<?xml version="1.0" encoding="utf-8"?>
<feed xmlns="http://www.w3.org/2005/Atom">
  <title>Ali Ömer Keser</title>
  <link href="https://keser.ai/"/>
  <link rel="self" href="https://keser.ai/feed.xml"/>
  <id>https://keser.ai/</id>
  <updated>2026-09-25T00:00:00Z</updated>
  <author><name>Ali Ömer Keser</name><uri>https://keser.ai/</uri></author>
  <entry>
    <title>I cut a Claude Code session&#x27;s cost by 34% with a symbol graph</title>
    <link href="https://keser.ai/blog/keeba-claude-code-cost/"/>
    <id>https://keser.ai/blog/keeba-claude-code-cost/</id>
    <published>2026-09-25T00:00:00Z</published>
    <updated>2026-09-25T00:00:00Z</updated>
    <summary>How keeba gives AI coding agents a symbol graph of your codebase instead of whole files, and what a real A/B run in Claude Code showed: $1.45 against $2.19, same two bugs found.</summary>
  </entry>
  <entry>
    <title>My gas-optimized ERC721 had an auth bypass. The fixed one still beats solmate.</title>
    <link href="https://keser.ai/blog/erc721-auth-bypass-then-beat-solmate/"/>
    <id>https://keser.ai/blog/erc721-auth-bypass-then-beat-solmate/</id>
    <published>2026-09-25T00:00:00Z</published>
    <updated>2026-09-25T00:00:00Z</updated>
    <summary>In 2022 I claimed my ERC721 was 61% cheaper than solmate. Re-reading it, the savings came from missing security checks. Here&#x27;s the fix, and the approval-epoch trick that makes the safe version cheaper than solmate and OpenZeppelin on every operation.</summary>
  </entry>
</feed>
