Wookash Podcast Mārtiņš Možeiko | Understanding Before Programming
16 snips
Nov 22, 2025 Mārtiņš Možeiko, a software engineer renowned for his work on low-level systems and profiling tools, shares insights from his unique journey in tech. He reminisces about self-teaching Basic on vintage PCs and transitioning through Pascal and C++. Mārtiņš discusses building obfuscation technology at Red Kryption, tackling compiler limitations, and his evolution at LG creating a self-driving simulator. His approach to learning emphasizes hands-on experimentation, and he shares thoughts on modern programming and AI coding tools, all while reflecting on his gaming preferences.
AI Snips
Chapters
Transcript
Episode notes
Building An Open Self-Driving Simulator
- At LG Mārtiņš moved from webOS build/security work to building an open-source Unity-based self-driving car simulator (LGSVL).
- The simulator emulated sensors and interfaced with ROS to test real autonomous software without real cars.
Synthetic Data Boosts Real Training
- Virtual simulated data is easier to generate with precise annotations and improves AI when mixed with real-world data.
- Purely synthetic training isn't sufficient, but combined datasets yield measurable gains.
Avoid AI Search Results For Critical Facts
- Turn off AI-generated search results when they introduce uncertainty; verify facts from primary sources instead.
- Mārtiņš avoids AI results because they often add noise and double his verification work.