📊 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 highlights that in AI-driven software development, the model itself accounts for only 10% of system behavior. The real focus should be on harnessing, verification, and context engineering, which dominate system performance and cost.
At a glance
reportWhen: published early 2026
The developmentGoogle’s new whitepaper reveals that the core of effective AI software development is not the model but the surrounding harness and verification processes.
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 the Focus on Harnesses and Verification Matters
This shift in understanding redefines how organizations should invest in AI development. Instead of chasing ever-better models, the emphasis should be on building robust harnesses, verification processes, and effective context management. This approach can lead to better system reliability, lower costs, and faster iteration cycles. For decision-makers, recognizing that the model is only 10% of the equation means reallocating resources toward configuration, testing, and security, which are critical for scalable and secure AI deployment.
Mastering Codex for Parallel AI Agents: Run multiple AI agents at once and verify their work — a non-engineer's guide to supervising Codex (Codex Mastery Series Book 2)
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Background of the AI Development Paradigm Shift
Historically, improvements in AI performance focused on developing larger, more capable models. However, recent developments, including the rise of AI coding agents, have shown that the surrounding infrastructure—prompts, tools, and verification—has a greater impact on system behavior. The whitepaper builds on early 2026 data indicating widespread adoption of AI in coding workflows, with over 80% of developers using AI tools regularly. The authors argue that the industry is moving toward a new software development lifecycle (SDLC) centered on intent expression and trust in configuration rather than raw model improvements. This reflects a broader trend toward modular, configurable AI systems that prioritize verification and context management over model size.“The model is only 10% of what determines behavior; the harness is 90%. The behavior you experience is dominated by scaffolding you can build, own, and improve.”
— Addy Osmani
Agentic Artificial Intelligence: Harnessing AI Agents to Reinvent Business, Work and Life
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Unresolved Questions About Implementation and Costs
It remains unclear how organizations will transition from vibe coding to disciplined agentic engineering at scale, and what the precise cost-benefit trade-offs will be for different industries and use cases. The long-term impact on security and maintenance costs also requires further observation as adoption accelerates.
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 Adoption Strategies
Organizations should focus on developing and testing robust harnesses, verification frameworks, and context management practices. Future research will likely explore standardized tools for context engineering and metrics for measuring configuration effectiveness. Industry adoption of these principles could reshape AI development workflows, emphasizing configuration over model size, with ongoing evaluation of cost savings and security improvements.
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.
Key Questions
Why is the model only 10% of system behavior?
According to the whitepaper, the model’s influence is minimal compared to the surrounding harness, verification, and context management, which shape the system’s performance and reliability.What is meant by ‘harness’ in AI systems?
The harness includes prompts, tools, rules, observability, and configuration settings that control how an AI model behaves in a specific application.How does this shift affect AI development costs?
While vibe coding appears cheaper upfront, it incurs higher ongoing costs due to token usage, maintenance, and security risks. Disciplined engineering requires higher initial investment but reduces long-term expenses.What skills should developers focus on?
Developers should prioritize skills in configuration, verification, context engineering, and designing effective harnesses to optimize AI system performance.Will this change how AI is integrated into products?
Yes, the emphasis on configuration and verification suggests a move toward more modular, controllable AI systems that are easier to maintain, secure, and scale.Source: ThorstenMeyerAI.com
