Show Notes
What does an AI model actually use when it reasons? This episode explores chain of thought, why written reasoning is useful but imperfect, and the research history behind "Neuralese," a term originally used for learned machine-to-machine communication protocols. It also looks at Coconut and continuous latent reasoning, and why less human-readable reasoning creates new interpretability and safety questions.
Voice narration is AI-generated.
REFERENCES
OpenAI — Reasoning models struggle to control their chains of thought, and that's good https://openai.com/index/reasoning-models-chain-of-thought-controllability/
OpenAI — Evaluating chain-of-thought monitorability https://openai.com/index/evaluating-chain-of-thought-monitorability/
OpenAI — Detecting misbehavior in frontier reasoning models https://openai.com/index/chain-of-thought-monitoring/
Anthropic — Measuring faithfulness in Chain-of-Thought reasoning https://www.anthropic.com/research/measuring-faithfulness-in-chain-of-thought-reasoning
Anthropic — Reasoning models don't always say what they think https://www.anthropic.com/research/reasoning-models-dont-say-think
ACL Anthology — Translating Neuralese, Andreas, Dragan, Klein, 2017 https://aclanthology.org/P17-1022/
ArXiv — Training Large Language Models to Reason in a Continuous Latent Space, Coconut https://arxiv.org/abs/2412.06769
ACL Anthology — Cognitively Inspired Developmental Trajectories Improve Explore-Exploit Dynamics in Neural Agent Emergent Communication, 2026 https://aclanthology.org/2026.conll-main.8/
ACL Anthology — Unsupervised Translation of Emergent Communication, 2025 https://ojs.aaai.org/index.php/AAAI/article/view/34489
More from Learn AI in Bits
Recursive self-improvement is the idea that an AI can improve the very process it uses to improve itself, and this episode separates what's been demonstrated so far from the dramatic predictions often attached to the term. It draws a clear line between self-correction (revising one answer), automated optimization (keeping the best of several tested variants), and true recursive improvement, where the system changes the process that generates future improvements, so the next round starts with a better improver.
This week's AI News Sunday Wrap-Up covers a week where the industry moved on several fronts at once: new flagship and open models, enterprise agent governance, a serious safety warning, legal pressure, and the physical infrastructure limits behind the AI race.
Meta's Muse story is moving fast, and this episode takes a closer look at Muse Spark, the multimodal reasoning model family behind Meta's push into AI agents, including yesterday's release of an open-weight version for developers.
When a large language model generates text one token at a time, even a fast GPU hits a sequential bottleneck. This episode explains speculative decoding, an inference technique that prepares several tokens at once and checks them together, and when the extra computation pays off.
When an AI model answers slowly, the problem usually isn't its intelligence — it's how the serving system moves data through a GPU. This episode goes one level below the basic inference loop covered in an earlier episode, into GPU memory, the key-value cache, attention kernels, and the scheduling techniques used by systems like vLLM and NVIDIA TensorRT-LLM.