OpenAI is reportedly developing a new architecture called "recurrent depth" for its upcoming GPT-6 model, which aims to improve reasoning by allowing the model to internally loop and refine its thoughts rather than generating lengthy text-based "chains of thought." This shift towards internal, non-textual computation could significantly alter the AI industry's token-based business models. Concurrently, a Russian startup called Mostik has demonstrated a method for inter-model communication that bypasses human language entirely, projecting high-dimensional states directly between models, drastically reducing computational costs and potentially enabling more efficient AI agent workflows. AI
IMPACT These advancements signal a potential shift away from token-based pricing and towards more efficient, internal computation and direct model-to-model communication, impacting AI development and business models.
RANK_REASON Report on OpenAI's new architecture for GPT-6 and a new inter-model communication technique. [lever_c_demoted from frontier_release: ic=1 ai=1.0]
- ARC-AGI
- Astra
- chains of thought
- GLM-5.2
- Google DeepMind
- GPT-6
- Jakub Pachocki
- Karl Tuyls
- Mostik
- OpenAI
- Qwen-3.5
- recurrent depth
- Sasha Malysheva
- Stanislav Smirnov
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
