<?xml version="1.0" encoding="UTF-8"?>
<urlset xmlns="http://www.sitemaps.org/schemas/sitemap/0.9" xmlns:image="http://www.google.com/schemas/sitemap-image/1.1">
<url>
<loc>https://www.khetwal.com</loc>
<changefreq>weekly</changefreq>
<priority>1</priority>
</url>
<url>
<loc>https://www.khetwal.com/news</loc>
<changefreq>weekly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://www.khetwal.com/papers</loc>
<changefreq>weekly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://www.khetwal.com/models</loc>
<changefreq>weekly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://www.khetwal.com/blog</loc>
<changefreq>weekly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://www.khetwal.com/off-hours</loc>
<changefreq>weekly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://www.khetwal.com/about</loc>
<changefreq>weekly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://www.khetwal.com/connect</loc>
<changefreq>weekly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://www.khetwal.com/blog/kimi-k3-a-2-8-trillion-parameter-open-model-that-bets-on-linear-attention</loc>
<image:image>
<image:loc>https://cdn.khetwal.com/blogs/posts/kimi-k3-8f198bce.jpg</image:loc>
</image:image>
<lastmod>2026-08-10T14:55:59.109Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.6</priority>
</url>
<url>
<loc>https://www.khetwal.com/papers/attention-is-all-you-need-the-paper-behind-modern-ai-in-plain-english</loc>
<image:image>
<image:loc>https://cdn.khetwal.com/blogs/posts/paper-cover-ba6f918c.jpg</image:loc>
</image:image>
<lastmod>2026-07-28T19:00:27.985Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.6</priority>
</url>
<url>
<loc>https://www.khetwal.com/blog/llm-system-design-architecture-the-end-to-end-guide-to-building-production-ai</loc>
<image:image>
<image:loc>https://cdn.khetwal.com/blogs/posts/llm-system-design-architecture-the-end-to-end-guide-to-building-production-ai-ee28911e69.png</image:loc>
</image:image>
<lastmod>2026-08-11T15:47:30.038Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.6</priority>
</url>
<url>
<loc>https://www.khetwal.com/blog/gitlost-how-a-github-ai-agent-leaked-private-repos-from-one-public-issue</loc>
<image:image>
<image:loc>https://cdn.khetwal.com/blogs/posts/gitlost-how-a-github-ai-agent-leaked-private-repos-from-one-public-issue-024cc8794f.png</image:loc>
</image:image>
<lastmod>2026-08-11T15:47:30.764Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.6</priority>
</url>
<url>
<loc>https://www.khetwal.com/papers/from-18-to-58-accuracy-the-simple-prompt-that-taught-ai-to-reason</loc>
<image:image>
<image:loc>https://cdn.khetwal.com/blogs/posts/chain-of-thought-prompting-explained-b3f56aef1d.png</image:loc>
</image:image>
<lastmod>2026-08-10T14:55:47.427Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.6</priority>
</url>
<url>
<loc>https://www.khetwal.com/blog/the-8-000-misunderstanding-how-to-cut-your-llm-bill-by-80</loc>
<image:image>
<image:loc>https://cdn.khetwal.com/blogs/posts/the-8-000-misunderstanding-how-to-cut-your-llm-bill-by-80-338e1f6b24.png</image:loc>
</image:image>
<lastmod>2026-08-10T14:56:02.255Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.6</priority>
</url>
<url>
<loc>https://www.khetwal.com/blog/the-trillion-parameter-tipping-point-open-weight-ai-is-now-a-c-suite-decision</loc>
<image:image>
<image:loc>https://cdn.khetwal.com/blogs/posts/open-weights-in-the-enterprise-a162cd22cd.png</image:loc>
</image:image>
