📊 Full opportunity report: The Model Is Only 10%: The Real Lesson of the New SDLC on ThorstenMeyerAI.com — validation score, market gap, and execution plan.
TL;DR
A recent whitepaper from Google emphasizes that in AI-assisted software development, the model itself is only about 10% of the system’s behavior. The real focus should be on harness design and context engineering, which account for 90%. This shift has significant implications for how companies approach AI integration.
A new Google whitepaper emphasizes that in AI-driven software development, the model constitutes only about 10% of the system’s behavior. Instead, harness and context engineering are the main determinants of performance. This insight challenges the common focus on model size and suggests a strategic shift for developers and organizations adopting AI tools.
The whitepaper, authored by Addy Osmani, Shubham Saboo, and Sokratis Kartakis, states that the biggest shift in software engineering is moving from writing code to expressing intent and trusting machines to interpret that intent. As of early 2026, 85% of professional developers use AI coding agents regularly, with 51% using them daily and about 41% of all new code being AI-generated.
The core argument is that the model itself is only a small part of the system’s behavior. The majority of performance depends on the harness—the prompts, tools, rules, and observability layers surrounding the model. Experiments cited in the paper show that changing only the harness or prompts can dramatically improve AI agent performance, even with the same underlying model.
The paper emphasizes that cost efficiency depends heavily on how the AI system is structured. While vibe coding—quick prompts with minimal review—seems inexpensive, it actually incurs high ongoing costs due to token inefficiency and maintenance. Conversely, disciplined, agentic engineering involves higher upfront investment but lower marginal costs over time.
The model is only 10%
A Google whitepaper argues software’s biggest shift is from writing code to expressing intent. Its sharpest claim: the model you obsess over is the smallest part of the system — the scaffolding around it does the real work.
The clearest map yet of how serious AI development works — and mostly tool-agnostic. But it’s a Google funnel: the concepts are neutral, the on-ramps point to Gemini, Jules & the ADK. If the harness is 90% and it’s yours, your moat and your costs both live there — so own your scaffolding, route across models, and remember: AI amplifies whatever engineering culture it lands in.
Implications for AI Development Strategies
This shift means organizations should prioritize harness design and context engineering over simply acquiring larger models. The focus on configuration, tools, and structured context can lead to more cost-effective and reliable AI systems. It challenges the industry narrative that model size correlates directly with performance and encourages a reevaluation of AI investment strategies.
For developers and CTOs, this underscores the importance of building flexible scaffolding around AI models, which can be owned and improved over time, rather than relying solely on the latest, largest models. This approach can provide a durable competitive advantage and mitigate risks associated with rapid model obsolescence.
The AI Prompt Playbook: Master AI Prompt Engineering with 140 Ready-to-Use Templates for ChatGPT, Claude, Gemini & Copilot
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Industry Shift Toward Harness and Context Engineering
The whitepaper builds on ongoing trends where AI adoption has accelerated, with a significant portion of code now generated by AI tools. Historically, the focus has been on model improvements—larger, more capable models. However, recent experiments and industry feedback indicate that the behavior of AI systems is more heavily influenced by how they are configured and the quality of their contextual information.
This aligns with broader industry movements toward modular AI architectures and cost management, emphasizing the importance of configuration, tooling, and dynamic context loading. The paper explicitly states that model upgrades alone are insufficient without corresponding improvements in harness and context strategies.
“The model is only about 10% of what determines behavior; the harness and context are the real drivers.”
— Addy Osmani
Observability in the AI-Native Era: Leveraging AIOps to build, observe, and operate resilient systems
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Unresolved Questions About Implementation and Industry Adoption
While the whitepaper provides compelling evidence and strategic guidance, it remains unclear how quickly organizations will adopt these insights at scale. Specific best practices for harness design and context engineering are still evolving, and the industry lacks standardized frameworks for measuring success in these areas. Additionally, the long-term impact of this shift on AI model development and hardware investment is still being evaluated.
AI-Powered Web Design Mastery: Harness the Power of Framer AI to Build, Customize, and Launch Stunning Websites—A Step-by-Step Guide for Beginners and Professionals
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Next Steps for Organizations and Developers
Organizations should begin reassessing their AI development processes, focusing on building and owning their harnesses and refining context engineering practices. Industry leaders are likely to develop new tools and frameworks to support this approach. Meanwhile, further research and case studies are expected to emerge, clarifying best practices and quantifying performance gains. Companies that adapt quickly may gain a significant strategic advantage in AI deployment.
Agentic Context Engineering: Master Agentic Context Engineering Through Example Stories (Generative AI Mastery)
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
Why is the model size less important than the harness?
The whitepaper shows that behavior and performance depend more on how the AI system is configured, including prompts, tools, and context, than on the underlying model size. Experiments confirm that tuning the harness can dramatically improve results.
What does this mean for companies investing in larger models?
It suggests that investment in larger models alone may be less effective than focusing on system configuration, tooling, and context management. Building durable, configurable harnesses can provide a better return on investment.
How can organizations improve their harness and context engineering?
Organizations should develop structured frameworks for prompt design, tool integration, and dynamic context loading. Training teams in configuration best practices and adopting modular architectures are key steps.
Is this shift applicable across all AI applications?
While most AI systems can benefit from better harness and context engineering, the extent varies depending on the use case. Critical applications requiring high reliability will especially benefit from disciplined configuration.
Source: ThorstenMeyerAI.com
