📊 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 the core of modern software development isn’t the AI model, but the surrounding harness and context engineering. The model accounts for only about 10% of the system’s behavior, shifting focus to configuration, verification, and judgment.
A new Google whitepaper titled The New SDLC With Vibe Coding states that the AI model constitutes only about 10% of the overall system behavior, shifting the focus to the harness and context engineering. This challenges the common perception that advances in models alone will drive software innovation, emphasizing instead the importance of configuration, verification, and judgment in AI development.
The whitepaper, authored by Addy Osmani, Shubham Saboo, and Sokratis Kartakis, reports that 85% of professional developers now use AI coding agents regularly, with 51% doing so daily. It notes that roughly 41% of all new code is AI-generated, but the key insight is that the model’s role is minor.
The authors argue that the behavior of AI agents depends predominantly on the harness — the prompts, rules, tools, and context around the model — which accounts for about 90% of the system’s effectiveness. Evidence from benchmarks shows that changing only the harness can significantly improve performance, even with the same model.
They emphasize that cost and reliability are driven more by how the AI is configured and integrated than by the choice of model, advocating for a disciplined approach called agentic engineering that emphasizes verification, structured context, and continuous oversight.
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 redefines how organizations should invest in AI. Instead of chasing the latest model, companies should focus on building robust harnesses and context management. This approach can lead to more cost-effective, reliable, and secure AI systems, as the whitepaper highlights that most failures are configuration issues rather than model limitations. It suggests that the future of AI in software engineering lies in structured engineering practices rather than solely in model innovation.
AI code verification tools
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Background on Evolving AI Development Practices
Over the past year, the AI industry has heavily promoted new models as the primary drivers of innovation. However, the whitepaper from Google signals a paradigm shift, emphasizing that the real work is in how models are integrated and controlled. The concept of vibe coding — quick prompts with minimal oversight — is contrasted with agentic engineering, which involves formal specs, testing, and verification. This reflects a broader trend toward disciplined AI practices that prioritize reliability and cost-efficiency.
Prior to this, the industry largely viewed model improvements as the main lever for progress, but recent experiments show that tweaking harnesses and context can yield greater performance gains at lower costs.
“The biggest shift isn’t the model itself but how we engineer around it, focusing on verification and context.”
— Addy Osmani
AI development environment with context engineering
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Uncertainties About Practical Implementation
It remains unclear how quickly organizations will adopt this disciplined approach at scale, and what specific tools or frameworks will emerge to support it. Further research is needed to quantify the cost savings and reliability improvements across different industries and use cases.automated testing software for AI systems
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Next Steps for AI-Driven SDLC Adoption
Organizations are likely to begin investing more in developing structured harnesses, verification protocols, and context management tools. Industry standards and best practices may evolve to emphasize configuration and testing over model selection. Future research and case studies will clarify how this approach impacts productivity, cost, and security in real-world deployments.
AI model harness configuration tools
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 indicates that the model’s behavior is heavily influenced by how it is configured, with the harness, prompts, tools, and context playing a larger role in determining output.
What is agentic engineering?
Agentic engineering refers to a disciplined approach that involves formal specs, verification, structured context, and oversight, moving beyond quick prompts to reliable AI systems.
How does this shift affect AI development costs?
Focusing on harness and context engineering can reduce costs by decreasing token burn, improving reliability, and lowering maintenance and security expenses, compared to vibe coding approaches.
Will this change how AI tools are built and sold?
Yes, it suggests that tools and frameworks supporting harness creation, context management, and verification will become more central, shifting value away from model innovation alone.
What are the main challenges in adopting this approach?
The main challenges include developing standardized practices, training teams in structured engineering, and integrating these methods into existing workflows.
Source: ThorstenMeyerAI.com
