The case for real-time robotics with large models
One post predicts that big models will soon challenge the idea that they’re too slow for robotics. A quote-post argues that most such systems need little on-the-fly reasoning for real-time behavior.
TLDR
A September 26, 2026 post predicts that within weeks, many DeepSeek V4 Flash–scale JEVs will run in real time and come surprisingly close to Astra through distillation. The author also argues that people underestimate how fast prefill can get. Quoting that prediction, another user points to Jev’s speed as an example of fast “System 1” behavior with language models, arguing that most such systems don’t need much reasoning on the fly.
The case for real-time robotics with large models
One post predicts that big models will soon challenge the idea that they’re too slow for robotics. A quote-post argues that most such systems need little on-the-fly reasoning for real-time behavior.
TLDR
A September 26, 2026 post predicts that within weeks, many DeepSeek V4 Flash–scale JEVs will run in real time and come surprisingly close to Astra through distillation. The author also argues that people underestimate how fast prefill can get. Quoting that prediction, another user points to Jev’s speed as an example of fast “System 1” behavior with language models, arguing that most such systems don’t need much reasoning on the fly.