Discussion

• @7uomoki Tuōmo on x

  @JeffDean this willbe huge for things like full length medical histories or biomedical omics data dumps etc! 💪

• @chenzoe0922 Zoe on x

  @JeffDean The Titans architecture is solid and well designed, integrating surprise based prioritization, momentum gating, and adaptive eviction into a fully trainable memory system in a clean and effective way. Its ability to outperform both Transformer++ and Mamba 2 on BABILong …

• @tmcirony22088 @tmcirony22088 on x

  @GoogleResearch Lovely step. If Titans really wants “deeper than 2M tokens”, the choke point isn't context length but how you compose bounded memory. Right now both RNNs and Transformers are doing ad-hoc additions on activations that live in a finite dynamic range. In a

• @brianemcgrath @brianemcgrath on x

  @JeffDean Extended reasoning over massive context windows is one of those capabilities that unlocks entirely new use cases we haven't even imagined yet. This is the kind of fundamental architectural innovation that compounds—each breakthrough in context + reasoning creates expone…

• @yafffesh @yafffesh on x

  @GoogleResearch It solves Retrieval, but it breaks Alignment. Titans uses “Surprise” (gradient magnitude) to decide what to write to memory. But to an AI, a typo is “surprising.” A predictable tragedy is low-gradient noise. Without a Human Anchor (\Gamma) to modulate that surpris…

• @codewithimanshu Himanshu Kumar on x

  @GoogleResearch Google, that's quite a feat, but will it be truly useful in the real world, you think?

• @bobbuitech Bob Bui on x

  @JeffDean Fascinating work. The test-time updates based on ‘surprise’ seem powerful. What's the mechanism to prevent the memory from getting polluted by noisy or even adversarial inputs over time?

• @econ_tech_vance Potato Terminator on x

  @JeffDean Great to see context window barriers breaking. Real test is whether compute and energy demands stay manageable for real-world use.

• @jeffdean Jeff Dean on x

  We have a long history of model architecture innovations, and we're continuing to explore new ideas. Here's some early work that shows promising results on extending the context length over which a model can reason over.

• @azize_rho Azize R. on x

  @JeffDean Great to see memory becoming a first-class citizen in model architectures. We've been exploring external cognitive memory systems using spatial semantic indexing to extend long-range reasoning, fascinating to watch internal and external memory paradigms converge.

• @ilyasiqbal Ilyas Iqbal on x

  @JeffDean This is a major step forward combining the efficiency of RNN-style recurrence with the expressive power of Transformers has been a long-standing goal in the field. Looks like Titans can truly deliver real-time learning with multi-million-token context windows, that resh…

• @arimedai @arimedai on x

  @GoogleResearch I'm impressed. MIRAS framing sequence modeling as online optimization? Smart. Titans might just solve the compute-efficiency bottleneck holding back long-context applications.

• @brianemcgrath @brianemcgrath on x

  @GoogleResearch This is genius! This is exactly the kind of architectural leap that enables AI systems to handle truly complex, long-horizon tasks. The implications for autonomous agents are massive.

• @googleresearch @googleresearch on x

  Today at #NeurIPS2025, we present Titans, a new architecture that combines the speed of RNNs with the performance of Transformers. It uses deep neural memory to learn in real-time, effectively scaling to contexts larger than 2 million tokens. More at: https://research.google/... …