a year of philosophy, that turned out to be one idea.</description><category>Architecture</category></item><item><title>Start small, stay narrow</title><link>https://tanveerhari.com/writing/start-small-stay-narrow.html</link><guid>https://tanveerhari.com/writing/start-small-stay-narrow.html</guid><description>Over the past three or four years, I have made a lot of mistakes. Basic mistakes. And the worst part is that I am still making them today. Simple rules to…</description><category>Essay</category></item><item><title>LLMs: a discovery more than an invention</title><link>https://tanveerhari.com/writing/llms-a-discovery.html</link><guid>https://tanveerhari.com/writing/llms-a-discovery.html</guid><description>Not good architecturally, but viable economically. On using LLMs while building towards systems that need them less.</description><category>Architecture</category></item><item><title>The agent memory problem: notes from building one</title><link>https://tanveerhari.com/writing/the-agent-memory-problem.html</link><guid>https://tanveerhari.com/writing/the-agent-memory-problem.html</guid><description>The first time I noticed it, I was watching our agent fail at a task it had succeeded at twenty minutes earlier. Same prompt, same environment, same…</description><category>Architecture</category></item><item><title>When the bottleneck moves: notes on consulting after LLMs</title><link>https://tanveerhari.com/writing/when-the-bottleneck-moves.html</link><guid>https://tanveerhari.com/writing/when-the-bottleneck-moves.html</guid><description>I have spent enough time watching small teams outbuild large ones, in 2026, that I have stopped treating it as the exception. It is the new shape of the…</description><category>Architecture</category></item><item><title>Explore then scale: an engineering discipline for AI-native companies</title><link>https://tanveerhari.com/writing/explore-then-scale.html</link><guid>https://tanveerhari.com/writing/explore-then-scale.html</guid><description>I have started to notice that &quot;AI-native&quot; has joined the small cemetery of phrases that mean less every time they are used. Cloud-native died there. So…</description><category>Architecture</category></item><item><title>The story behind these pages</title><link>https://tanveerhari.com/writing/the-story-behind-these-pages.html</link><guid>https://tanveerhari.com/writing/the-story-behind-these-pages.html</guid><description>Some of the earliest memories I have are of maps. I would follow rivers with a finger until they ran out at a coastline, trace roads through villages…</description><category>Essay</category></item><item><title>From engineer to architect: what AI changed</title><link>https://tanveerhari.com/writing/from-engineer-to-architect.html</link><guid>https://tanveerhari.com/writing/from-engineer-to-architect.html</guid><description>For years, the primary obstacle to my ideas was resources. I had the design instinct. I had the taste. I didn&#x27;t have the hands. Every project past a…</description><category>Architecture</category></item><item><title>Habit beats goals: the pedometer principle</title><link>https://tanveerhari.com/writing/habit-beats-goals.html</link><guid>https://tanveerhari.com/writing/habit-beats-goals.html</guid><description>I use a pedometer app. It has a daily step goal. I rarely hit it. No calamity has ever followed from missing it. And yet the app has changed how much I…</description><category>Essay</category></item><item><title>MusCom, the failure I keep coming back to</title><link>https://tanveerhari.com/writing/muscom.html</link><guid>https://tanveerhari.com/writing/muscom.html</guid><description>A final-year project that failed in almost every way, and the music composer it is still trying to become.</description><category>Essay</category></item><item><title>Design your DSL for the layperson, not the expert</title><link>https://tanveerhari.com/writing/design-your-dsl-for-the-layperson.html</link><guid>https://tanveerhari.com/writing/design-your-dsl-for-the-layperson.html</guid><description>The natural instinct when designing a tool for a domain is to design it for domain experts. It feels respectful. It feels rigorous. It&#x27;s almost always…</description><category>Architecture</category></item><item><title>Domain-driven design for creative computing</title><link>https://tanveerhari.com/writing/domain-driven-design-for-creative-computing.html</link><guid>https://tanveerhari.com/writing/domain-driven-design-for-creative-computing.html</guid><description>When I was sketching a DSL for a creative domain (in this case, a computational-composition project), I kept running into the same design decision. Where…</description><category>Architecture</category></item><item><title>Formal languages are the bridge between programming and intelligence</title><link>https://tanveerhari.com/writing/formal-languages-are-the-bridge.html</link><guid>https://tanveerhari.com/writing/formal-languages-are-the-bridge.html</guid><description>In 2023 I wrote a note to myself that formal languages were the bridge between programming and intelligent systems. I was speculating. Three years on, the…</description><category>Architecture</category></item><item><title>MVP paralysis when the domain is too broad</title><link>https://tanveerhari.com/writing/mvp-paralysis.html</link><guid>https://tanveerhari.com/writing/mvp-paralysis.html</guid><description>I once spent months trying to define an MVP for a product idea in the assistive-AI space. Every candidate MVP I sketched kept expanding. To work at all…</description><category>Architecture</category></item><item><title>Deadlines as complexity control</title><link>https://tanveerhari.com/writing/deadlines-as-complexity-control.html</link><guid>https://tanveerhari.com/writing/deadlines-as-complexity-control.html</guid><description>I once had a task I estimated at two weeks. Over a year later, it hadn&#x27;t shipped. In the post-mortem, the two-week estimate turned out to be correct. What…</description><category>Architecture</category></item><item><title>Start with objectives, not with grammars</title><link>https://tanveerhari.com/writing/start-with-objectives-not-with-grammars.html</link><guid>https://tanveerhari.com/writing/start-with-objectives-not-with-grammars.html</guid><description>The most common failure mode I see in technically capable builders (including in myself, regularly) is jumping to implementation detail before settling…</description><category>Architecture</category></item><item><title>A life around curiosity</title><link>https://tanveerhari.com/writing/a-life-around-curiosity.html</link><guid>https://tanveerhari.com/writing/a-life-around-curiosity.html</guid><description>Maps, made-up planets, crude poetry, then programming. The harder part was building a life around the curiosity.