📊 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 highlights that in AI-assisted software development, the core challenge is not the AI model itself but the surrounding harness and context. This shift impacts how companies should invest in AI tools and development strategies.
A Google whitepaper released in early 2026 states that the AI model constitutes only about 10% of the behavior in AI-assisted coding systems. The report emphasizes that the real control lies in the harness and context engineering, which account for roughly 90%. This insight suggests that companies should focus more on configuring and managing their AI systems rather than solely investing in the latest models.
The whitepaper, authored by Addy Osmani, Shubham Saboo, and Sokratis Kartakis, argues that the traditional focus on acquiring or upgrading AI models is misplaced. Instead, the behavior of AI agents depends heavily on the surrounding harness — prompts, rules, tools, and observability. Evidence from benchmark experiments shows that changing the harness can significantly improve performance without upgrading the model itself. For example, moving a coding agent from outside the Top 30 to the Top 5 was achieved solely through harness adjustments, with the same model.
The report introduces the concept of agentic engineering, where AI is integrated into formal specifications, automated testing, and oversight, contrasting with the more casual vibe coding approach. The authors stress that cost and effectiveness are driven more by configuration and context management than by the AI model, urging organizations to invest in these areas for long-term advantages.
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 in understanding impacts how organizations should approach AI integration. Focusing on harness and context engineering can lead to more reliable, secure, and cost-effective AI systems. It challenges the notion that adopting the latest AI model is the key to success, highlighting instead the importance of configuration, verification, and ongoing management. As AI becomes embedded in critical workflows, these insights could redefine best practices and investment priorities in the industry.
AI harness configuration tools
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Evolution of AI Coding Practices and Industry Insights
The whitepaper builds on recent trends where AI-generated code now makes up approximately 41% of new code, with 85% of developers regularly using AI agents. The narrative shifts from vibe coding — quick prompts with minimal review — toward agentic engineering, which emphasizes formal specifications, testing, and oversight. This evolution reflects a broader understanding that the model itself is only a small part of the system’s behavior, with the majority influenced by how the AI is configured and managed.
Historically, the industry has focused on model improvements, but recent experiments show that tuning the harness can yield performance gains comparable to or exceeding those from model upgrades. This understanding is reshaping development priorities and costs, with a focus on long-term system stability and security.
“The behavior you experience in AI coding tools is dominated by scaffolding you can build, own, and improve — not the model provider.”
— 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
It is not yet clear how quickly organizations will shift their focus toward harness and context engineering, or how this approach will vary across different industries and use cases. Specific best practices and standards are still emerging, and the long-term impact on AI development costs and security remains to be fully understood.
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.
Next Steps for Organizations and Developers in AI Integration
Organizations should evaluate their current AI workflows, emphasizing the configuration, tooling, and oversight mechanisms. Developing best practices for harness design and context management will be critical. Industry-wide, expect increased investment in formal specifications, testing frameworks, and security measures to support this paradigm shift. Further research and case studies will likely clarify how to optimize these strategies for different applications.
AI Assignments and Assessment for Teachers: Create Quizzes, Tests, Feedback, and Grading Systems in Minutes Using ChatGPT (No Tech Skills Required) (Classroom AI Success Series)
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
Why is the model only 10% of AI system behavior?
The whitepaper argues that the behavior of AI systems is primarily determined by how they are configured and managed — the harness and context — rather than the underlying model itself. Experiments show that tuning these elements can significantly improve performance without changing the model.
What is agentic engineering?
Agentic engineering refers to integrating AI systems into formal specifications, automated testing, and oversight processes, making them more reliable and controllable compared to casual or vibe coding approaches.
How does this shift affect AI development costs?
Focusing on harness and context engineering may require higher initial investment but leads to lower marginal costs, better security, and more predictable performance over time, reversing the traditional cost curve of AI development.
Will this change how AI tools are marketed?
Yes, the emphasis is moving from model performance to configuration, security, and verification, which could influence how vendors position their AI solutions and what features they highlight.
What should organizations do now?
They should assess their current AI workflows, prioritize building robust harnesses, improve context management, and invest in testing and security frameworks to adapt to this new paradigm.
Source: ThorstenMeyerAI.com
