📊 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 highlights that in AI-driven software development, the model accounts for only 10% of system behavior. The focus shifts to harness design and context engineering, which significantly influence outcomes and costs.
A Google whitepaper published in early 2026 states that the AI model constitutes only about 10% of the behavior of AI-powered systems. The document argues that the real value lies in harness design and context engineering, which determine system performance and cost efficiency. This shifts the focus from developing larger models to optimizing the surrounding infrastructure and processes, a change with significant implications for software engineering teams.
The whitepaper, authored by Addy Osmani, Shubham Saboo, and Sokratis Kartakis, emphasizes that the dominant factor in AI system behavior is not the model but the harness—the prompts, rules, tools, and observability layers built around the model. Concrete experiments cited show that simply tweaking the harness can dramatically improve performance, sometimes by over 13 points on benchmark scores, even with the same model.
The authors introduce the concept of agentic engineering, where AI is integrated within a framework of verification, testing, and structured context, rather than vibe coding—quick prompts with minimal oversight. They argue that verification and judgment are now the new craft in AI development, replacing reliance solely on the model’s raw capabilities. The whitepaper also highlights that cost considerations favor investing in harness and context engineering, as ad-hoc prompting can be 3-10 times more expensive over time due to inefficiencies and security risks.
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 Harness and Context Engineering
This shift in emphasis from models to harness and context has broad implications for software development strategies. Companies that prioritize optimizing their harness and structuring the context effectively can achieve better performance at lower costs, even with smaller models. It also means that long-term competitive advantage will come from mastering these areas, rather than chasing the latest model frontier. For developers and leaders, understanding this paradigm change is crucial to managing costs, security, and system reliability in AI-driven projects.
AI-Powered Web Design Mastery: Harness the Power of Framer AI to Build, Customize, and Launch Stunning Websites—A Step-by-Step Guide for Beginners and Professionals
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Background of the AI Development Shift
Since 2025, AI adoption in software engineering has rapidly increased, with reports indicating that 85% of developers use AI coding agents regularly. The industry has often focused on acquiring larger models, believing that bigger models yield better results. However, recent experiments and industry insights suggest that the performance gains are largely driven by how the AI is integrated and controlled. The whitepaper builds on this evolving understanding, framing the AI development process as one where the harness and context are more critical than the model itself.
This perspective aligns with ongoing industry debates about cost, security, and system reliability, emphasizing that investment in infrastructure and process design can yield more sustainable benefits than solely chasing model size improvements.
“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 interpret that intent.”
— Addy Osmani
Non-Deterministic Software Engineering: How to Build Reliable Software with AI Assistants Without Losing Quality, Security, or Control
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Unresolved Questions About Practical Implementation
While the whitepaper emphasizes the importance of harness and context engineering, it remains unclear how organizations will effectively scale these practices across diverse projects. Specific methodologies, best practices, and tooling support for widespread adoption are still emerging. Additionally, the long-term impact on model development priorities and industry standards is yet to be fully understood.
AI Harness Engineering Interview Preparation Handbook: Runtime, Control Layer, Guardrails, MCP, Evals, and Observability for Production AI Agents
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Next Steps for Software Teams and Industry Leaders
Organizations should begin evaluating their current AI workflows, focusing on harness configuration and context management. Investing in training and tools that facilitate structured engineering practices will be critical. Industry forums and standards bodies are likely to develop guidelines around best practices in harness design and verification, shaping future AI development norms. Monitoring how leading firms adapt to this shift will inform broader adoption strategies.
AI Chat Pen for Tests | Smart Study Tool with Integrated Scanner | Answer Questions in Math & More | Perfect for Students & Travelers | AI-Powered Learning Aid (1Set)
【Effortless Digitization】Easily convert physical books, documents, and handwritten notes into clear, searchable digital files with the AI Smart…
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 most AI system behavior depends on how the model is integrated, controlled, and verified through harness design and context engineering, not just the model itself.
What is harness engineering?
Harness engineering involves designing prompts, rules, tools, and observability layers that surround and control the AI model, significantly impacting its output and reliability.
How does this shift affect AI development costs?
Focusing on harness and context can reduce long-term costs by improving efficiency, security, and system performance, despite higher initial investment in structured engineering.
Will larger models become obsolete?
Not necessarily, but the whitepaper suggests that improvements in harness and context engineering can deliver more value than simply increasing model size, making strategic focus on these areas more cost-effective.
What should organizations do now?
Organizations should assess their current AI workflows, invest in developing robust harnesses, and adopt best practices for context management to stay competitive in AI-driven development.
Source: ThorstenMeyerAI.com
