The week in AI research
Lattice reads the newest models, breakthroughs, and industry studies, then publishes clear summaries and deep-dive explainers with proper citations and a save-for-later list.
% Featured research breakdown
@article{ lattice2026,
title = "Compute-Optimal Training: Scaling Laws Revisited",
topic = "LLMs" }
A fresh look at how to spend a fixed compute budget between model size and training tokens, and why many recent models were undertrained.
Read the breakdownFresh from the lab
A fresh look at how to spend a fixed compute budget between model size and training tokens, and why many recent models were undertrained.
A practical guide to component-level responsiveness: when to use container queries, how to structure containers, and where they outperform viewport breakpoints.
How routing tokens to a small subset of expert networks delivers larger effective capacity without proportional compute cost.
A no-hype framework for adding AI to research, writing, support, and internal workflows without creating a second operating system to maintain.
Replacing the U-Net backbone in diffusion models with a transformer improves scaling and sample quality across image benchmarks.
Training assistants to critique and revise their own outputs against a written set of principles, reducing reliance on human preference labels.