<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"><channel><title>manish.sh — AI Tips &amp; Tricks</title><description>Manish Shahi is a software engineer and AI developer who writes the Inside LLMs series — reverse-engineering assistants by interviewing them, then auditing claims against papers — along with MCP server guides, agent tooling walkthroughs, and practical notes on models and developer workflows.</description><link>https://manish.sh/</link><language>en</language><item><title>Inside Character.ai: The Technical Story of What Keeps Users Hooked</title><link>https://manish.sh/writings/ai-tools/inside-character-ai-the-technical-story-of-what-keeps-users-hooked/</link><guid isPermaLink="true">https://manish.sh/writings/ai-tools/inside-character-ai-the-technical-story-of-what-keeps-users-hooked/</guid><description>From c.ai to Character.ai — what it is, the problem it solves, then Kaiju, memory, Lorebook, and the infra behind long chats.</description><pubDate>Wed, 22 Jul 2026 19:30:00 GMT</pubDate><category>ai-tools</category><category>Character.ai</category><category>Kaiju</category><category>Noam Shazeer</category><category>Daniel De Freitas</category><category>LLM</category><category>KV Cache</category><category>MQA</category><category>Quantization</category><category>Memory</category><category>Lorebook</category><category>DPO</category><category>Inference</category></item><item><title>Inside Qwen 3.8-Max-Preview: Reverse Engineering an AI Assistant by Interviewing Itself</title><link>https://manish.sh/writings/models/inside-qwen-3-8-max-preview-reverse-engineering-an-ai-assistant-by-interviewing-itself/</link><guid isPermaLink="true">https://manish.sh/writings/models/inside-qwen-3-8-max-preview-reverse-engineering-an-ai-assistant-by-interviewing-itself/</guid><description>I interviewed Qwen 3.8-Max-Preview on memory, tools, context, and hallucinations, then checked published research. Plain English, diagrams, transcript.</description><pubDate>Wed, 22 Jul 2026 07:30:00 GMT</pubDate><category>models</category><category>Qwen</category><category>Qwen3.8-Max-Preview</category><category>Qwen3</category><category>Tongyi Qianwen</category><category>Alibaba</category><category>LLM</category><category>MoE</category><category>Context Window</category><category>Tool Calling</category><category>Memory</category><category>Hallucination</category><category>Transformer</category></item><item><title>Inside DeepSeek: Reverse Engineering an AI Assistant by Interviewing Itself</title><link>https://manish.sh/writings/models/inside-deepseek-reverse-engineering-an-ai-assistant-by-interviewing-itself/</link><guid isPermaLink="true">https://manish.sh/writings/models/inside-deepseek-reverse-engineering-an-ai-assistant-by-interviewing-itself/</guid><description>Interviewed DeepSeek on how it works, then checked against papers. Same method as Kimi — plainer English and concept diagrams.</description><pubDate>Tue, 21 Jul 2026 07:30:00 GMT</pubDate><category>models</category><category>DeepSeek</category><category>LLM</category><category>MoE</category><category>MLA</category><category>Context Window</category><category>Hallucination</category><category>RLHF</category><category>Transformer</category><category>Tool Calling</category><category>Hidden Reasoning</category></item><item><title>Inside Kimi K2.6: Reverse Engineering an AI Assistant by Interviewing Itself</title><link>https://manish.sh/writings/models/inside-kimi-k2-6-reverse-engineering-an-ai-assistant-by-interviewing-itself/</link><guid isPermaLink="true">https://manish.sh/writings/models/inside-kimi-k2-6-reverse-engineering-an-ai-assistant-by-interviewing-itself/</guid><description>I interviewed Kimi on memory, tools, and safety, then checked published research on K2.6. Plain English, diagrams, sources.</description><pubDate>Mon, 20 Jul 2026 18:00:00 GMT</pubDate><category>models</category><category>Kimi</category><category>Kimi K2.6</category><category>Moonshot AI</category><category>LLM</category><category>Context Window</category><category>Tool Calling</category><category>Memory</category><category>Hallucination</category><category>MoE</category><category>MLA</category><category>Agent Swarm</category></item><item><title>How I Built an MCP Server in PHP for Fayyaz Travels</title><link>https://manish.sh/writings/tips/how-i-built-mcp-in-php-for-fayyaz-travels/</link><guid isPermaLink="true">https://manish.sh/writings/tips/how-i-built-mcp-in-php-for-fayyaz-travels/</guid><description>Fayyaz Travels is an Inncelerator client. I shipped PHP MCP for site analytics first. Flights, hotels, holidays, and ticket email come next.</description><pubDate>Sun, 19 Jul 2026 12:00:00 GMT</pubDate><category>tips</category><category>MCP</category><category>PHP</category><category>Fayyaz Travels</category><category>Inncelerator</category><category>AI Agents</category><category>Travel Tech</category><category>JSON-RPC</category><category>SSE</category><category>Redis</category><category>MySQL</category><category>Analytics</category></item><item><title>Give Your Agents Power to Purchase: Agentcard.sh Go-To Guide</title><link>https://manish.sh/writings/ai-tools/give-your-agents-power-to-purchase-agentcard-sh-go-to-guide/</link><guid isPermaLink="true">https://manish.sh/writings/ai-tools/give-your-agents-power-to-purchase-agentcard-sh-go-to-guide/</guid><description>Agents fill carts then stall at checkout. Agentcard.sh issues capped one-time Visas so they can finish buying — MCP, Companies wizard, and when not to use it.</description><pubDate>Sat, 18 Jul 2026 17:30:00 GMT</pubDate><category>ai-tools</category><category>Agentcard</category><category>AI Agents</category><category>Payments</category><category>MCP</category><category>Visa</category><category>Card Issuing</category></item></channel></rss>