<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Finance on cloud_affects</title><link>https://cloudaffects.com/tags/finance/</link><description>Recent content in Finance on cloud_affects</description><generator>Hugo</generator><language>en</language><copyright>CC BY-NC-SA 4.0</copyright><lastBuildDate>Fri, 08 May 2026 00:00:00 +0000</lastBuildDate><atom:link href="https://cloudaffects.com/tags/finance/index.xml" rel="self" type="application/rss+xml"/><item><title>Dexter — A Financial Research Agent in Bun</title><link>https://cloudaffects.com/gems/dexter/</link><pubDate>Fri, 08 May 2026 00:00:00 +0000</pubDate><guid>https://cloudaffects.com/gems/dexter/</guid><description>&lt;p&gt;&lt;a href="https://github.com/virattt/dexter"&gt;Dexter&lt;/a&gt; is an autonomous research agent for financial analysis. You ask a question, it plans the steps, pulls income statements, balance sheets, and cash flows from the Financial Datasets API, validates its own answer, and iterates until it&amp;rsquo;s done. The README pitches it as &amp;ldquo;Claude Code, but for financial research&amp;rdquo; — that framing is closer to honest than most.&lt;/p&gt;
&lt;p&gt;The part I&amp;rsquo;d steal first is the scratchpad. Every query writes a &lt;code&gt;.dexter/scratchpad/&amp;lt;timestamp&amp;gt;_&amp;lt;hash&amp;gt;.jsonl&lt;/code&gt; file with newline-delimited entries: the original question, every tool call with its raw result, the LLM summary of that result, and the agent&amp;rsquo;s reasoning steps. JSONL on disk, not a hosted trace UI, not OpenTelemetry — just a file you can grep. For an agent that touches numbers people will act on, an auditable log isn&amp;rsquo;t optional, and most agent projects bolt this on later.&lt;/p&gt;</description></item></channel></rss>