Cloud agentsLet AI research in the cloud — handling multi-step tasks or running on a schedule.
Research skillsLet AI pick the right methods and tools for each task, and extend it through the skill market.
Paper setsSearch, filter and combine papers, then analyze within a clearly scoped body of material.
Research panelSee sources, papers and results beside the conversation — check the evidence and steer as you go.

Literature discoveryStart from a research question, find the relevant papers, and build a list worth reading closely.
Topic reviewsMap technical approaches, compare evidence, and turn many papers into a topic report.
Ongoing trackingFollow a direction over time, with sourced research briefs and papers queued to read.
Topic researchBring topic materials and results together so AI can watch progress, revisit what you know, and suggest next leads.

Model Architecture

1 report

SurveysOctober 14, 2026

Looped and Recurrent-Depth Transformers: Architectures, Training, Adaptive Inference, Theory and Systems

A looped (or recurrent-depth) transformer applies the same weight-tied block repeatedly to a token's hidden state, turning "depth" from a fixed property of the parameter count into a variable that can be spent at inference time.

Model Architecture

Model Architecture - Parallax