The AI world has been so caught up in the GPT-5.6 Sol / Mythos 5 / two-tier system drama that you'd be forgiven for missing what happened on June 29 — a quiet Monday that saw three genuinely important releases across video, foundation models, and physical AI.
Here's your rapid-fire briefing.
🎬 ByteDance: Seedance 2.5
ByteDance dropped Seedance 2.5, the latest iteration of its AI video generation model. While OpenAI and Google battle over reasoning benchmarks, ByteDance has been quietly eating the video generation market. Seedance 2.5 brings improved temporal consistency (fewer flickering objects), better multi-scene coherence for longer clips, and a reported 40% speedup over Seedance 2.0. It's already rolling out to CapCut and select enterprise partners.
🧪 Liquid AI: LFM 2.5 (230M)
The team behind the Liquid Neural Network architecture released Liquid Foundation Models 2.5 — a compact 230-million-parameter non-transformer model. Don't let the size fool you: LFM 2.5 uses Liquid's state-space architecture that achieves transformer-comparable quality at a fraction of the compute cost. The 230M model is designed for edge deployment — think smartphone assistants, IoT devices, and real-time inference in resource-constrained environments. It's the quiet counter-narrative to "bigger is better."
🤖 WIRobotics: Allex Simulation Model
In Seoul, WIRobotics announced the beginning of its Physical AI development ecosystem, with the first technology release being the Allex Simulation Model. This is a simulated humanoid robot environment designed for training AI agents in realistic physics. The idea: train your robot brain in Allex, then deploy to the real world. It's part of a growing trend of simulation-first robotics that we saw with NVIDIA's Cosmos 3 — but WIRobotics is betting on open access to accelerate the ecosystem.
The Takeaway
While the world watches the geopolitical theatre of frontier model restrictions, the actual AI industry keeps shipping. Better video generation, leaner edge models, and physical AI simulation — June 29 was a reminder that the most interesting AI progress often happens outside the headlines.
Which of these three will have the biggest impact? Drop your thoughts in the comments 👇
Top comments (0)