📊 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 highlights that in AI-driven software development, the model accounts for only 10% of system behavior. The true focus should be on harness setup and context engineering, which are key to effective, cost-efficient AI applications.
A new Google whitepaper, titled The New SDLC With Vibe Coding, states that the AI model accounts for only about 10% of a system’s behavior, shifting focus toward harness design and context engineering. This insight challenges common assumptions and underscores the importance of configuration and oversight in AI development, making it a critical lesson for software engineers and organizations relying on AI tools.
The whitepaper, authored by Addy Osmani, Shubham Saboo, and Sokratis Kartakis, emphasizes that the most significant shift in software engineering is moving from writing code to expressing intent and trusting machines to interpret that intent. It reports that as of early 2026, 85% of professional developers use AI coding agents regularly, with 51% using them daily, and roughly 41% of all new code being AI-generated.
The core argument is that the model itself is a small part of the equation. Instead, the harness—the prompts, tools, rules, and observability around the model—determines system behavior much more. Experiments cited in the paper show that changing only the harness or context can dramatically improve AI performance, even with the same model. For example, one team moved a coding agent from outside the Top 30 to the Top 5 on a benchmark by optimizing only the harness.
The whitepaper advocates for a shift from vibe coding—quick prompts with minimal review—to agentic engineering, which involves structured verification, testing, and oversight. It also highlights that the economics favor investing in harness development and context engineering, as this approach reduces long-term costs and security risks compared to ad-hoc prompting.
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 perspective matters because it redefines where organizations should invest their resources. Instead of chasing the latest model improvements, companies that emphasize harness design and context management can achieve better, more reliable results at lower costs. This approach also enhances security and maintainability, reducing vulnerabilities and technical debt associated with unstructured prompt-based workflows. As AI becomes embedded in core systems, understanding that the smallest model is only a fraction of the system’s behavior is crucial for strategic planning and competitive advantage.
Building AI-Powered Products: The Essential Guide to AI and GenAI Product Management
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Background on the Evolution of AI Development Practices
Over the past few years, the AI development landscape has been dominated by rapid model improvements and hype around new architectures. However, the whitepaper contextualizes this trend by showing that actual system performance often hinges on configuration, prompts, and tooling—collectively known as the harness. Prior to this, many organizations focused on adopting the latest models, assuming that model size and complexity directly correlated with effectiveness. The paper signals a paradigm shift, emphasizing that the real challenge and opportunity lie in how these models are integrated and controlled.
“The model is only 10% of the system; the rest is harness and context.”
— Addy Osmani
Observability in the AI-Native Era: Leveraging AIOps to build, observe, and operate resilient systems
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 whitepaper provides compelling evidence and experiments, it remains to be seen how quickly and broadly organizations will adopt this paradigm shift. Specific best practices for harness design and context engineering are still evolving, and the long-term impact on AI development workflows and security standards is not yet fully understood. Additionally, the extent to which this approach applies across different AI tasks and industries is still under investigation.
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 Organizations and Developers
Organizations should evaluate their current AI workflows, focusing on harness and context management. Developing best practices, tooling, and training around these areas will be critical. Future research and industry standards are likely to emerge, guiding how to systematically implement and measure the effectiveness of harness and context engineering. Monitoring early adopters’ results will be essential to understanding the full impact of this shift.
Patriola's Guide: Cross AI Workflows: Seamless Multi-AI Tool Orchestration (Patriola's Guide to Claude Book 33)
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 surrounding harness—prompts, tools, rules, and observability—has a much larger influence on how the AI performs than the model itself. Experiments confirm that optimizing these configurations can significantly improve outcomes with the same model.
How does this change AI development strategies?
Instead of focusing solely on acquiring or developing larger models, organizations should invest in designing better harnesses and managing context effectively. This approach offers better control, security, and cost-efficiency.
What is agentic engineering?
Agentic engineering involves structuring AI workflows with formal specifications, verification, testing, and oversight—moving beyond quick prompts to disciplined, reliable systems.
Will this approach reduce AI costs?
Yes, by emphasizing harness design and context management, organizations can lower operational costs, reduce token waste, and improve security, especially compared to ad-hoc vibe coding workflows.
What are the risks of ignoring this shift?
Ignoring the importance of harness and context could lead to unreliable AI outputs, higher maintenance costs, security vulnerabilities, and missed opportunities for optimizing system performance.
Source: ThorstenMeyerAI.com
