<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Rudi-Renoir Appoldt on Medium]]></title>
        <description><![CDATA[Stories by Rudi-Renoir Appoldt on Medium]]></description>
        <link>https://medium.com/@rrenoir?source=rss-d9f41929d746------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*Id_iCYkLf--YpzOEdxArbw.jpeg</url>
            <title>Stories by Rudi-Renoir Appoldt on Medium</title>
            <link>https://medium.com/@rrenoir?source=rss-d9f41929d746------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sun, 22 Nov 2026 12:37:55 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@rrenoir/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Things 3 becomes Task Backend for Your Local AI Agent]]></title>
            <link>https://medium.com/@rrenoir/turn-things-3-into-the-task-backend-for-your-local-ai-agent-fbc5a6228274?source=rss-d9f41929d746------2</link>
            <guid isPermaLink="false">https://medium.com/p/fbc5a6228274</guid>
            <category><![CDATA[claude-code]]></category>
            <category><![CDATA[openai-codex]]></category>
            <category><![CDATA[things]]></category>
            <category><![CDATA[agentic-ai]]></category>
            <category><![CDATA[gtd-getting-things-done]]></category>
            <dc:creator><![CDATA[Rudi-Renoir Appoldt]]></dc:creator>
            <pubDate>Thu, 02 Jul 2026 07:04:18 GMT</pubDate>
            <atom:updated>2026-07-02T13:57:14.838Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="A Man stands in front of a computer in a big room. Around him are collages of lists. All very colorfull. The Image suggests that AI is helping with dayli tasks." src="https://cdn-images-1.medium.com/max/1024/1*uBr66BAecbdUzZ5RBFtqNg.png" /><figcaption>Life might become easier to manage if an AI Agents knows about your to do’s</figcaption></figure><p><strong>How I gave Codex safe, controlled access to my task manager — without building a new to-do system.</strong></p><p>I’m building a lightweight AI Chief of Staff with Codex. Things remains my operational task manager, while Codex helps me understand projects, identify conflicts, and decide what deserves attention.</p><p>The goal was not full automation. I wanted controlled access with clear safety boundaries.</p><h3>1. Locating the Things database</h3><p>On macOS, Things stores its local data inside:</p><pre>~/Library/Group Containers/<br>JLMPQHK86H.com.culturedcode.ThingsMac/</pre><p>Inside this container is a directory such as ThingsData-xxxxx, containing the SQLite database.</p><p>Cultured Code documents this location for local backups and data exports. The exported data is available as an SQLite database. <a href="https://culturedcode.com/things/support/articles/2982272/">Things: Exporting Your Data</a></p><h3>2. Never querying the live database directly</h3><p>Codex does not perform analysis against the live database.</p><p>Before each read, it creates a consistent temporary SQLite snapshot:</p><pre>sqlite3 -readonly &quot;/path/to/main.sqlite&quot; \<br>  &quot;.backup &#39;/private/tmp/things-snapshot.sqlite&#39;&quot;</pre><p>It then opens this snapshot in immutable, read-only mode:</p><pre>sqlite3 -readonly \<br>  &quot;file:/private/tmp/things-snapshot.sqlite?immutable=1&quot; \<br>  &quot;SELECT ...;&quot;</pre><p>This allows Codex to inspect information such as:</p><ul><li>open tasks and projects</li><li>deadlines</li><li>scheduled tasks</li><li>project relationships</li><li>completion status</li><li>modification dates</li></ul><p>The original Things database remains untouched.</p><h3>3. Treating the database schema as unstable</h3><p>The database is accessible, but its internal schema is not a guaranteed public API.</p><p>Before using the data, Codex verifies that the required tables and columns still exist. If the schema changes or a field cannot be interpreted reliably, the analysis stops.</p><p>The rule is simple:</p><blockquote><em>It is better to provide no answer than to provide a confident but incorrect answer.</em></blockquote><p>Codex also avoids writing private task contents into diagnostic logs.</p><h3>4. Separating task management from project context</h3><p>My Things structure has evolved over time. It contains real projects, old ideas, reference collections, and other material that should not automatically be treated as active work.</p><p>For this reason, Codex maintains a small set of Markdown files containing strategic project context.