</description><category>Essay</category></item><item><title>Inspiration needs attention</title><link>https://tanveerhari.com/writing/inspiration-needs-attention.html</link><guid>https://tanveerhari.com/writing/inspiration-needs-attention.html</guid><description>A specific failure mode I&#x27;ve watched unfold multiple times, more than once: a small team of talented engineers, each holding a demanding role elsewhere…</description><category>Essay</category></item><item><title>The IT job’s moderation tax</title><link>https://tanveerhari.com/writing/the-it-jobs-moderation-tax.html</link><guid>https://tanveerhari.com/writing/the-it-jobs-moderation-tax.html</guid><description>For a chunk of my early career, I worked as an engineer in the standard IT-corporate mode and told myself the money justified the misfit. It didn&#x27;t, at…</description><category>Essay</category></item><item><title>Intelligence is relative to its world</title><link>https://tanveerhari.com/writing/intelligence-is-relative.html</link><guid>https://tanveerhari.com/writing/intelligence-is-relative.html</guid><description>Intelligence is the ability to form, update and use concepts, and every one of those is bounded by the system doing it.</description><category>Architecture</category></item><item><title>Worlds, models and minds that evolve</title><link>https://tanveerhari.com/writing/worlds-models-and-minds.html</link><guid>https://tanveerhari.com/writing/worlds-models-and-minds.html</guid><description>A system with a fixed model is a world. A system whose model evolves is a mind. The difference is where intelligence lives.</description><category>Architecture</category></item><item><title>Software that thinks with you</title><link>https://tanveerhari.com/writing/software-that-thinks-with-you.html</link><guid>https://tanveerhari.com/writing/software-that-thinks-with-you.html</guid><description>From commands to conversation: why development should be interactive, and what makes an application intelligent at all.</description><category>Architecture</category></item><item><title>The Indian IT bubble that never burst</title><link>https://tanveerhari.com/writing/the-indian-it-bubble-that-never-burst.html</link><guid>https://tanveerhari.com/writing/the-indian-it-bubble-that-never-burst.html</guid><description>Is a typical IT services company in India a business built out of technology, or a business that merely uses technology? The two sound alike. They are not.</description><category>Essay</category></item><item><title>Art and craft</title><link>https://tanveerhari.com/writing/art-and-craft.html</link><guid>https://tanveerhari.com/writing/art-and-craft.html</guid><description>Art asks you to forget the rules. Craft asks you to keep them. Most institutions teach only one.</description><category>Essay</category></item><item><title>A workshop for stories</title><link>https://tanveerhari.com/writing/a-workshop-for-stories.html</link><guid>https://tanveerhari.com/writing/a-workshop-for-stories.html</guid><description>What a story is, why one thought can take many mediums, and why it needs a place to be made.</description><category>Essay</category></item><item><title>Why IoT is hard to test</title><link>https://tanveerhari.com/writing/iot-1-why-iot-is-hard-to-test.html</link><guid>https://tanveerhari.com/writing/iot-1-why-iot-is-hard-to-test.html</guid><description>Scale, variety and the internet itself make the usual QA process fall apart. What to do instead.</description><category>Architecture</category></item><item><title>An end-to-end test harness</title><link>https://tanveerhari.com/writing/iot-2-an-end-to-end-test-harness.html</link><guid>https://tanveerhari.com/writing/iot-2-an-end-to-end-test-harness.html</guid><description>Simulated devices, a broker in the middle, and a framework that checks every hop: how to automate an IoT system’s tests.</description><category>Architecture</category></item><item><title>Edge AI in the physical world</title><link>https://tanveerhari.com/writing/iot-3-edge-ai-attendance.html</link><guid>https://tanveerhari.com/writing/iot-3-edge-ai-attendance.html</guid><description>An attendance system that combines a sensor, a Raspberry Pi and face recognition, and what its architecture teaches.</description><category>Architecture</category></item><item><title>The shape of the problem</title><link>https://tanveerhari.com/writing/netconf-1-the-shape-of-the-problem.html</link><guid>https://tanveerhari.com/writing/netconf-1-the-shape-of-the-problem.html</guid><description>What a Netconf manager has to do, written as one small equation, and why it splits into a library and a service.</description><category>Architecture</category></item><item><title>The library</title><link>https://tanveerhari.com/writing/netconf-2-the-library.html</link><guid>https://tanveerhari.com/writing/netconf-2-the-library.html</guid><description>Four layers in the RFC become four components: operations, session, RPC and transport, with the session at the centre.</description><category>Architecture</category></item><item><title>The service</title><link>https://tanveerhari.com/writing/netconf-3-the-service.html</link><guid>https://tanveerhari.com/writing/netconf-3-the-service.html</guid><description>Growing the design one use case at a time: from one client and one node to many clients and many nodes.</description><category>Architecture</category></item><item><title>Subscriptions and notifications</title><link>https://tanveerhari.com/writing/netconf-4-subscriptions.html</link><guid>https://tanveerhari.com/writing/netconf-4-subscriptions.html</guid><description>Why a good architecture absorbs RFC 5277 with almost no change, and the one place it should bend.</description><category>Architecture</category></item></channel></rss>
