AIThis post was created with the assistance of artificial intelligence (AI).

📊 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.

Buying for a business?Offer from Amazon

Get business pricing on tech for your team

  • Business-only prices and quantity discounts
  • Tax-exempt purchasing
  • Multiple users, one account, clear invoices
As an affiliate, we earn on qualifying purchases.

TL;DR

A recent whitepaper from Google highlights that in AI-assisted software development, the actual AI model is just 10% of the system. The focus should be on harness design and context engineering, which determine system behavior and cost efficiency.

A new whitepaper from Google, authored by Addy Osmani, Shubham Saboo, and Sokratis Kartakis, states that the core shift in software engineering driven by AI is not the development of new models but the emphasis on harness design and context engineering. The paper emphasizes that the AI model itself accounts for only about 10% of system behavior, with the remaining 90% determined by how the model is integrated and guided.

The whitepaper, titled The New SDLC With Vibe Coding, underscores that the dominant factor in AI-assisted development is the ‘harness’—the prompts, tools, rules, and observability layers surrounding the model. Evidence cited includes experiments where tweaking only the harness or context improved performance significantly, despite using the same underlying model. For example, one team moved from outside the Top 30 to Top 5 on a benchmark by changing only the harness.

Furthermore, the paper introduces the concept of ‘agentic engineering,’ where AI systems are built with structured verification, testing, and guardrails, contrasting with ‘vibe coding,’ which involves quick prompts and minimal oversight. It argues that the cost-effectiveness of AI development depends heavily on upfront design of schemas, context loading strategies, and modular skills, rather than solely on accessing the latest model.

At a glance
reportWhen: published March 2026
The developmentThe new Google whitepaper reveals that in AI-driven SDLC, the model is only 10% of the system; the majority of performance depends on harness and context engineering.
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 shift means organizations should prioritize building robust harnesses and context management over constantly chasing the latest model upgrades. The insight challenges the common perception that model improvements alone drive AI performance, highlighting instead that system design, configuration, and verification are the real levers for efficiency, cost savings, and reliability. It also suggests that competitive advantage lies in how well teams engineer their AI workflows, not just in their access to cutting-edge models.

Amazon

AI model harness design tools

As an affiliate, we earn on qualifying purchases.

As an affiliate, we earn on qualifying purchases.

Evolution of AI-Assisted Software Engineering

The whitepaper builds on the rapid adoption of AI coding agents, with 85% of developers using them regularly as of early 2026. It follows a broader trend where AI is integrated into the software lifecycle, shifting focus from model development to system engineering. Previous discussions have centered on model performance benchmarks, but this new perspective emphasizes the importance of system architecture, verification, and cost management, marking a paradigm shift in AI-driven development.

“The model is only 10% of what determines behavior; the harness is the other 90%. This redefines how we should approach AI system design.”

— Addy Osmani

Amazon

AI context engineering software

As an affiliate, we earn on qualifying purchases.

As an affiliate, we earn on qualifying purchases.

Unanswered Questions About Implementation and Impact

It is not yet clear how organizations will operationalize these insights at scale or how quickly they will shift their development practices. The paper provides strong evidence for the importance of harness design but does not specify exact methodologies or tools for widespread adoption. Additionally, the long-term impact on model development and the AI market remains to be seen, as the focus pivots toward system engineering.

Amazon

AI system observability tools

As an affiliate, we earn on qualifying purchases.

As an affiliate, we earn on qualifying purchases.

Next Steps for Developers and Organizations

Organizations are expected to reevaluate their AI workflows, investing more in harness and context engineering. Future research and tool development may focus on simplifying and standardizing these practices. Additionally, industry leaders might prioritize training teams on system configuration, verification, and cost management strategies to capitalize on these insights. Monitoring how this paradigm influences AI product development and operational costs will be crucial in the coming months.

Amazon

AI development testing tools

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 according to the whitepaper?

The whitepaper argues that the model itself provides the core capabilities, but how it is integrated, guided, and verified through harness and context engineering determines actual system behavior and effectiveness.

What is harness design in AI systems?

Harness design includes prompts, tools, rules, guardrails, and observability layers that shape how the model functions within a system, significantly influencing performance and reliability.

How does this insight affect AI development costs?

While initial setup of harness and context strategies may be more costly upfront, it reduces ongoing costs by improving efficiency, reducing token waste, and minimizing maintenance and security issues.

Will this shift change the AI market landscape?

Potentially, as organizations focus less on acquiring the latest models and more on building robust system architectures, the value may shift toward expertise in system engineering and configuration tools.

What should teams prioritize after reading this whitepaper?

Teams should invest in developing strong harnesses, effective context management, and verification processes, emphasizing system design over model chasing.

Source: ThorstenMeyerAI.com

HALLOWEEN

Halloween Picks

As an affiliate, we earn on qualifying purchases.

You May Also Like

Simplify Tech Trends: Develop A Signal Monitor With 500 Lines Of C++

A new lightweight signal monitor filters platform and tooling updates, highlighting ‘Software rendering in 500 lines of C++’ for small software teams.

List Of 200+ WebGPU Kernels In @Huggingface/kernels For Local AI Innovation

Hugging Face’s WebAI team releases @huggingface/kernels with 207 optimized WebGPU kernels and Fleet benchmarking tool, advancing in-browser AI inference.

A 2027 Guide To 15 AI Image Generators For Creative Projects

A 2027 guide compares 15 AI image generators, prompt aids and learning resources by their stated purposes—and explains what buyers should verify.

Pi 1.0

Earendil released Pi 1.0, adding Codemode, extension support and other features, alongside the experimental Pi Durable package.