📊 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 emphasizes that AI models are only a small part of the software lifecycle. The focus is shifting toward harnessing, verification, and context engineering, which are more critical for effective AI deployment.
A new whitepaper from Google, authored by Addy Osmani, Shubham Saboo, and Sokratis Kartakis, states that the model constitutes only about 10% of an AI system’s behavior. The report argues that the real focus should be on harnessing, configuration, and verification, which account for the remaining 90%. This shift has significant implications for how organizations develop and deploy AI systems, emphasizing the importance of system design over model selection alone.
The whitepaper, titled The New SDLC With Vibe Coding, challenges the common perception that upgrading to the latest AI model automatically improves system performance. Instead, it highlights that most failures and inefficiencies stem from configuration errors, missing tools, or poor context management. Experiments cited in the paper show that changing the harness—such as prompts, tools, and middleware—can drastically improve AI performance, even with the same underlying model. For example, one team moved a coding agent from outside the Top 30 to the Top 5 on a benchmark by tweaking only the harness, not the model.
The authors advocate for a disciplined approach called agentic engineering, where AI is embedded within a framework of verification, structured context, and continuous oversight. This approach contrasts sharply with vibe coding, which involves minimal review and quick prompts, often leading to higher long-term costs and security risks. The paper emphasizes that costs are driven more by how AI is configured and used than by the model itself. High upfront investments in system design and context engineering yield lower marginal costs over time, making AI deployment more sustainable and secure.
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 System Design and Configuration Matter More Than the Model
This whitepaper shifts the AI development paradigm, highlighting that system configuration, harnessing, and verification are more impactful than simply upgrading models. For organizations, this means that investing in system architecture, tools, and context management can lead to better performance, lower costs, and enhanced security. It challenges teams to focus on building durable, configurable frameworks rather than chasing the latest model versions, which often provide diminishing returns without proper configuration.
LEAN PROGRAMMING FOR FORMAL SOFTWARE VERIFICATION: Mathematical proof systems and logical frameworks for verified computation
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Shift Toward System-Centric AI Development
As of early 2026, AI adoption is widespread, with 85% of professional developers using AI coding agents regularly, and 41% generating most of their code via AI. Previously, the emphasis was on acquiring the most advanced models; now, the focus is on how those models are integrated, configured, and verified within larger systems. The whitepaper builds on ongoing discussions about the total cost of ownership in AI, emphasizing that the model itself is only a small component of the overall system performance and security.
This perspective aligns with recent experiments demonstrating that tweaks to prompts, tools, and context loading can significantly outperform simply upgrading to newer models. The authors argue that the real skill lies in system design and context engineering, which has been historically undervalued.
“The model constitutes only about 10% of an AI system’s behavior; the rest is determined by harnessing, configuration, and verification.”
— Addy Osmani
YAML Made Simple: A Beginner’s Guide to Configuration and Data Structuring
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
What Aspects of System Configuration Are Still Unclear
While the whitepaper provides strong evidence that configuration and harnessing are critical, it remains unclear how organizations can best scale these practices across complex systems. Specific methodologies for systematically measuring and improving harness design are still evolving. Additionally, the long-term impact of this shift on AI model development cycles and industry standards has yet to be fully established.
Context Engineering for Claude: A Practitioner's Guide to CLAUDE.md, Memory Tools, and Three-Layer Workflows for Solopreneurs, Freelancers, and Product Managers
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Next Steps for AI Development and System Optimization
Organizations are likely to invest more in system architecture, tooling, and context engineering practices. Expect to see increased focus on developing frameworks for verifying and maintaining AI systems, along with new standards for harness design. Researchers and practitioners will explore best practices for scalable context management and automated configuration, shaping the future of AI engineering. Further studies are anticipated to quantify the cost savings and performance improvements achievable through these approaches.
AI Model Validation & Testing: Ensuring Reliable AI Systems — Bias Testing, Robustness Evaluation & Regulatory Compliance (AI Compliance Toolkit)
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 AI system’s behavior?
The whitepaper explains that most of an AI system’s behavior depends on how it is configured, harnessed, and verified, not just the underlying model. The model provides a base, but the surrounding system determines how effectively it performs.
How can organizations improve their AI systems based on this insight?
By investing in system design, including tools, prompts, context management, and verification processes, organizations can significantly enhance AI performance without always needing the latest model upgrades.
Does this mean model upgrades are no longer important?
Model upgrades remain valuable but are now seen as one part of a broader system strategy. The whitepaper emphasizes that system configuration and harness design often have a larger impact on results.
What are the risks of focusing too much on system configuration?
Over-reliance on configuration without understanding foundational model capabilities could lead to security vulnerabilities or maintenance challenges. Balanced development that includes ongoing model evaluation is recommended.
What is agentic engineering?
Agentic engineering involves embedding AI within a framework of structured verification, tools, and context management, enabling more reliable and scalable AI deployment.
Source: ThorstenMeyerAI.com
