The deeper you go, the better it gets. At 6 layers: Grassmann trails by 15%
At 12 layers: Grassmann trails by only 11% The model learns to approximate richer interactions through depth and local multi-scale windows. This is proof of concept that attention isn't necessary.
@godofprompt
-
Depth and local windows reduce Grassmann trails, replacing attention
By
–
-
Grassmann model trails in perplexity but leads in SNLI accuracy
By
–
Does it actually work in practice? Wikitext-2 (language modeling): 13M param Grassmann model: 275.7 perplexity
13M param Transformer: 248.4 perplexity
Gap: ~11% behind SNLI (natural language inference): Grassmann head: 85.50% validation accuracy
Transformer head: 85.45% -
Attention quadratic vs Grassmann linear scaling complexity gap
By
–
The complexity story is WILD: Standard attention: O(L²d) – quadratic scaling with sequence length
Grassmann mixing: O(Ld²) – LINEAR scaling (for fixed rank r) This isn't just theoretical. As sequences get longer, the gap becomes exponential. -
Grassmann flows replace attention with geometric evolution
By
–
The alternative is mathematically beautiful: Grassmann flows replace attention with controlled geometric evolution on a manifold. – Reduce hidden states to r=32 dimensions (from d=256)
– Encode token pairs as 2D subspaces on Gr(2,r)
– Use Plücker coordinates to create geometric -
Massive attention matrices cause Transformer interpretability crisis
By
–
Here's the interpretability crisis no one talks about: Every Transformer layer creates an L×L attention matrix. Across 96 layers and 96 heads in GPT-4, that's a MASSIVE evolving tensor cloud. The paper proves this is why we can't interpret these models not their size, but their
-
Transformer performance matched without computing attention weights
By
–
This paper just murdered the foundation of every AI model you've ever used. A researcher proved you can match Transformer performance WITHOUT computing a single attention weight. Here's what changed (and why this matters now):
-
Anthropic CEO ranks CCP, democracies, AI companies as threats
By
–
The CEO of Anthropic just called AI companies a civilizational threat. His own industry. Dario ranks who he’s worried about. CCP first. Democratic governments second (he’s scared the tools we build to fight authoritarians get turned inward). Third: AI companies. He ranks
-
MCP servers deliver interactive interfaces for seamless AI workflows
By
–
This means:
— God of Prompt (@godofprompt) 27 janvier 2026
Any MCP server can now deliver interactive interfaces wherever users work.
Developers get a universal way to bring their tools into AI conversations.
And users get seamless workflows without tab-switching. pic.twitter.com/DJiy4dH8cDThis means: Any MCP server can now deliver interactive interfaces wherever users work. Developers get a universal way to bring their tools into AI conversations. And users get seamless workflows without tab-switching.
-
MCP Apps extends Model Context Protocol for interactive UI in AI products
By
–
The tech behind this is MCP Apps. Built on the Model Context Protocol (MCP), the open standard for connecting AI to tools. MCP Apps extends this further so developers can build interactive UI that works in any supporting AI product not just Claude.
-
Claude enables Asana timelines, Slack messages, Figma diagrams from text
By
–
Build project timelines in Asana. Draft and send Slack messages with formatting preview. Create Figma diagrams from text.
— God of Prompt (@godofprompt) 27 janvier 2026
All of this happens inside your conversation with Claude.
This isn't just integration. It's transformation. pic.twitter.com/WWm215YEqTBuild project timelines in Asana. Draft and send Slack messages with formatting preview. Create Figma diagrams from text. All of this happens inside your conversation with Claude. This isn't just integration. It's transformation.