AI moved fast in November, with several major releases worth noting. Let me catch you up on what I missed.
GeoSpy’s AI tool that finds exact locations from photos – no GPS or metadata needed, got an update. Their free version gives city-level accuracy, while their Superbolt model pinpoints specific street corners. Beyond the cool factor, this raises serious privacy questions about what our photos reveal.
DeepSeek-R1 emerged as a Chinese competitor to OpenAI’s reasoning models. It takes time to work through complex problems but matches top AI benchmarks. The model runs under China’s strict AI guidelines, which adds an interesting wrinkle to the global AI race.
Prime Intellect released a 10 billion parameter AI model trained across three continents. This decentralized approach minimizes communication between nodes while leveraging global computing power. They plan an open-source release soon.
Vidu’s 1.5 update added Multi-Entity Consistency to AI video generation. This means characters, objects, and backgrounds maintain visual coherence throughout scenes. While not perfect, it’s getting better. They’re working on pushing resolution to 1080p.
Stable Diffusion 3.5 dropped unexpectedly with mixed results. It generates images quickly but quality varies wildly. Compared to Flux 1.1 Pro it falls short, though it matches Flux Schnell’s capabilities.
For more details on recent AI developments, check out my coverage of OpenAI’s planned releases at https://adam.holter.com/openai-plans-12-days-of-ai-releases-including-sora-and-new-reasoning-model/
I’ll keep tracking these developments as they unfold. The AI space moves fast, but the real value comes from understanding which advances matter for practical applications.