<lastmod>2026-08-11T17:32:26.568Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.6</priority>
</url>
<url>
<loc>https://www.khetwal.com/blog/why-your-rag-pilot-won-t-survive-production</loc>
<image:image>
<image:loc>https://cdn.khetwal.com/blogs/posts/why-rag-breaks-at-scale-b8a56cc543.png</image:loc>
</image:image>
<lastmod>2026-08-10T14:55:44.487Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.6</priority>
</url>
<url>
<loc>https://www.khetwal.com/blog/failover-amnesia-why-your-llm-chatbot-forgets-during-an-outage</loc>
<image:image>
<image:loc>https://cdn.khetwal.com/blogs/posts/llm-failover-amnesia-stateful-continuity-c4df2dd051.png</image:loc>
</image:image>
<lastmod>2026-08-10T23:27:53.961Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.6</priority>
</url>
<url>
<loc>https://www.khetwal.com/blog/your-llm-judge-s-secret-drift-why-80-agreement-isn-t-enough</loc>
<image:image>
<image:loc>https://cdn.khetwal.com/blogs/posts/llm-judge-drift-calibration-63e3c07faa.png</image:loc>
</image:image>
<lastmod>2026-08-10T22:13:59.925Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.6</priority>
</url>
<url>
<loc>https://www.khetwal.com/blog/your-retry-loop-is-a-time-bomb-from-a-50-api-call-to-a-5-000-bill</loc>
<image:image>
<image:loc>https://cdn.khetwal.com/blogs/posts/llm-retry-loop-hidden-costs-45bcd51e6e.png</image:loc>
</image:image>
<lastmod>2026-08-10T23:06:12.025Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.6</priority>
</url>
<url>
<loc>https://www.khetwal.com/blog/your-ai-team-is-a-bottleneck-why-llms-belong-on-the-platform</loc>
<image:image>
<image:loc>https://cdn.khetwal.com/blogs/posts/why-llms-belong-on-the-platform-not-with-an-ai-team-12e4d17b95.png</image:loc>
</image:image>
<lastmod>2026-08-11T17:30:28.506Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.6</priority>
</url>
<url>
<loc>https://www.khetwal.com/blog/your-first-eval-build-a-30-example-test-set-this-afternoon</loc>
<image:image>
<image:loc>https://cdn.khetwal.com/blogs/posts/build-a-30-example-ai-eval-set-this-afternoon-b226ea9c76.png</image:loc>
</image:image>
<lastmod>2026-08-10T14:55:49.006Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.6</priority>
</url>
<url>
<loc>https://www.khetwal.com/blog/the-1m-token-budget-why-1-in-10-queries-fail-at-scale</loc>
<image:image>
<image:loc>https://cdn.khetwal.com/blogs/posts/the-1m-token-budget-context-cost-c271bf5d55.png</image:loc>
</image:image>
<lastmod>2026-08-10T14:55:54.542Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.6</priority>
</url>
<url>
<loc>https://www.khetwal.com/models/claude-s-1m-tokens-90-recall-what-it-solves-what-it-doesn-t</loc>
<image:image>
<image:loc>https://cdn.khetwal.com/blogs/posts/claude-1m-token-context-window-explained-9bd5722d35.png</image:loc>
</image:image>
<lastmod>2026-08-10T23:29:51.163Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.6</priority>
</url>
<url>
<loc>https://www.khetwal.com/papers/train-a-70b-model-on-one-gpu-a-lora-walkthrough</loc>
<image:image>
<image:loc>https://cdn.khetwal.com/blogs/posts/train-70b-model-on-one-gpu-lora-walkthrough-16e6ff9aa7.png</image:loc>
</image:image>
<lastmod>2026-08-10T18:55:35.275Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.6</priority>
</url>
<url>
<loc>https://www.khetwal.com/models/codex-max-and-the-100-mistake-24-hours-in</loc>
<image:image>
<image:loc>https://cdn.khetwal.com/blogs/posts/codex-max-long-horizon-agent-cost-9bbf2e89b0.png</image:loc>
</image:image>
