Startups Chase Next Big LLM Architectures Beyond Transformers
ryanallred · x · 2026-08-11
MIT Technology Review explores the bottlenecks of the Transformer architecture underlying current LLMs. While Transformers support the entire AI industry, recent advances in reasoning and long contexts are largely workarounds for its flaws. A wave of startups, like Subquadratic, is attempting to build entirely new underlying architectures to replace the powerful but inefficient model.
Related event: Startups Race to Challenge Transformer with Next-Gen LLM Architectures(3 posts)→
More from Companies & People
- Ex-Kimi Co-founder JustinLin Launches Pragmatik Labs, Backed by Sequoia & Tencent — JustinLin610 · 2026-08-12
- Palantir CTO: AI Boosts Productivity and Drives Hiring in Supply-Constrained Economy — MartinGTobias · 2026-08-12
- OpenAI Co-founder Brad Lightcap Departs to Start New Venture — bradlightcap · 2026-08-12
- Microsoft VP: AI Will Radically Change Software Engineering, Devs Need Editorial Judgment — clamanna · 2026-08-12
- Oumi AI Highlights Enterprise AI Paradox, Advocates for Specialized Intelligence — prfsanjeevarora · 2026-08-12
- OpenAI to Co-host Git Merge 2026 in Lisbon This September — charliermarsh · 2026-08-12