<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>Akash Panchal</title>
    <link>https://akashpanchal.com</link>
    <description>Senior AI Software Engineer. Tutorials and notes on building AI agents, voice systems, and scalable platforms.</description>
    <language>en-us</language>
    <atom:link href="https://akashpanchal.com/rss.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>Voice Agents: The Hard Problems Nobody Warns You About</title>
      <link>https://akashpanchal.com/articles/voice-agents-hard-problems</link>
      <guid>https://akashpanchal.com/articles/voice-agents-hard-problems</guid>
      <description>The demo is easy — speech-to-text, LLM, text-to-speech, done. Then a real person talks to it and everything falls apart: the silences are too long, it talks over them, it can&apos;t tell when they&apos;re done. The hard problems in voice aren&apos;t the models. They&apos;re latency, barge-in, and turn-taking.</description>
      <category>Article</category>
      <pubDate>Sat, 04 Jul 2026 00:00:00 GMT</pubDate>
    </item>
    <item>
      <title>The AI SDK v7 Features I&apos;m Actually Excited About</title>
      <link>https://akashpanchal.com/articles/ai-sdk-v7-features</link>
      <guid>https://akashpanchal.com/articles/ai-sdk-v7-features</guid>
      <description>v7 isn&apos;t a coat of paint. It moves the hard parts of agent-building — the multi-step loop, tool approvals, structured output, telemetry — out of your app and into the SDK. Here are the features I reached for the same week they landed, and why.</description>
      <category>Article</category>
      <pubDate>Wed, 01 Jul 2026 00:00:00 GMT</pubDate>
    </item>
    <item>
      <title>I Let an Agent Write Code and Run It — Safely — in a Vercel Sandbox</title>
      <link>https://akashpanchal.com/articles/agent-runs-code-vercel-sandbox</link>
      <guid>https://akashpanchal.com/articles/agent-runs-code-vercel-sandbox</guid>
      <description>An agent that writes code you can&apos;t run is a fancy autocomplete. But running model-generated code on your own box is how you get a crypto miner. I build a mini-app where the agent writes, executes, and self-corrects code inside a Vercel Sandbox microVM — and walk the parts that surprised me.</description>
      <category>Article</category>
      <pubDate>Sun, 28 Jun 2026 00:00:00 GMT</pubDate>
    </item>
    <item>
      <title>Mastra, Part 7: Evals &amp; Scorers — Proving the Agent Is Good</title>
      <link>https://akashpanchal.com/tutorials/mastra-evals</link>
      <guid>https://akashpanchal.com/tutorials/mastra-evals</guid>
      <description>The finale. An agent that ships without evals is a vibe with a deploy button. This part puts numbers on agent quality — deterministic checks, model-graded scorers, pass/fail gates in CI, and live sampling in production — so &apos;it seems better&apos; becomes &apos;it scored 0.86, up from 0.71&apos;.</description>
      <category>Tutorial</category>
      <pubDate>Tue, 23 Jun 2026 00:00:00 GMT</pubDate>
    </item>
    <item>
      <title>Mastra, Part 6: Long-Running &amp; Durable Agents</title>
      <link>https://akashpanchal.com/tutorials/mastra-durable-agents</link>
      <guid>https://akashpanchal.com/tutorials/mastra-durable-agents</guid>
      <description>Some agent work doesn&apos;t fit in one request. It scrapes forty pages, waits on a human, or runs on a schedule. This part covers the machinery for work that outlives the HTTP request: background tasks, durable agents that survive a crash, and heartbeats that run on a cron.</description>
      <category>Tutorial</category>
      <pubDate>Tue, 16 Jun 2026 00:00:00 GMT</pubDate>
    </item>
    <item>
      <title>Mastra, Part 5: RAG — Giving the Agent Something Real to Say</title>
      <link>https://akashpanchal.com/tutorials/mastra-rag</link>
      <guid>https://akashpanchal.com/tutorials/mastra-rag</guid>
      <description>A streaming agent that answers from the model&apos;s memory is still guessing. In this part I build a retrieval pipeline — chunk, embed, store, query — and wire it under the agent so it answers from your actual documents, with citations.</description>
      <category>Tutorial</category>
      <pubDate>Tue, 09 Jun 2026 00:00:00 GMT</pubDate>
    </item>
    <item>
      <title>Mastra, Part 4: Streaming to a Real UI</title>
      <link>https://akashpanchal.com/tutorials/mastra-streaming</link>
      <guid>https://akashpanchal.com/tutorials/mastra-streaming</guid>
      <description>A blocking generate() call is fine for a script. An interactive agent needs to stream — tokens, tool calls, and your own custom progress events. I wire agent.stream() into a UI and learn where the interesting data actually lives.</description>
      <category>Tutorial</category>
      <pubDate>Tue, 02 Jun 2026 00:00:00 GMT</pubDate>
    </item>
    <item>
      <title>Mastra, Part 3: The Harness</title>
      <link>https://akashpanchal.com/tutorials/mastra-harness</link>
      <guid>https://akashpanchal.com/tutorials/mastra-harness</guid>
      <description>Part three of the Mastra series. An agent and some tools aren&apos;t an application yet. Mastra&apos;s harness — the AgentController runtime — adds the session, modes, subagents, and tool approvals that turn an agent into something users can actually depend on.</description>
      <category>Tutorial</category>
      <pubDate>Tue, 26 May 2026 00:00:00 GMT</pubDate>
    </item>
    <item>
      <title>Mastra, Part 2: Workflows</title>
      <link>https://akashpanchal.com/tutorials/mastra-workflows</link>
      <guid>https://akashpanchal.com/tutorials/mastra-workflows</guid>
      <description>Part two of the Mastra series. Agents are great at deciding; workflows are great at guaranteeing. I build typed, multi-step workflows with sequencing, parallelism, branching, loops, and human-in-the-loop suspend &amp; resume.</description>
      <category>Tutorial</category>
      <pubDate>Tue, 19 May 2026 00:00:00 GMT</pubDate>
    </item>
    <item>
      <title>Mastra, Part 1: Agents</title>
      <link>https://akashpanchal.com/tutorials/mastra-agents</link>
      <guid>https://akashpanchal.com/tutorials/mastra-agents</guid>
      <description>The first part of a hands-on series on Mastra. I start where everything else builds from — defining a typed agent, giving it tools with createTool, and wiring memory so it remembers across turns.</description>
      <category>Tutorial</category>
      <pubDate>Tue, 12 May 2026 00:00:00 GMT</pubDate>
    </item>
  </channel>
</rss>