📊 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-driven software development, the model itself accounts for only 10% of system behavior. The real focus should be on harness design and context engineering, which dominate performance and cost.
A new Google whitepaper titled The New SDLC With Vibe Coding emphasizes that the core of AI-driven software development is not the model itself, but the harness and context engineering surrounding it. The document states that the model accounts for only about 10% of the system’s behavior, with the remaining 90% determined by configuration, verification, and judgment, marking a significant shift in how organizations should approach AI integration.
The whitepaper, authored by Addy Osmani, Shubham Saboo, and Sokratis Kartakis, argues that the dominant factor in AI system performance is the harness—the prompts, tools, rules, and observability layers built around the model. Concrete experiments cited in the paper show that modifying only the harness can dramatically improve an agent’s performance, often more than upgrading the model itself.
Additionally, the paper emphasizes the importance of context engineering: providing the right instructions, knowledge, memory, and guardrails to the AI. The authors introduce the concept of Agent Skills, which involves loading procedural knowledge only when necessary, enabling more efficient and scalable AI systems.
From an economic perspective, the whitepaper states that ad-hoc prompting (vibe coding) appears inexpensive upfront but incurs high ongoing costs, including token burn, maintenance, and security risks. Conversely, disciplined, structured approaches—referred to as agentic engineering—require higher initial 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.
Impact of Harness and Context on AI Development
This shift in focus from the model to the harness and context engineering has major implications for organizations adopting AI. It suggests that building durable, configurable scaffolding around AI systems offers a competitive advantage, as it allows for continuous improvement and cost management. The insight challenges the common perception that upgrading to the latest model is the primary driver of AI performance, instead highlighting the importance of system configuration and verification.
For leaders, this means prioritizing investments in system architecture, tooling, and process discipline rather than solely chasing new models, which are now only a small part of the overall system.
AI Model Risk Blueprint: Model Validation Testing | Ethical Considerations in AI Models | Integrating AI with Business Risk Plans | Real-World AI Model … Strategies | AI Governance Tools & Resource
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Evolution of AI System Design and Cost Structures
The whitepaper builds on recent trends where AI adoption has skyrocketed, with 85% of professional developers using AI coding agents as of early 2026. Previously, the focus was on acquiring more powerful models; now, the emphasis is shifting toward how these models are integrated and controlled.
Past developments include the rise of vibe coding—quick, minimal prompts used for rapid prototyping—and the move toward more structured, disciplined engineering practices. Experiments cited in the paper show that even when using the same model, performance can vary dramatically based on configuration, tools, and context management.
This evolution reflects a broader understanding that AI systems are complex, configurable ecosystems, not just isolated models, and that system-level design determines success more than raw model capabilities.
“The biggest shift in software engineering isn’t a new language or framework; it’s moving from writing code to expressing intent and trusting machines to turn that intent into software.”
— Addy Osmani
AI Engineering: Building Applications with Foundation Models
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Unclear Aspects of Implementation and Cost Savings
While the paper provides compelling evidence that harness and context are dominant, it remains to be seen how widely organizations will adopt this approach and how it will impact long-term costs and system robustness. Specific strategies for scaling these practices across different industries are still developing, and real-world case studies are limited.
AI-Powered Observability: From Noise to Insight: Transforming How We Monitor, Detect, and Respond
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Next Steps for AI System Optimization
Organizations are likely to begin investing more heavily in system architecture, tooling, and verification processes. Further research and case studies will clarify best practices for harness design and context engineering at scale. Industry leaders may also develop new standards and frameworks to guide disciplined AI development, emphasizing configuration over model upgrades.
AI verification and testing platforms
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 that the model’s core generative capabilities are only a small part of what determines overall behavior; the surrounding harness, prompts, tools, and configuration dominate system performance and reliability.
How does this shift affect AI development costs?
While adopting complex harnesses and structured processes requires higher upfront investment, it reduces ongoing costs related to token consumption, maintenance, and security vulnerabilities, making AI systems more cost-effective over time.
What is meant by ‘agentic engineering’?
Agentic engineering refers to designing AI systems with structured, verified components, including schemas, tools, and guardrails, to improve reliability and control, moving away from ad-hoc prompt-based methods.
Will organizations shift focus from models to system design?
Yes, the whitepaper suggests that system configuration, harness design, and context engineering are now the primary areas for strategic investment and competitive advantage.
What remains uncertain about this approach?
It is still unclear how quickly and broadly organizations will implement these practices, and how they will perform at scale across different industries and use cases.
Source: ThorstenMeyerAI.com
