Episodes from Latent Space about Open Source AI Tools.

⚡️Making DeepSeek v4 outperform Opus 4.7 with Taste — @Ahmad Awais , CommandCode.ai
Jun 6, 2026 · 40:41
Ahmad Awais explains how his open-source CLI CommandCode uses a 'validate-then-repair' layer to fix tool-calling errors in open models like DeepSeek, allowing them to outperform premium models like Opus 4.7 in 6 of 10 evaluations. He argues that perceived weaknesses are harness/contract issues, not capability gaps, and extends the same repair logic to combat 'design slop' by encoding designer frameworks. Awais also shares plans to open-source CommandCode while keeping it focused on the best models.

⚡️ Google's Open AI Strategy — Omar Sanseviero, Google DeepMind
May 25, 2026 · 29:59
Omar Sanseviero, head of Developer Experience at Google DeepMind, breaks down Gemma 4's novel architecture with per-layer embeddings that enable parameter offloading, allowing a 2B active parameter model to run fast on devices. He explains trade-offs between dense and MoE models, notes fine-tuning is declining as base models improve, and highlights Gemma 4's native multimodal support for audio, images, and short video. The team is growing in Singapore and India, and Kaggle's recent integration will help benchmark agent capabilities.