CI/CD Is Dead, Agents Need Continuous Compute and Computers — Hugo Santos and Madison Faulkner

2026-05-13 · Source: AI Engineer · Field: Technology & Digital — Software Development & Engineering, Artificial Intelligence & Machine Learning, Cloud Computing & IT Infrastructure · Depth: Advanced, long

Summary

Madison, a partner at NEA, and Hugo Santos, CEO of Namespace, propose that traditional CI/CD pipelines are becoming obsolete due to the rise of agentic software, advocating for a "continuous compute" paradigm. They highlight that while human developers submit few pull requests (PRs) weekly, AI agents generate thousands of short-lived branches and PRs, overwhelming current CI/CD systems and making merges impossible. The proposed solution involves accelerating build, test, and deploy times through hardware/software co-design, with caching becoming an orchestration layer. This new architecture eliminates PRs, starting with an "intent and plan" that agents continuously validate internally. The future vision includes fully automated external validation by specialized agents (e.g., security LLMs) and a "pre-merge queue" to reconcile numerous parallel changes before human approval of the intent and result, not individual code.

Key takeaway

For CTOs and VPs of Engineering grappling with scaling development in an AI-driven landscape, your current CI/CD infrastructure will soon be a bottleneck. You should prioritize investing in continuous compute solutions that support agentic workflows, focusing on accelerating internal validation and exploring agent-driven external validation to manage the exponential increase in code changes. This shift will enable your teams to maintain velocity and control as agentic software adoption grows.

Key insights

Agentic software's high-volume, continuous code generation necessitates a shift from traditional CI/CD to continuous compute.

Principles

• Code generation is cheap, continuous, and requires inner-loop evaluation.
• Time to merge is critical, akin to high-performance database serialization.
• CI principles shift from separate phases to continuous, in-loop validation.

Method

A new architecture replaces PRs with an "intent and plan" that agents continuously validate internally. External validation shifts to specialized agents, with a pre-merge queue for human approval of results.

In practice

• Implement hardware/software co-design for faster build/test/deploy.
• Utilize agentic identity for software and retries at scale.
• Employ a stateful environment for agentic loops to avoid delays.

Topics

• Agentic Software Development
• Continuous Compute
• CI/CD Transformation
• Pre-Merge Queues
• LLM-driven Automation

Best for: CTO, VP of Engineering/Data, Director of AI/ML, MLOps Engineer, AI Engineer, AI Architect

Related on AIssential

• The End of CI/CD Pipelines: The Dawn of Agentic DevOps — Agentic DevOps delegates judgment to AI, automating complex tasks beyond traditional scripting but introducing novel failure modes.
• Claude Code on desktop can now preview your running apps, review your code & handle CI failures, PRs in background — Prompt repetition, specialized hardware, and agentic coding tools are rapidly transforming AI development and software architecture.
• The Agentic AI Gap: Why Your AI Workforce is Still Stuck in Orientation — Agentic orchestration, not just more agents, is crucial for bridging the enterprise AI vision-reality gap.
• not much happened today — AI development is rapidly advancing across agentic systems, inference efficiency, and model architectures, while grappling with practical deployment and ethical challenges.
• Your Agent Passed Every Test. It's Still Going to Break in Production. — AgentOps requires a continuous evaluation loop to manage non-deterministic AI agent behavior in production.
• Day 2 - Agentic Engineering/VIbe Coding an AI App from Zero To Production | 🔴 Live — AI agents can rapidly generate functional prototypes and manage project workflows from high-level requirements.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by AI Engineer.