<lastmod>2026-08-10T18:55:31.302Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.6</priority>
</url>
<url>
<loc>https://www.khetwal.com/blog/cut-your-llm-bill-by-70-a-guide-to-four-caching-strategies-for-2026</loc>
<image:image>
<image:loc>https://cdn.khetwal.com/blogs/posts/four-llm-caching-strategies-2026-guide-2615ee6339.png</image:loc>
</image:image>
<lastmod>2026-08-10T23:04:36.575Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.6</priority>
</url>
<url>
<loc>https://www.khetwal.com/blog/fixing-your-recurring-outage-ask-for-a-schema-not-for-json</loc>
<image:image>
<image:loc>https://cdn.khetwal.com/blogs/posts/ask-for-a-schema-not-for-json-a8b6502adb.png</image:loc>
</image:image>
<lastmod>2026-08-10T14:57:46.463Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.6</priority>
</url>
<url>
<loc>https://www.khetwal.com/blog/chunking-is-a-retrieval-decision-up-to-9-recall-is-on-the-line</loc>
<image:image>
<image:loc>https://cdn.khetwal.com/blogs/posts/rag-chunking-strategy-recall-boost-a3b863e6d7.png</image:loc>
</image:image>
<lastmod>2026-08-10T14:55:46.003Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.6</priority>
</url>
<url>
<loc>https://www.khetwal.com/models/opus-4-5-vs-gemini-3-pro-which-model-for-which-workload</loc>
<image:image>
<image:loc>https://cdn.khetwal.com/blogs/posts/gemini-3-pro-vs-claude-4-5-evaluation-plan-fb49e32ccc.png</image:loc>
</image:image>
<lastmod>2026-08-10T23:25:36.496Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.6</priority>
</url>
<url>
<loc>https://www.khetwal.com/blog/role-confusion-the-61-attack-success-rate-exposing-llm-s-core-flaw</loc>
<image:image>
<image:loc>https://cdn.khetwal.com/blogs/posts/role-confusion-llm-authorization-bug-prompt-injection-f4dd8b60b4.png</image:loc>
</image:image>
<lastmod>2026-08-10T23:29:04.114Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.6</priority>
</url>
<url>
<loc>https://www.khetwal.com/blog/the-12-agent-task-why-production-ai-needs-a-harness-not-a-loop</loc>
<image:image>
<image:loc>https://cdn.khetwal.com/blogs/posts/production-ai-agent-harness-architecture-4bec529f26.png</image:loc>
</image:image>
<lastmod>2026-08-10T18:55:29.232Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.6</priority>
</url>
<url>
<loc>https://www.khetwal.com/models/gemini-3-1-pro-is-a-1m-token-window-worth-a-blind-upgrade</loc>
<image:image>
<image:loc>https://cdn.khetwal.com/blogs/posts/gemini-3-1-pro-production-upgrade-guide-c32bc18a02.png</image:loc>
</image:image>
<lastmod>2026-08-10T23:29:22.764Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.6</priority>
</url>
<url>
<loc>https://www.khetwal.com/blog/prompts-are-production-code-version-them-like-it</loc>
<image:image>
<image:loc>https://cdn.khetwal.com/blogs/posts/prompt-versioning-is-production-engineering-1697f84b06.png</image:loc>
</image:image>
<lastmod>2026-08-10T14:55:55.922Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.6</priority>
</url>
<url>
<loc>https://www.khetwal.com/papers/flashattention-how-a-memory-trick-unlocked-today-s-giant-ai-models</loc>
<image:image>
<image:loc>https://cdn.khetwal.com/blogs/posts/flashattention-memory-io-aware-llm-context-754c32e262.jpg</image:loc>
</image:image>
<lastmod>2026-08-11T17:29:52.845Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.6</priority>
</url>
<url>
<loc>https://www.khetwal.com/models/claude-opus-4-5-at-5-mtok-when-to-upgrade-your-agent-s-brain</loc>
<image:image>
