📊 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 Google whitepaper reveals that in AI-driven software development, the model accounts for only 10% of system behavior. The focus should shift to harness design and context engineering, which have greater impact and cost implications.
A recent Google whitepaper titled The New SDLC With Vibe Coding highlights a counterintuitive but crucial insight: the AI model itself accounts for only about 10% of the system’s behavior. The paper underscores that the real control lies in the harness and context engineering, which together determine 90% of outcomes. This shift has significant implications for how organizations develop and manage AI systems, emphasizing the importance of configuration, scaffolding, and strategic context over the choice of model alone.
The whitepaper, authored by Addy Osmani, Shubham Saboo, and Sokratis Kartakis, argues that the dominant factor in AI system performance is how the model is integrated and guided through the harness. Experiments cited show that changing only the harness — prompts, tools, rules, and observability — can dramatically improve performance, even with the same underlying model. For example, a team improved a coding agent’s ranking from outside the Top 30 to within the Top 5 by adjusting only the harness, not the model itself.
The paper introduces the concept of agentic engineering: a disciplined approach that involves structured context, verification, and judgment, contrasting with vibe coding, which relies on quick prompts and minimal review. It emphasizes that costs and risks associated with AI development are primarily driven by configuration, maintenance, and security, not the raw model size or capabilities. The authors warn that many failures are configuration issues, not model failures, and that the durable advantage lies in harness design and context management.
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 Harness and Context Are More Critical Than the Model
This insight shifts the focus from chasing the latest model to investing in system architecture, configuration, and context engineering. It suggests that organizations can achieve better performance and cost efficiency by mastering harness design and context management, rather than simply upgrading models. This approach reduces long-term operating costs, enhances security, and provides a competitive advantage through tailored system behavior. For developers and leaders, it underscores the importance of strategic system design over model selection alone.
MUCAR 892BT AI-Assisted Bidirectional Scan Tool, Full System OBD2 Scanner, Bi-Directional OBD2 Scanner Diagnostic Tool,ECU Coding, 35 Services, FCA Autoauth, CANFD and DOIP, Free Lifetime Upgrade
【Powerful Performance】: OBD2 scanner, featuring an 8-inch ultra-large display, the MUCAR 892BT runs on Android 10 with a…
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Background of the SDLC and AI Development Shifts
Traditional software development has long focused on code quality, architecture, and testing. With AI, especially large language models, the landscape has shifted toward integrating models into systems through prompts, tools, and rules. The whitepaper builds on recent trends where 85% of developers use AI coding agents, and 41% of code is AI-generated, emphasizing that the model is just one component. The core challenge now is how to structure, verify, and control AI behavior, which the authors argue is more impactful than model improvements alone.
“The model you’re paying so much attention to is the smallest part of the system.”
— 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 Industry Adoption
While the paper presents compelling evidence that harness and context are critical, it is still unclear how organizations will systematically implement these strategies at scale. The precise methods for measuring and optimizing harness design, and how quickly these practices will become standard, remain to be seen. Additionally, the long-term impact on model development priorities and industry standards is still developing.
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 Development and System Design
Organizations are likely to begin investing more in harness and context engineering, developing best practices and tooling for system configuration, verification, and security. Future research and industry standards may focus on formalizing these practices, with potential shifts in AI development budgets and team structures. Monitoring how these strategies influence system performance, cost, and security over the coming months will be key.
Designing Instruction with Generative AI: 24/7 Support for Optimizing Teaching and Learning
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 system behavior?
The whitepaper demonstrates that most of an AI system’s behavior is determined by how the model is integrated, guided, and constrained through prompts, tools, and rules — collectively called the harness. Experiments show that tweaking these elements can significantly outperform simply upgrading the model.
What is agentic engineering?
Agentic engineering is a disciplined approach that involves designing structured contexts, verification methods, and judgment frameworks to control AI behavior, moving beyond quick prompt-based vibe coding.
How does this shift affect AI development costs?
Focusing on harness and context can reduce long-term costs by minimizing token waste, improving security, and decreasing maintenance, despite higher initial investment in system design.
Will this change how models are built?
The focus is shifting from model size and capabilities to system integration and control. Model development may still advance, but system design and configuration will become more central to performance and safety.
When will industry standards adopt these practices?
It is still uncertain, but early adoption by leading organizations and ongoing research suggest that harness and context engineering will become standard practice within the next year or two.
Source: ThorstenMeyerAI.com
