In the news
GPT-6 Astra, Looped Transformers, and Hidden Reasoning
Sebastian Raschka · Sebastian Raschka, PhD · Published · 3 min read
In 30 seconds
- What happened
- OpenAI released GPT-6 Astra, which uses looped transformers that reuse the same transformer blocks multiple times instead of stacking unique blocks.
- Why it matters
- Engineers building or optimizing large language models need to understand this architectural approach for reasoning and inference efficiency improvements.
- Watch out
- Looped transformers are not new; the concept dates to 2018 Universal Transformers. Claims about hidden reasoning chains require careful evaluation beyond architectural changes alone.
- reasoning
- gpt
The patterns behind this
- Energy-Efficient Inference
- Agentic Context Engineering (Evolving Playbook)
- Query Transformation Retrieval
Each one covers how the technique works, when it earns its cost, and where it breaks.
The Agent Architect
One pattern, one tradeoff, one production failure story. A short weekly briefing for people building agentic systems.
Weekly email, one-click unsubscribe. We only use your address to send the briefing.