<image:loc>https://cdn.khetwal.com/blogs/posts/claude-opus-4-5-routing-tier-upgrade-87b2e224e6.png</image:loc>
</image:image>
<lastmod>2026-08-10T14:55:57.573Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.6</priority>
</url>
<url>
<loc>https://www.khetwal.com/papers/the-r-in-rag-was-meant-to-be-trained-not-just-searched</loc>
<image:image>
<image:loc>https://cdn.khetwal.com/blogs/posts/the-r-in-rag-was-trained-not-just-searched-9fc2f50b07.png</image:loc>
</image:image>
<lastmod>2026-08-10T18:55:24.162Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.6</priority>
</url>
<url>
<loc>https://www.khetwal.com/blog/how-to-log-your-llm-unlock-99-automation-with-better-data</loc>
<image:image>
<image:loc>https://cdn.khetwal.com/blogs/posts/how-to-log-your-llm-unlock-automation-6e1c239a0a.png</image:loc>
</image:image>
<lastmod>2026-08-10T23:28:19.922Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.6</priority>
</url>
<url>
<loc>https://www.khetwal.com/papers/more-data-smaller-models-the-2022-paper-that-resized-every-llm</loc>
<image:image>
<image:loc>https://cdn.khetwal.com/blogs/posts/chinchilla-scaling-laws-smaller-models-0ffa948f72.png</image:loc>
</image:image>
<lastmod>2026-08-10T21:52:35.206Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.6</priority>
</url>
<url>
<loc>https://www.khetwal.com/blog/the-100x-cost-gap-when-to-replace-your-llm-with-a-simple-rule</loc>
<image:image>
<image:loc>https://cdn.khetwal.com/blogs/posts/the-0-01-query-when-a-rule-beats-an-llm-call-92eee9fdc6.png</image:loc>
</image:image>
<lastmod>2026-08-10T23:06:14.473Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.6</priority>
</url>
<url>
<loc>https://www.khetwal.com/blog/the-genai-divide-why-95-of-enterprise-ai-projects-delivered-no-value</loc>
<image:image>
<image:loc>https://cdn.khetwal.com/blogs/posts/enterprise-ai-silent-failures-2025-e4686644f3.png</image:loc>
</image:image>
<lastmod>2026-08-10T23:06:13.293Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.6</priority>
</url>
<url>
<loc>https://www.khetwal.com/blog/your-llm-is-a-bug-a-5-point-test-for-when-not-to-use-ai</loc>
<image:image>
<image:loc>https://cdn.khetwal.com/blogs/posts/when-not-to-use-an-llm-decision-test-4b570ee6e9.png</image:loc>
</image:image>
<lastmod>2026-08-11T18:07:13.056Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.6</priority>
</url>
<url>
<loc>https://www.khetwal.com/papers/react-s-loop-the-simple-pattern-behind-every-modern-ai-agent</loc>
<image:image>
<image:loc>https://cdn.khetwal.com/blogs/posts/react-loop-ai-agent-pattern-f220de8c6d.png</image:loc>
</image:image>
<lastmod>2026-08-11T17:31:04.107Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.6</priority>
</url>
<url>
<loc>https://www.khetwal.com/blog/the-76-mistake-why-ai-build-vs-buy-is-a-layer-by-layer-choice</loc>
<image:image>
<image:loc>https://cdn.khetwal.com/blogs/posts/ai-stack-build-vs-buy-layer-by-layer-choice-7d25821304.png</image:loc>
</image:image>
<lastmod>2026-08-11T17:31:18.509Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.6</priority>
</url>
<url>
<loc>https://www.khetwal.com/blog/why-your-first-ai-invoice-is-a-trap-a-cost-model-for-survival</loc>
<image:image>
<image:loc>https://cdn.khetwal.com/blogs/posts/ai-unit-economics-cost-model-d84609e21a.png</image:loc>
</image:image>
<lastmod>2026-08-11T17:32:10.082Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.6</priority>
</url>
<url>
<loc>https://www.khetwal.com/blog/most-enterprise-rags-have-no-access-control-how-to-fix-yours</loc>
