<?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: FearedLynx</title>
    <description>The latest articles on DEV Community by FearedLynx (@fearedlynx).</description>
    <link>https://dev.to/fearedlynx</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.us-east-2.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F4253596%2F0f7b248b-8be0-47cb-a77b-163ec1be7c60.jpg</url>
      <title>DEV Community: FearedLynx</title>
      <link>https://dev.to/fearedlynx</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/fearedlynx"/>
    <language>en</language>
    <item>
      <title>We Built a Bot That Runs a Car Transport Company for a Week, and Beat the Organisers' Own System</title>
      <dc:creator>FearedLynx</dc:creator>
      <pubDate>Thu, 19 Nov 2026 10:50:28 +0000</pubDate>
      <link>https://dev.to/fearedlynx/we-built-a-bot-that-runs-a-car-transport-company-for-a-week-and-beat-the-organisers-own-system-50h9</link>
      <guid>https://dev.to/fearedlynx/we-built-a-bot-that-runs-a-car-transport-company-for-a-week-and-beat-the-organisers-own-system-50h9</guid>
      <description>&lt;p&gt;&lt;em&gt;This is a submission for the &lt;a href="https://dev.to/mlh-hackathon"&gt;MLH x DEV Writing Challenge&lt;/a&gt;&lt;/em&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  What I Built
&lt;/h2&gt;

&lt;p&gt;At &lt;strong&gt;Junction 2026&lt;/strong&gt; in Helsinki, our team &lt;strong&gt;Elysa's Secret&lt;/strong&gt; took on the &lt;strong&gt;Bahn Express "Spread" challenge&lt;/strong&gt;. The task sounded simple: build a bot that runs a car transport company for one week, completely on its own.&lt;/p&gt;

&lt;p&gt;Car dealers ask for a price to move a car from A to B. Our bot has a few seconds to answer, and the dealer is also asking a competitor. If we win the car, we offer the job to freelance drivers, and they decide for themselves whether to take it. The goal is to make as much money as possible. But there's a catch: if fewer than 90% of your cars arrive on time, you don't get a score at all.&lt;/p&gt;

&lt;p&gt;That one rule changed everything. Winning lots of cars is easy. Getting enough drivers to actually deliver them is the hard part.&lt;/p&gt;

&lt;h3&gt;
  
  
  What made our bot work
&lt;/h3&gt;

&lt;p&gt;We ended up with three simple ideas:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Don't take cars you can't deliver.&lt;/strong&gt; Cars were almost only late when they got picked up late, so the bot learned to say no to the risky ones.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Pay drivers for the drive.&lt;/strong&gt; Drivers care about how long a trip takes, not about the suggested price, so we set our pay based on drive time.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Send drivers home.&lt;/strong&gt; Drivers are much happier to take a job that ends near their home, so the bot combined two cars into one trip that does exactly that.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;The bot doesn't rely on memorised answers. It learns during the week what prices and fees work, so it can handle a week it has never seen.&lt;/p&gt;

&lt;h3&gt;
  
  
  Result
&lt;/h3&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;6th of 57&lt;/strong&gt; in our best practice round, and we beat the organisers' own bot on the same weeks.&lt;/li&gt;
&lt;li&gt;In the live final we finished &lt;strong&gt;16th of 58 ranked teams&lt;/strong&gt; in Europe's Biggest hackathon.&lt;/li&gt;
&lt;/ul&gt;

&lt;h3&gt;
  
  
  What I learned
&lt;/h3&gt;

&lt;p&gt;Most of our ideas didn't work. Lowering prices to win more cars just lost money, because there weren't enough drivers to deliver them. The lesson I'm taking home: &lt;strong&gt;look at your worst result, not your average.&lt;/strong&gt; When one bad week means you score nothing, playing it safe is the smart move.&lt;/p&gt;

&lt;h2&gt;
  
  
  Demo
&lt;/h2&gt;

&lt;p&gt;  &lt;iframe src="https://www.youtube.com/embed/1q8kELOUgoc" width="710" height="399"&gt;
  &lt;/iframe&gt;
&lt;/p&gt;

&lt;p&gt;Code: &lt;a href="https://github.com/emermelada/bahn-junction2026" rel="noopener noreferrer"&gt;github.com/emermelada/bahn-junction2026&lt;/a&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  Partner Technologies
&lt;/h2&gt;

&lt;p&gt;We built the whole project with &lt;strong&gt;Claude Code&lt;/strong&gt; as a coding partner.&lt;/p&gt;

&lt;p&gt;Before writing any code, we wrote a short "rulebook" for the AI: what our goal was, what mattered most (deliver on time first, earn money second), and a few rules it had to follow every time. We also split the work into two parts, so two AI sessions could work in parallel without getting in each other's way.&lt;/p&gt;

&lt;p&gt;The best habit we picked up was &lt;strong&gt;testing every idea before trusting it&lt;/strong&gt;. We told the AI to check our assumptions with experiments instead of just building on them. That's how we discovered that two of our own beliefs about drivers were wrong, and fixing those became our best ideas.&lt;/p&gt;

&lt;p&gt;What surprised me most: with an AI writing code this fast, the hard part isn't building things anymore. It's being honest about whether they actually work.&lt;/p&gt;

&lt;h2&gt;
  
  
  Hackathon Experience
&lt;/h2&gt;

&lt;p&gt;Junction 2026 in Helsinki, 13–15 November. The challenge had practice rounds every few hours, all through Friday night and Saturday, so there was always a new score to chase, even at 3 a.m.&lt;/p&gt;

&lt;p&gt;We started as a team of two. On Saturday morning my teammate moved to another challenge and left me a clear list of everything we'd tested, and I carried the bot through to the end.&lt;/p&gt;

&lt;p&gt;The moment I'll remember most is Sunday at 8:00. The final started, and for 34 minutes I wasn't allowed to touch anything. I just watched the bot make its own decisions, hoping everything we'd built would hold up. It did: we stayed ranked and finished 16th out of 58.&lt;/p&gt;

&lt;p&gt;My tip for next time: test early, and don't be afraid to throw away your own "great" ideas.&lt;/p&gt;

</description>
      <category>mlhacks</category>
      <category>devchallenge</category>
      <category>hackathon</category>
    </item>
  </channel>
</rss>