</p><p>For example:</p><pre>---<br>status: active<br>working_deadline: 2026-07-31<br>things_projects:<br>  - THINGS-PROJECT-ID-1<br>  - THINGS-PROJECT-ID-2<br>---</pre><p>This makes it possible to express that two Things projects actually belong to the same larger initiative.</p><p>The responsibilities remain separate:</p><ul><li>Things is authoritative for tasks and their status.</li><li>Markdown contains project goals, current state, risks, and strategic meaning.</li><li>Codex combines both sources when making recommendations.</li></ul><p>Tasks are never copied into Markdown.</p><h3>5. Adding new tasks through the official URL scheme</h3><p>Initially, Codex had read-only access.</p><p>I later added one tightly controlled exception: Codex may create a new task inside one designated Things project, but only after I explicitly approve the task.</p><p>New tasks are created through the official Things URL scheme, not through SQLite. <a href="https://culturedcode.com/things/support/articles/2803573/">Things URL Scheme</a></p><p>Example:</p><pre>open &#39;things:///add?title=Prepare%20Project%201%20deliverable&amp;list=PROJECT%201&amp;when=today&#39;</pre><p>The URL scheme supports parameters including:</p><ul><li>title</li><li>notes</li><li>when</li><li>deadline</li><li>list or list-id</li><li>heading</li></ul><p>Codex is not allowed to update, move, complete, or delete existing tasks.</p><h3>6. Verifying every write</h3><p>After creating a task, Codex produces another fresh database snapshot.</p><p>It then verifies that:</p><ul><li>exactly the intended task was created,</li><li>the title is correct,</li><li>it belongs to the correct project,</li><li>its scheduling state is correct.</li></ul><p>A write is not considered successful merely because the URL command executed without an error.</p><h3>7. Persisting the safety contract</h3><p>The operating rules are stored in the project’s AGENTS.md file so they remain available in future Codex conversations.</p><p>The contract includes:</p><ul><li>Read only from a consistent temporary snapshot.</li><li>Never modify the SQLite database.</li><li>Detect schema changes before analysis.</li><li>Stop when field semantics are uncertain.</li><li>Never copy Things tasks into Markdown.</li><li>Create tasks only in one designated Things project.</li><li>Require explicit confirmation before every task creation.</li><li>Never update or delete existing Things data.</li><li>Verify every created task using a new snapshot.</li><li>Never create tasks silently or in bulk.</li></ul><h3>What this enables</h3><p>I can now ask Codex questions such as:</p><ul><li>What requires attention today?</li><li>Is my current workload realistic?</li><li>Which projects have become inactive?</li><li>What am I repeatedly postponing?</li><li>Which task has the greatest immediate consequence?</li><li>What should I deliberately not work on today?</li></ul><p>Things remains the task manager. Codex acts as the reasoning layer above it: reading project context, detecting conflicts, and helping me make deliberate choices.</p><p>The useful part is not merely giving an AI access to tasks. It is defining exactly what the AI may observe, what it may infer, and what it is allowed to change.</p><p>I just started to use this. We will see if it works in the longrun. But it seams like a nice way to expand on the existing workflow.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=fbc5a6228274" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Music, AI, and the Language of Emotion]]></title>
            <link>https://medium.com/@rrenoir/music-ai-and-the-language-of-emotion-619f02dc3520?source=rss-d9f41929d746------2</link>
            <guid isPermaLink="false">https://medium.com/p/619f02dc3520</guid>
            <dc:creator><![CDATA[Rudi-Renoir Appoldt]]></dc:creator>
            <pubDate>Sun, 22 Sep 2024 20:09:31 GMT</pubDate>
            <atom:updated>2024-09-22T20:09:31.528Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*vfDH0P4f-6fRHZUJQ5_vMQ.jpeg" /><figcaption>Can Music inspire a new approach towards AI?</figcaption></figure><p>As someone with a music background, I find it fascinating to explore the connections between music, artificial intelligence (AI), and computation. There is a surprising overlap between the language of music theory and computer programming. Both fields use systems, structures, and concepts that share common ground, especially when it comes to audio processing and data processing. This shared language has helped me understand programming better, and in turn, offers insights into the world of AI.