<?xml version='1.0' encoding='utf-8'?>
<rss version="2.0"><channel><title>Writing by Nipun Khanna</title><link>https://nipunkhanna.me/writing/</link><description>Essays on AI agents, product decisions, research and consulting.</description><language>en</language><item><title>Jev: the AI model that makes decisions instead of writing</title><link>https://nipunkhanna.me/writing/jev-ai-model-decisions/</link><guid>https://nipunkhanna.me/writing/jev-ai-model-decisions/</guid><description>A new AI model is getting attention for something odd: it doesn’t write a single sentence. It’s called Jev.</description><pubDate>Tue, 29 Sep 2026 00:00:00 +0000</pubDate></item><item><title>How much of your job would an AI need to do before you called it AGI?</title><link>https://nipunkhanna.me/writing/ai-computer-use-agi/</link><guid>https://nipunkhanna.me/writing/ai-computer-use-agi/</guid><description>Computer use got very real this year. Agents now open your apps, click the buttons and move through websites almost like you do.</description><pubDate>Thu, 24 Sep 2026 00:00:00 +0000</pubDate></item><item><title>AI got better at design when it worked in its own language</title><link>https://nipunkhanna.me/writing/ai-design-own-language/</link><guid>https://nipunkhanna.me/writing/ai-design-own-language/</guid><description>The poster needs visual atmosphere. The slide needs to carry a message in an order someone can follow.</description><pubDate>Thu, 24 Sep 2026 00:00:00 +0000</pubDate></item><item><title>What changes when an agent starts building the slides?</title><link>https://nipunkhanna.me/writing/agent-building-slides/</link><guid>https://nipunkhanna.me/writing/agent-building-slides/</guid><description>What I wanted was an agent beside a live canvas, so the agent and the human could work on the same deck together.</description><pubDate>Sun, 13 Sep 2026 00:00:00 +0000</pubDate></item><item><title>Agent observability: how to find where an AI agent went wrong</title><link>https://nipunkhanna.me/writing/agent-observability/</link><guid>https://nipunkhanna.me/writing/agent-observability/</guid><description>When an AI agent returns the wrong result, the final answer is usually only the symptom.</description><pubDate>Sun, 30 Aug 2026 00:00:00 +0000</pubDate></item><item><title>Loops explained: graph engineering does not replace loops</title><link>https://nipunkhanna.me/writing/graph-engineering-loops/</link><guid>https://nipunkhanna.me/writing/graph-engineering-loops/</guid><description>Graph engineering does not replace loop engineering.</description><pubDate>Sat, 08 Aug 2026 00:00:00 +0000</pubDate></item><item><title>Building better AI agents with harness-driven loops</title><link>https://nipunkhanna.me/writing/harness-driven-loops/</link><guid>https://nipunkhanna.me/writing/harness-driven-loops/</guid><description>A loop alone is a liability. An unconstrained model with tools can wander, overspend, or confidently do the wrong thing.</description><pubDate>Wed, 05 Aug 2026 00:00:00 +0000</pubDate></item><item><title>Build a multi-model relay: Claude architects, Codex verifies, Antigravity implements</title><link>https://nipunkhanna.me/writing/multi-model-relay/</link><guid>https://nipunkhanna.me/writing/multi-model-relay/</guid><description>The first rule: expensive models think, cheap models type.</description><pubDate>Thu, 23 Jul 2026 00:00:00 +0000</pubDate></item><item><title>How I Built a Jail for My OpenClaw Agent and Run It for Free</title><link>https://nipunkhanna.me/writing/openclaw-agent-jail/</link><guid>https://nipunkhanna.me/writing/openclaw-agent-jail/</guid><description>Two problems with OpenClaw and how I fixed both in one weekend.</description><pubDate>Tue, 21 Apr 2026 00:00:00 +0000</pubDate></item><item><title>I Built an Agent-Operated Design Canvas. Here is What I Learned About Agentic Architecture.</title><link>https://nipunkhanna.me/writing/agent-operated-design-canvas/</link><guid>https://nipunkhanna.me/writing/agent-operated-design-canvas/</guid><description>How JSON-driven state, LLM orchestration, and Fabric.js come together to let an AI design in real time.</description><pubDate>Tue, 31 Mar 2026 00:00:00 +0000</pubDate></item><item><title>What happens when AI stops answering questions and starts running your tools?</title><link>https://nipunkhanna.me/writing/ai-starts-running-tools/</link><guid>https://nipunkhanna.me/writing/ai-starts-running-tools/</guid><description>We’re moving from models that answer questions to systems that take sustained, multi-step actions across tools.</description><pubDate>Thu, 26 Feb 2026 00:00:00 +0000</pubDate></item><item><title>I Thought I’d Automated Content Creation. I Hadn’t.</title><link>https://nipunkhanna.me/writing/automated-content-creation/</link><guid>https://nipunkhanna.me/writing/automated-content-creation/</guid><description>A few months of building AI workflows taught me more about human judgment than automation.</description><pubDate>Sun, 15 Feb 2026 00:00:00 +0000</pubDate></item><item><title>AI Agents Won’t Kill Automation. They’ll Stand on Top of It.</title><link>https://nipunkhanna.me/writing/ai-agents-and-automation/</link><guid>https://nipunkhanna.me/writing/ai-agents-and-automation/</guid><description>The future isn’t one or the other; it’s hybrid systems.</description><pubDate>Sun, 30 Nov 2025 00:00:00 +0000</pubDate></item><item><title>The Leverage Mindset: Why One Person + Great System Beats Five People + Mediocre System</title><link>https://nipunkhanna.me/writing/the-leverage-mindset/</link><guid>https://nipunkhanna.me/writing/the-leverage-mindset/</guid><description>Most people think about scaling wrong.</description><pubDate>Sun, 09 Nov 2025 00:00:00 +0000</pubDate></item></channel></rss>