<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:dc="http://purl.org/dc/elements/1.1/">
  <channel>
    <title>DEV Community: keesan.eth</title>
    <description>The latest articles on DEV Community by keesan.eth (@cryptokeesan).</description>
    <link>https://dev.to/cryptokeesan</link>
    <image>
      <url>https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F918212%2F4a0730ab-6568-4e42-ae73-5088e9b37b59.jpg</url>
      <title>DEV Community: keesan.eth</title>
      <link>https://dev.to/cryptokeesan</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/cryptokeesan"/>
    <language>en</language>
    <item>
      <title>AI coding agents need receipts, not just better prompts</title>
      <dc:creator>keesan.eth</dc:creator>
      <pubDate>Mon, 11 May 2026 17:46:15 +0000</pubDate>
      <link>https://dev.to/cryptokeesan/ai-coding-agents-need-receipts-not-just-better-prompts-838</link>
      <guid>https://dev.to/cryptokeesan/ai-coding-agents-need-receipts-not-just-better-prompts-838</guid>
      <description>&lt;p&gt;AI coding agents are getting good enough to run real engineering tasks, but not safe enough to run without guardrails.&lt;/p&gt;

&lt;p&gt;The failure mode is not always dramatic.&lt;/p&gt;

&lt;p&gt;Sometimes the agent just keeps working.&lt;/p&gt;

&lt;p&gt;It retries.&lt;br&gt;
It rewrites.&lt;br&gt;
It spends tokens.&lt;br&gt;
It changes files.&lt;br&gt;
It says it is done.&lt;/p&gt;

&lt;p&gt;Then another engineer opens the diff and realizes the agent solved the wrong problem.&lt;/p&gt;

&lt;p&gt;That creates a new engineering question:&lt;/p&gt;

&lt;p&gt;Can another engineer audit this run later?&lt;/p&gt;

&lt;p&gt;That is why I’m building MartinLoop.&lt;/p&gt;

&lt;p&gt;MartinLoop is an open-source control plane for AI coding agents. The goal is to make every agent run bounded, inspectable, and test-verifiable.&lt;/p&gt;

&lt;p&gt;The first version focuses on:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;hard budget caps&lt;/li&gt;
&lt;li&gt;JSONL run records&lt;/li&gt;
&lt;li&gt;audit trails&lt;/li&gt;
&lt;li&gt;failure classification&lt;/li&gt;
&lt;li&gt;test-verified completion&lt;/li&gt;
&lt;li&gt;reproducible agent runs&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;The thesis is simple:&lt;/p&gt;

&lt;p&gt;The next layer of AI coding is not only better prompts.&lt;/p&gt;

&lt;p&gt;It is governance.&lt;/p&gt;

&lt;p&gt;Before agents touch serious repos, teams need receipts:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;what the agent tried&lt;/li&gt;
&lt;li&gt;what it changed&lt;/li&gt;
&lt;li&gt;how much it spent&lt;/li&gt;
&lt;li&gt;what commands it ran&lt;/li&gt;
&lt;li&gt;what tests passed&lt;/li&gt;
&lt;li&gt;what failed&lt;/li&gt;
&lt;li&gt;why it stopped&lt;/li&gt;
&lt;li&gt;whether a human can resume, revert, or rerun it&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;I’m looking for feedback from developers using Claude Code, Codex, Cursor, Devin-style agents, or custom coding agents in real repos.&lt;/p&gt;

&lt;p&gt;What would you want in the default “agent receipt”?&lt;/p&gt;

&lt;p&gt;GitHub: &lt;a href="https://github.com/Keesan12/Martin-Loop" rel="noopener noreferrer"&gt;https://github.com/Keesan12/Martin-Loop&lt;/a&gt;&lt;br&gt;
Site: &lt;a href="https://martinloop.com" rel="noopener noreferrer"&gt;https://martinloop.com&lt;/a&gt;&lt;/p&gt;

</description>
      <category>ai</category>
      <category>opensource</category>
      <category>devops</category>
      <category>programming</category>
    </item>
    <item>
      <title>Come Build on Concordium</title>
      <dc:creator>keesan.eth</dc:creator>
      <pubDate>Tue, 30 Aug 2022 19:57:03 +0000</pubDate>
      <link>https://dev.to/cryptokeesan/come-build-on-concordium-2jgh</link>
      <guid>https://dev.to/cryptokeesan/come-build-on-concordium-2jgh</guid>
      <description>&lt;p&gt;Concordium Blockchain is the only public blockchain with a privacy based ID-layer at the protocol level built using RUST. It is the only blockchain with user attributes accessible from Smart Contracts built to be enterprise grade and compliant by nature. &lt;/p&gt;

&lt;p&gt;We welcome all #rustdevs to test us out and help us build out our bounties as we look to create new tools for integrations, interoperability, and Dapps. &lt;/p&gt;

&lt;p&gt;The blockchain for the future has arrived!&lt;/p&gt;

</description>
      <category>rust</category>
      <category>blockchain</category>
      <category>webdev</category>
      <category>bounties</category>
    </item>
  </channel>
</rss>