</p><p>However, what makes music unique is its foundation in emotional expression. Music, unlike programming, doesn’t primarily solve problems; it speaks to our emotions. Jazz, for instance, started as a form of emotional and spontaneous performance, and only later was it analyzed and systematized through theory. What I’ve learned is that jazz harmony, which feels so free and expressive, is deeply rooted in mathematics.</p><h4>Music and Mathematics: A Connection Through Jazz</h4><p>During my time at Berklee College of Music, I was trained in jazz harmony, which derives from the Schillinger System of Musical Composition. Joseph Schillinger, a professor in New York, taught a method of composition that relied on mathematical processes. He worked with musicians like George Gershwin and Glenn Miller, and his system focused on describing music through mathematical principles. What struck me most was that this system didn’t prescribe how music should be made but rather described the underlying structure of musical expression.</p><p>Now that I’m studying programming and AI, I see similar references to this structured approach. If mathematics can serve as the foundation for jazz harmony, a genre rooted in emotion, then perhaps it can also serve as a language for emotional expression in AI and programming. This idea makes me wonder if programming, traditionally focused on logic and problem-solving, could be approached like music, where the focus is on the journey of emotional storytelling rather than just solving a problem.</p><h4>Nature, Aesthetics, and the Foundation of Music</h4><p>Another essential aspect of music, especially for me, is its connection to nature. When I began studying music, I wanted to understand the “magic” behind it — what is the guiding principle that makes music so powerful? I found my answer in nature. The dynamics of the natural world — waves crashing, winds rising and falling — parallel the dynamics of music, like crescendos and decrescendos.</p><p>When I took advanced classes in modal harmony, I learned to create synthetic scales to push the boundaries of music theory. At this theoretical limit, music becomes more mathematical, but still, we rely on references to nature to keep it aesthetically pleasing. This insight has led me to believe that many systems, including music, computation, and AI, are connected at their core. Mathematics might be the key to translating between these systems, but it’s incomplete without the emotional component that music brings.</p><h4>Bridging the Gap: Can Programming Learn from Music?</h4><p>The emotional journey in music, with its tension and release, is what captivates listeners. A successful musical composition or performance navigates this journey in a way that resonates with the audience. In contrast, programming and AI tend to be linear, aiming for efficiency and solutions rather than expression. So, what if we could apply the principles of emotional expression found in music to programming?</p><p>In AI, there’s a tendency to “average out” results, pushing everything toward the middle. In music, we strive to avoid this. Music embraces tension, surprises, and unpredictability — all elements that create emotional depth. As humans, we are drawn to indications of life and evolution. We expect artists to grow and change; when they don’t, we lose interest. I believe the same may be true for AI — if it continues to produce “mediocre” results, people will eventually reject it.</p><h4>AI, Evolution, and Aesthetic Perception</h4><p>This brings me to another idea: our perception of consciousness might be tied to our emotional connection with living things. Even though AI may never achieve consciousness, it might still evolve in ways that help us perceive emotional depth. If we can teach AI that evolution is linked to aesthetics and life, we might create systems that are more adaptive, more human-like in their thinking and perceiving.</p><p>Looking at the history of AI, it’s clear that many scientists have focused on solving the problems of learning and consciousness. But maybe it’s time to include an angle that comes from the world of artistic expression. If AI could hold some of the magic that a great musical performance does — if it could impose on us the feeling that it wants to connect — this could be the first step toward creating a system that becomes an integral part of our human experience.</p><p>Otherwise, just like a tasteless artificial tomato, AI might be rejected if it fails to deliver that essential emotional connection.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=619f02dc3520" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>