New review paper argues code is how AI agents think and act, not just what they produce
Quick Answer
A new review paper posits that the software layer around AI models is the key to their functionality, not the models themselves.
Quick Take
A new review paper posits that the software layer around AI models is the key to their functionality, not the models themselves. Deepseek is establishing a 'Harness' team in Beijing, emphasizing that the combination of a model and its harness creates an effective AI agent.
Key Points
- The software layer includes tools, memory, testing, and permission boundaries.
- Deepseek's 'Harness' team aims to enhance AI agent capabilities.
- The thesis suggests that a model plus harness equals a functional AI agent.
- The review challenges the notion that language models alone drive AI behavior.
Article Excerpt
From source RSS / original summaryA new review paper argues that the real bottleneck for autonomous AI agents isn't the language model itself but the software layer wrapped around it. Tools, memory, testing, and permission boundaries turn a stateless model into a working agent. Deepseek is already building a dedicated "Harness" team in Beijing with a core formula that confirms the thesis: model plus harness equals AI agent.
The article New review paper argues code is how AI agents think and act, not just what they produce appeared first on The Decoder.
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from The Decoder
See more →
An AI model programmed nonstop for 19 days on a single MirrorCode task that cost $2,600 to run
Epoch AI's MirrorCode benchmark reveals Claude Opus 4.7 as the leader with a 56% solve rate, reconstructing a 16,000-line toolkit in 14 hours. Despite this, all models tested struggle with the most complex tasks, highlighting limitations in current AI capabilities. The single task consumed $2,600 over 19 days, raising questions about cost-effectiveness in AI development.
