AI agents can talk to each other without words in 4,096-dimensional space
This language has no symbols or sounds: it is geometry in 4,096-dimensional space. A single vector in a multi-dimensional space carries as much meaning as hundreds of words in English.
Princeton, Stanford2 and Microsoft3 independently reported versions of this AI-to-AI language (2024-2025).
The cited system is 24x faster4 than natural-language exchange, with higher reported accuracy and precision.
The parallel with Villeneuve's Arrival5 is hard to miss - the heptapods also thought in whole concepts, not in sequential words, but instead of ink circles we have tensors in latent space.
Ted Chiang was closer to reality than he probably thought.
But a large language model is not a writer; it's not even a user of language.7
Ted ChiangScience fiction writerPhotograph by Arturo Villarrubia, cropped, CC BY-SA 2.0References
- Johnson, W. B. and Lindenstrauss, J., Extensions of Lipschitz Mappings into a Hilbert Space, Contemporary Mathematics 26, 189-206 (American Mathematical Society, 1984) - the theorem behind the packing argument - https://doi.org/10.1090/conm/026/737400↩
- Zou, J., Qiu, R., Li, G., Yang, X., et al., Latent Collaboration in Multi-Agent Systems (arXiv:2511.20639, 25 November 2025; ICML 2026 Spotlight) - Princeton and Stanford - https://arxiv.org/abs/2511.20639↩
- Liu, Y., Huang, Y., Yao, J., Feng, S., et al., DroidSpeak: KV Cache Sharing for Cross-LLM Communication and Multi-LLM Serving (arXiv:2411.02820, 5 November 2024) - Microsoft Research - https://arxiv.org/abs/2411.02820↩
- Du, Z., Wang, R., Bai, H., Cao, Z., et al., Enabling Agents to Communicate Entirely in Latent Space (arXiv:2511.09149, 12 November 2025; accepted to ACL 2026) - the 24x figure - https://arxiv.org/abs/2511.09149↩
- Chiang, T., Story of Your Life, in Starlight 2, ed. Patrick Nielsen Hayden (Tor, 1998)↩
- Ricci-Curbastro, G. and Levi-Civita, T., Méthodes de calcul différentiel absolu et leurs applications, Mathematische Annalen 54, 125-201 (1900) - the founding paper of tensor calculus - https://doi.org/10.1007/BF01454201↩
- Chiang, T., Why A.I. Isn't Going to Make Art, The New Yorker, 31 August 2024 - source of the quotation - https://www.newyorker.com/culture/the-weekend-essay/why-ai-isnt-going-to-make-art↩