The Model Is Only 10%: The Real Lesson of the New SDLC

📊 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 emphasizes that in AI-assisted software development, the focus should shift from the AI model to the surrounding harness and context engineering. The model itself is only about 10% of system behavior, with verification and configuration being more crucial.

A new Google whitepaper, The New SDLC With Vibe Coding, emphasizes that the most significant shift in software engineering is moving from writing code to expressing intent and trusting machines to generate software. The paper states that the model accounts for only about 10% of system behavior, with the remaining 90% determined by the harness, configuration, and context engineering. This insight challenges traditional focus areas and suggests a strategic pivot for development teams.

The whitepaper, authored by Addy Osmani, Shubham Saboo, and Sokratis Kartakis, reports that 85% of professional developers now use AI coding agents regularly, with 51% using them daily. Additionally, approximately 41% of all new code is generated by AI. Despite the hype around models, the paper argues that the true differentiator lies in how these models are integrated and managed through the harness, which includes prompts, tools, rules, and observability.

Concrete evidence from experiments cited in the paper shows that changing the harness—such as prompts, tools, or middleware—can significantly improve an AI agent’s performance, even when using the same underlying model. For example, a team improved a coding agent’s ranking from outside the top 30 to the top 5 by adjusting only the harness. This underscores that most failures are configuration issues rather than model deficiencies.

At a glance
reportWhen: published early 2026
The developmentThe publication of a Google whitepaper highlights that the primary driver of AI system performance is not the model but the harness and context engineering, redefining best practices in SDLC.
The Model Is Only 10% — The New SDLC With Vibe Coding
AI Dispatch · Field Notes
Google · Osmani, Saboo & Kartakis · May 2026

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.

A spectrum, not a binary — the differentiator is how outputs get verified
Vibe Coding
Casual prompts · “does it seem to work?” · disposable code · high risk
Structured AI-Assisted
Detailed prompts + constraints · manual testing · features in real codebases
Agentic Engineering
Formal specs · automated tests + evals + CI gates · production scale · low risk
Tests verify the deterministic; evals verify the rest. Without both, it’s vibe coding — however clever the prompt.
The idea worth building your strategy around
Agent = Model + Harness
~10%
HARNESS — prompts · tools · context · hooks · sandboxes · observability
MODEL~90% IS YOUR SURFACE AREA, NOT THE PROVIDER’S
Outside Top 30 → Top 5 on Terminal Bench 2.0 by changing only the harness — same model.
“Most agent failures, examined honestly, are configuration failures” — a missing tool, a vague rule, a noisy context.
The economics: it’s a token-cost problem (CapEx vs OpEx)
Vibe Coding
Low CapEx · High OpEx
Looks free, hides debt: token burn (fix-it loops), maintenance tax (AI spaghetti), security remediation. Crosses over to 3–10× more per feature.
Agentic Engineering
High CapEx · Low OpEx
Pay upfront (specs, evals, context), then ship cheaply. Levers: context engineering for first-pass success + intelligent model routing — cheap models for the easy work.
85%
of devs use AI coding agents (51% daily)
41%
of all new code is AI-generated
~90%
of agent behavior is the harness, not the model
+19%
longer on some tasks (METR) — verification is the cost
The read

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.

Source: Osmani, Saboo & Kartakis, “The New SDLC With Vibe Coding,” Google (May 2026). Figures are the paper’s own, incl. METR & LangChain. Analysis is the author’s.
thorstenmeyerai.com

Why Focusing on Harness and Context Matters

This shift has major implications for AI development strategies. By recognizing that 90% of behavior is determined by how the AI is configured and integrated, organizations can achieve better performance and cost efficiency by investing in harness development and context engineering. It also means that competitive advantage comes from mastery over these aspects, not just access to the latest model.

Moreover, the paper warns that relying solely on the model can lead to higher costs, increased maintenance, and security vulnerabilities. Proper harnessing and context management enable scalable, reliable, and secure AI systems, which are critical as AI becomes central to software development.

Amazon

AI development harness tools

As an affiliate, we earn on qualifying purchases.

As an affiliate, we earn on qualifying purchases.

Evolution of AI in Software Development

The whitepaper builds on the recent surge in AI-assisted coding, where 85% of developers now use AI agents, and 41% of new code is AI-generated. It reflects a broader industry shift from vibe coding—quick prompts with minimal oversight—to agentic engineering, which involves structured, verified, and well-managed AI workflows. This evolution is driven by the realization that the model itself is only a small part of the system’s success.

Prior to this, the focus was largely on acquiring powerful models. Now, the emphasis is on how to best integrate, verify, and control these models within the development process, highlighting the importance of configuration, context, and scaffolding.

“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 generate software.”

— Addy Osmani

Amazon

AI prompt engineering tools

As an affiliate, we earn on qualifying purchases.

As an affiliate, we earn on qualifying purchases.

Unclear Aspects of Model-Harness Dynamics

While the whitepaper provides compelling evidence that harness and configuration dominate behavior, it does not specify precise thresholds or best practices for all contexts. The exact proportion of behavior influenced by harness may vary across different applications and models, and the long-term impact of this shift remains to be seen as organizations adopt these insights.

Amazon

software configuration management tools

As an affiliate, we earn on qualifying purchases.

As an affiliate, we earn on qualifying purchases.

Next Steps for AI-Driven Software Strategies

Organizations are likely to prioritize developing robust harnesses, including tools, prompts, and guardrails, to improve AI system performance. Future research and industry efforts may focus on establishing standardized frameworks for context engineering and configuration management. Monitoring how these practices influence cost, security, and reliability will be key as AI becomes more integrated into mainstream software development.

Amazon

AI observability and monitoring software

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 argues that most AI failures and behaviors are determined by how the model is integrated, configured, and managed through prompts, tools, and rules, which together comprise the harness.

What is the main takeaway for development teams?

Focus on building and refining the harness and context management rather than relying solely on the latest AI models for better performance and cost efficiency.

How does this shift affect AI project costs?

While initial setup and configuration may be more intensive, disciplined harnessing reduces ongoing token costs, maintenance, and security risks, leading to lower total cost of ownership.

Will this change how AI models are developed?

Yes, it emphasizes the importance of designing flexible, well-structured integration layers and context management strategies over solely improving model capabilities.

Source: ThorstenMeyerAI.com

Nothing in this article is financial or investment advice. Cryptocurrency and precious-metal investments carry significant risk — do your own research and consider a licensed advisor.
You May Also Like

The Atlas. What the framework is.

An analysis of the Post-Labor Transition Atlas, a new empirical framework assessing AI-driven labor displacement, policy responses, and structural alternatives as of 2026.

Best Quiet CPU Coolers for Sustained AI/Compute Loads

Discover top quiet CPU coolers optimized for long AI and compute workloads, including air and liquid options, for reliable, low-noise performance in 2026.

The labor share. Is value really moving from labor to capital? The data isn’t on anyone’s side yet.

New data shows that while some signals suggest AI may be shifting value from labor to capital at the margins, the overall labor share remains stable, leaving the debate unresolved.

The Local-First Agentic Operator

A single operator, using agentic AI, now builds and manages a broad portfolio of software products, traditionally requiring organizations.