<image:image>
<image:loc>https://cdn.khetwal.com/blogs/posts/enterprise-rag-access-control-fix-5ab446d108.png</image:loc>
</image:image>
<lastmod>2026-08-11T17:31:51.081Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.6</priority>
</url>
<url>
<loc>https://www.khetwal.com/papers/your-llm-is-a-reward-model-how-dpo-cut-alignment-costs</loc>
<image:image>
<image:loc>https://cdn.khetwal.com/blogs/posts/dpo-alignment-without-the-reward-model-1839540695.png</image:loc>
</image:image>
<lastmod>2026-08-11T17:28:53.008Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.6</priority>
</url>
<url>
<loc>https://www.khetwal.com/models/qualifying-claude-opus-4-8-from-shadowing-to-go-no-go-decision</loc>
<image:image>
<image:loc>https://cdn.khetwal.com/blogs/posts/claude-opus-4-8-upgrade-guide-386a648873.jpg</image:loc>
</image:image>
<lastmod>2026-08-11T18:06:36.241Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.6</priority>
</url>
<url>
<loc>https://www.khetwal.com/blog/the-alignment-tax-why-making-llms-safer-can-make-them-less-capable</loc>
<image:image>
<image:loc>https://cdn.khetwal.com/blogs/posts/alignment-tax-llm-safety-capability-tradeoff-e1b75b250f.jpg</image:loc>
</image:image>
<lastmod>2026-08-11T18:56:15.368Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.6</priority>
</url>
<url>
<loc>https://www.khetwal.com/blog/the-bowling-alley-test-a-sanity-check-for-ai-vendor-claims</loc>
<image:image>
<image:loc>https://cdn.khetwal.com/blogs/posts/bowling-alley-test-ai-vendor-benchmarks-8630914906.jpg</image:loc>
</image:image>
<lastmod>2026-08-11T18:07:37.585Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.6</priority>
</url>
<url>
<loc>https://www.khetwal.com/blog/stop-fine-tuning-your-llm-why-retrieval-is-cheaper-faster-more-reliable</loc>
<image:image>
<image:loc>https://cdn.khetwal.com/blogs/posts/stop-fine-tuning-llm-prompt-retrieve-tune-workflow-766c6c681c.png</image:loc>
</image:image>
<lastmod>2026-08-11T18:57:20.778Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.6</priority>
</url>
<url>
<loc>https://www.khetwal.com/blog/your-ai-is-slow-a-latency-budgeting-guide-for-sub-second-agents</loc>
<image:image>
<image:loc>https://cdn.khetwal.com/blogs/posts/ai-latency-budget-guide-for-agents-534f6c4a15.jpg</image:loc>
</image:image>
<lastmod>2026-08-11T18:57:45.827Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.6</priority>
</url>
<url>
<loc>https://www.khetwal.com/blog/hitl-that-works-make-review-faster-than-the-original-task</loc>
<image:image>
<image:loc>https://cdn.khetwal.com/blogs/posts/hitl-that-works-make-review-faster-3e71a90a00.jpg</image:loc>
</image:image>
<lastmod>2026-08-11T18:58:08.030Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.6</priority>
</url>
<url>
<loc>https://www.khetwal.com/models/2-8-trillion-parameters-free-to-download-inside-moonshot-ai-s-kimi-k3</loc>
<image:image>
<image:loc>https://cdn.khetwal.com/blogs/posts/kimi-k3-model-note-card-test-acf33bd061.png</image:loc>
</image:image>
<lastmod>2026-08-11T16:36:36.590Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.6</priority>
</url>
<url>
<loc>https://www.khetwal.com/blog/code-churn-doubles-why-retyping-llm-code-is-a-2x-not-10x-win</loc>
<image:image>
<image:loc>https://cdn.khetwal.com/blogs/posts/code-churn-doubles-retyping-llm-code-74f4487a77.png</image:loc>
</image:image>
<lastmod>2026-08-10T21:56:10.109Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.6</priority>
</url>
</urlset>
