📊 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
The latest whitepaper emphasizes that in AI-driven software development, the model itself is only about 10% of the system. The focus should be on harnesses, context engineering, and verification, which are the true sources of value and control.
A new Google whitepaper titled The New SDLC With Vibe Coding emphasizes that the AI model accounts for only about 10% of the system’s behavior. Instead, the paper argues that harnesses, context engineering, and verification are where the real value and control lie, marking a significant shift in software engineering practices as AI becomes more integrated into development workflows.
The whitepaper, authored by Addy Osmani, Shubham Saboo, and Sokratis Kartakis, reports that as of early 2026, 85% of professional developers use AI coding agents regularly, with 51% using them daily. It states that roughly 41% of all new code is generated by AI, but emphasizes that the model itself is only a small part of the overall system. The majority of the system’s behavior is shaped by the harness — including prompts, rules, tools, and observability — which constitutes about 90% of the system’s effectiveness. Concrete examples include experiments where tweaking only the harness improved agent performance significantly, while changing the model had minimal impact.
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.
Why Focus on Harnesses and Context Matters
This shift redefines where development teams should invest their resources. Instead of chasing the latest model advancements, organizations should prioritize building robust harnesses, optimizing context management, and verifying outputs. This approach can lead to lower costs, better security, and more reliable AI systems, fundamentally changing how AI-driven software is built and maintained.
AI development harness tools
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Evolution of AI in Software Development
The whitepaper builds on the recent surge in AI adoption, where by early 2026, a majority of developers integrate AI tools into their workflows. It challenges the common misconception that the model’s capabilities are the primary driver of AI performance. Previous practices focused heavily on model improvements, but emerging evidence suggests that configuration, scaffolding, and context engineering have a far greater impact. This perspective aligns with recent experiments showing that simple adjustments to prompts and tools can outperform major model upgrades.
“The biggest shift isn’t a new language or framework; it’s moving from writing code to expressing intent and trusting machines to turn that into working software.”
— Addy Osmani
software verification tools
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Unanswered Questions About Implementation and Impact
While the whitepaper presents compelling data and experiments, it remains unclear how universally applicable these findings are across different industries and scales. Specific strategies for building effective harnesses and context management at enterprise levels are still being developed, and real-world case studies are limited. Additionally, the long-term implications for AI model development priorities are still evolving.
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 Developers and Organizations
Organizations should begin evaluating their current AI workflows, focusing on harness design, context management, and verification processes. Future research and industry collaborations are expected to refine best practices for scalable harness construction. Monitoring how these principles influence AI system reliability, cost, and security will be key in the coming months.
AI observability software
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 the system’s behavior?
The whitepaper shows through experiments that the harness, prompts, tools, and context management play a much larger role in shaping AI behavior than the model itself, which is why the focus should shift there.
How does this change the way organizations should develop AI systems?
Instead of investing primarily in acquiring or upgrading models, organizations should prioritize building robust harnesses, optimizing context loading, and verifying outputs. This can reduce costs and improve system reliability.
What are harnesses and why are they so important?
Harnesses include prompts, rules, tools, and observability mechanisms that control and shape AI behavior. They are crucial because they determine how effectively an AI system performs, often more than the underlying model.
Are these findings applicable to all AI development projects?
The whitepaper presents strong evidence, but applicability may vary based on project scale, industry, and existing infrastructure. More case studies are needed to confirm universal relevance.
What should developers focus on first based on this new insight?
Developers should start by evaluating their current harnesses and context management strategies, then experiment with improvements to reduce costs and increase reliability.
Source: ThorstenMeyerAI.com
