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 whitepaper from Google emphasizes that in AI-driven software development, the model itself is only 10% of the system. The focus should shift to harness design and context engineering, which are responsible for 90% of behavior. This challenges traditional views and impacts how organizations invest in AI tools.

A Google whitepaper released in early 2026 states that the model accounts for only 10% of an AI system’s behavior, with the remaining 90% determined by the harness and context engineering. This shifts the focus away from model improvements toward better configuration and contextual design, impacting how organizations approach AI development and investment.

The whitepaper, authored by Addy Osmani, Shubham Saboo, and Sokratis Kartakis, argues that the dominant factor in AI system performance is the harness — the prompts, tools, rules, and observability layers surrounding the model. Evidence from benchmarks shows that changing only the harness can significantly improve outcomes, while model upgrades contribute less to performance gains.

It emphasizes that most failures in AI agents stem from configuration issues such as missing tools or vague rules, rather than the underlying model. The authors recommend that organizations prioritize context engineering — carefully designing instructions, knowledge, and guardrails — as the core skill for effective AI deployment. This approach shifts the economic consideration from model costs to the design and maintenance of the harness, which can be more cost-effective long-term.

At a glance
reportWhen: published early 2026
The developmentGoogle’s new whitepaper highlights that the core of AI system performance lies in harness and context, not the model itself.
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

Implications for AI Development Strategies

This paradigm shift means organizations should invest more in harness and context design rather than solely chasing the latest model advancements. It suggests that durable competitive advantages lie in configuration and setup, not in model selection, which can be rapidly replaced or upgraded. Understanding this can lead to more cost-efficient, reliable, and secure AI systems, especially as AI becomes integral to business operations.

The AI Prompt Playbook: Master AI Prompt Engineering with 140 Ready-to-Use Templates for ChatGPT, Claude, Gemini & Copilot

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.

Background on AI System Design and Evolving Practices

Prior to this, the focus in AI development was heavily on improving models, with frequent upgrades from providers like OpenAI and Google. The whitepaper builds on recent findings that AI use is widespread, with over 85% of developers regularly using AI coding agents. The shift toward recognizing the importance of harness and context reflects a maturation in AI engineering, emphasizing configuration, verification, and judgment over raw model power. This aligns with earlier trends toward more disciplined, structured AI workflows.

“The model is only 10% of what determines behavior; the harness is 90%. Focus on configuration and context engineering for real impact.”

— Addy Osmani

Observability in the AI-Native Era: Leveraging AIOps to build, observe, and operate resilient systems

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.

What Aspects of the Model and Harness Are Still Unclear?

While the whitepaper presents compelling evidence that harness and context are more influential than the model, it remains unclear how these insights will translate across different AI applications and industries. The precise methods for optimizing harness design at scale and the long-term economic impacts are still being studied and debated.

YAML Made Simple: A Beginner’s Guide to Configuration and Data Structuring

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.

Next Steps for Organizations Adopting AI Development Practices

Organizations are likely to reevaluate their AI strategies, investing more in harness development, context engineering, and configuration. Future research and industry practices will focus on creating standardized frameworks for harness design, testing, and maintenance. Monitoring the impact of these shifts on performance, cost, and security will be critical over the coming months.

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

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.

Key Questions

Why is the model only 10% of system behavior?

The whitepaper shows that the surrounding infrastructure — prompts, tools, rules, and observability — determines most of how an AI behaves in practice, making the model itself only a small part of the overall system.

How should organizations change their AI development approach?

They should focus less on upgrading models and more on designing and optimizing the harness and context, including instructions, tools, guardrails, and evaluation processes.

Does this mean model improvements are no longer valuable?

Model improvements remain important, but their relative impact is smaller compared to configuration and context design, which are more controllable and cost-effective.

What are the economic implications of this shift?

Focusing on harness and context can reduce long-term costs by minimizing token waste, improving security, and decreasing maintenance, shifting the economic advantage from model upgrades to configuration excellence.

Will this change how AI tools are marketed?

Yes, vendors may emphasize configurability, tools, and setup over raw model capabilities, aligning with the insight that the surrounding system determines performance much more than the model itself.

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

How Portable Monitors Help Traders Stay Flexible

Portable monitors help you stay flexible by expanding your screen space, allowing…

Best Quiet CPU Coolers for Sustained AI/Compute Loads

Discover top quiet CPU coolers suited for long AI and compute workloads, including air and liquid options, with expert insights for 2026.

The SSD Squeeze: Why Storage Joined the Party

Enterprise and consumer SSD prices surge due to NAND shortages driven by AI demand and fab competition, impacting the entire storage market in 2026.

One Model, a Whole Portfolio: What Ten Days on Fable Mean for a Business Building on Frontier AI

A detailed report on how running multiple systems through Anthropic’s Fable model transformed business operations, highlighting benefits, costs, and risks.