<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0"><channel><title>Transportation LLM Watch</title><link>https://mohsin-mahmud.github.io/blogs/</link><description>Evidence-led essays on language models and agentic systems in transportation by Mohsin Mahmud Topu.</description><language>en</language><lastBuildDate>Fri, 25 Sep 2026 14:30:00 +0600</lastBuildDate>
<item><title>The Trophy Is Not the Test</title><link>https://mohsin-mahmud.github.io/blogs/the-trophy-is-not-the-test.html</link><guid>https://mohsin-mahmud.github.io/blogs/the-trophy-is-not-the-test.html</guid><pubDate>Fri, 25 Sep 2026 00:00:00 +0600</pubDate><description>Why this week's transportation-AI research moved beyond leaderboard scores toward counterfactual diagnosis, bounded model authority, inspectable traffic reasoning and faster closed-loop simulation.</description></item>
<item><title>Who Holds the Keys When AI Enters the Vehicle?</title><link>https://mohsin-mahmud.github.io/blogs/who-holds-the-keys.html</link><guid>https://mohsin-mahmud.github.io/blogs/who-holds-the-keys.html</guid><pubDate>Fri, 18 Sep 2026 00:00:00 +0600</pubDate><description>Four new transportation-AI studies point toward the same design lesson: language models may advise, diagnose and simulate, while independent systems retain execution authority.</description></item>
<item><title>The AI Should Not Hold the Steering Wheel. Let It Build the Crash Test.</title><link>https://mohsin-mahmud.github.io/blogs/who-tests-the-tester.html</link><guid>https://mohsin-mahmud.github.io/blogs/who-tests-the-tester.html</guid><pubDate>Fri, 11 Sep 2026 00:00:00 +0600</pubDate><description>A cluster of new studies moved language models away from direct control and toward a more defensible role: generating tests, measuring causal reasoning and deciding when autonomy should yield.</description></item>
<item><title>Twenty-Two AI Drivers Enter a Ring Road. Traffic Learns a New Kind of Instability</title><link>https://mohsin-mahmud.github.io/blogs/ai-drivers-ring-road.html</link><guid>https://mohsin-mahmud.github.io/blogs/ai-drivers-ring-road.html</guid><pubDate>Fri, 04 Sep 2026 00:00:00 +0600</pubDate><description>This week’s most provocative result was not that language models can control cars. It was that individually plausible choices can become a collective traffic wave.</description></item>
<item><title>Teach the Car Where It Fails, Not Only Where Experts Succeed</title><link>https://mohsin-mahmud.github.io/blogs/teach-the-car.html</link><guid>https://mohsin-mahmud.github.io/blogs/teach-the-car.html</guid><pubDate>Fri, 28 Aug 2026 00:00:00 +0600</pubDate><description>RoG-DAgger turns an old driving-school truth into a post-training strategy: the dangerous lesson begins after the learner drifts away from the ideal route.</description></item>
<item><title>The Most Honest Transportation-AI Story This Week Is an Empty Road</title><link>https://mohsin-mahmud.github.io/blogs/the-empty-road.html</link><guid>https://mohsin-mahmud.github.io/blogs/the-empty-road.html</guid><pubDate>Fri, 21 Aug 2026 00:00:00 +0600</pubDate><description>No new transportation-specific LLM work cleared the evidence bar in this watch window. That quiet week is itself a useful signal about where the research pipeline remains thin.</description></item>
<item><title>The Camera Saw the Crash. Could It Explain What Happened?</title><link>https://mohsin-mahmud.github.io/blogs/after-the-crash.html</link><guid>https://mohsin-mahmud.github.io/blogs/after-the-crash.html</guid><pubDate>Fri, 14 Aug 2026 00:00:00 +0600</pubDate><description>A week of research exposed the difference between noticing an anomaly and understanding a transport system well enough to act without punishing innocent behaviour.</description></item>
</channel></rss>
