<?xml version="1.0" encoding="utf-8"?>
<feed xmlns="http://www.w3.org/2005/Atom">
  <title>Thoughts — Shahdad Fard</title>
  <subtitle>Notes on systems thinking, applied AI, graph data models, geometry, and decision intelligence.</subtitle>
  <link href="https://shahdad.dev/thoughts" rel="alternate" type="text/html" />
  <link href="https://shahdad.dev/feed.xml" rel="self" type="application/atom+xml" />
  <id>https://shahdad.dev/thoughts</id>
  <updated>2026-07-01T00:00:00Z</updated>
  <author>
    <name>Shahdad Fard</name>
    <uri>https://shahdad.dev/about</uri>
  </author>
  <entry>
    <title>Applied AI is not software decoration</title>
    <link href="https://shahdad.dev/thoughts/applied-ai-is-not-software-decoration" rel="alternate" type="text/html" />
    <id>https://shahdad.dev/thoughts/applied-ai-is-not-software-decoration</id>
    <updated>2026-07-01T00:00:00Z</updated>
    <published>2026-07-01T00:00:00Z</published>
    <summary>Adding AI on top of a fragmented process does not make the process intelligent. It adds another layer that still cannot see the system.</summary>
    <author>
      <name>Shahdad Fard</name>
      <uri>https://shahdad.dev/about</uri>
    </author>
  </entry>
  <entry>
    <title>Systems-of-systems thinking in AI</title>
    <link href="https://shahdad.dev/thoughts/systems-of-systems-thinking-in-ai" rel="alternate" type="text/html" />
    <id>https://shahdad.dev/thoughts/systems-of-systems-thinking-in-ai</id>
    <updated>2026-06-17T00:00:00Z</updated>
    <published>2026-06-17T00:00:00Z</published>
    <summary>AI becomes useful when geometry, data, people, standards, tools, constraints, machines, and decisions are connected into something the system can actually reason with.</summary>
    <author>
      <name>Shahdad Fard</name>
      <uri>https://shahdad.dev/about</uri>
    </author>
  </entry>
  <entry>
    <title>Decision memory is not documentation</title>
    <link href="https://shahdad.dev/thoughts/decision-memory-is-not-documentation" rel="alternate" type="text/html" />
    <id>https://shahdad.dev/thoughts/decision-memory-is-not-documentation</id>
    <updated>2026-06-03T00:00:00Z</updated>
    <published>2026-06-03T00:00:00Z</published>
    <summary>Documentation records what was decided. Decision memory preserves why it was decided, what was rejected, and what should inform the next move.</summary>
    <author>
      <name>Shahdad Fard</name>
      <uri>https://shahdad.dev/about</uri>
    </author>
  </entry>
  <entry>
    <title>The workflow is the interface</title>
    <link href="https://shahdad.dev/thoughts/the-workflow-is-the-interface" rel="alternate" type="text/html" />
    <id>https://shahdad.dev/thoughts/the-workflow-is-the-interface</id>
    <updated>2026-05-20T00:00:00Z</updated>
    <published>2026-05-20T00:00:00Z</published>
    <summary>The real interface of complex work is not a screen. It is the sequence of decisions, handovers, constraints, and forgotten context that shape what happens next.</summary>
    <author>
      <name>Shahdad Fard</name>
      <uri>https://shahdad.dev/about</uri>
    </author>
  </entry>
  <entry>
    <title>Why graph data models matter for applied AI</title>
    <link href="https://shahdad.dev/thoughts/why-graph-data-models-matter-for-applied-ai" rel="alternate" type="text/html" />
    <id>https://shahdad.dev/thoughts/why-graph-data-models-matter-for-applied-ai</id>
    <updated>2026-05-06T00:00:00Z</updated>
    <published>2026-05-06T00:00:00Z</published>
    <summary>Graphs make relationships explicit. That matters because AI cannot reason well when the structure of the work is hidden across files, tools, and people.</summary>
    <author>
      <name>Shahdad Fard</name>
      <uri>https://shahdad.dev/about</uri>
    </author>
  </entry>
  <entry>
    <title>Geometry is a decision system</title>
    <link href="https://shahdad.dev/thoughts/geometry-is-a-decision-system" rel="alternate" type="text/html" />
    <id>https://shahdad.dev/thoughts/geometry-is-a-decision-system</id>
    <updated>2026-04-22T00:00:00Z</updated>
    <published>2026-04-22T00:00:00Z</published>
    <summary>In complex design-to-manufacturing workflows, geometry is not only shape. It carries fabrication logic, cost, risk, tolerance, assembly, and performance.</summary>
    <author>
      <name>Shahdad Fard</name>
      <uri>https://shahdad.dev/about</uri>
    </author>
  </entry>
  <entry>
    <title>AI starts before the prompt</title>
    <link href="https://shahdad.dev/thoughts/ai-starts-before-the-prompt" rel="alternate" type="text/html" />
    <id>https://shahdad.dev/thoughts/ai-starts-before-the-prompt</id>
    <updated>2026-04-08T00:00:00Z</updated>
    <published>2026-04-08T00:00:00Z</published>
    <summary>A prompt cannot fix a fragmented workflow. Before AI can reason, the system needs context, relationships, constraints, and memory.</summary>
    <author>
      <name>Shahdad Fard</name>
      <uri>https://shahdad.dev/about</uri>
    </author>
  </entry>
  <entry>
    <title>Every serious project needs its own thinking system</title>
    <link href="https://shahdad.dev/thoughts/every-serious-project-needs-its-own-thinking-system" rel="alternate" type="text/html" />
    <id>https://shahdad.dev/thoughts/every-serious-project-needs-its-own-thinking-system</id>
    <updated>2026-03-25T00:00:00Z</updated>
    <published>2026-03-25T00:00:00Z</published>
    <summary>After 14+ years working with complex geometry, DfMA, façades, and structures, I do not believe in one-size-fits-all workflows. Complex projects need custom systems of reasoning.</summary>
    <author>
      <name>Shahdad Fard</name>
      <uri>https://shahdad.dev/about</uri>
    </author>
  </entry>
</feed>
