Cisco's DJ Sampath: "Every New Feature 100% Written by Codex"

I've been watching AI coding assistants for years, but this week's OpenAI-Cisco announcement made me do a double-take. We're not talking about autocomplete anymore.

DJ Sampath, Cisco's AI security leader, just dropped this bomb: "Every new feature that we're building is 100% written by Codex." Features that used to take quarters? Now shipping in weeks.

That's not hyperbole from a marketing deck. That's a technical executive talking about production systems at one of the world's largest networking companies.

The Agent That Actually Ships Code

Codex isn't your typical GitHub Copilot cousin. This thing operates as a full engineering agent:

• Independently navigates massive codebases
• Implements and tests changes autonomously
• Proposes pull requests for human review
• Runs in isolated cloud sandboxes (1-30 minutes per task)
• Handles everything from bug fixes to new feature development

Cisco is using it across three critical areas: scaling AI-native development, accelerating their AI Defense platform, and automating defect remediation. The last one is pure gold for enterprise teams drowning in technical debt.

This partnership goes deeper than a typical vendor relationship. Cisco became OpenAI's design partner, feeding real-world enterprise constraints back into Codex development. Think security compliance, multi-service dependencies, and those gnarly legacy systems that make enterprise engineering so fun.

Beyond the Marketing Hype

Look, I'm excited but not naive. That "100% written by Codex" claim needs context. Sampath likely means first-pass implementation, not the full engineering lifecycle. Architecture decisions, security reviews, and release management still need humans.

But even with that caveat, the implications are massive:

1. DevSecOps acceleration - AI generates patches and tests at enterprise scale

2. Infrastructure automation - Think router configs, switch management, firewall rules

3. Security workflow optimization - Cisco's AI Defense platform needed this speed boost

The technical architecture is smart too. Each Codex task runs in its own sandbox, so teams can parallelize work without exposing production environments. The AI reads files, runs linters, executes tests, then proposes changes through normal PR workflows.

Jeetu Patel, Cisco's Chief Product Officer, focused on practical value: helping network engineers "write, test, and build code more effectively." No dramatic job replacement rhetoric - just better tools for complex work.

The Enterprise Reality Check

Here's what makes this different from typical AI coding demos: Cisco operates at real enterprise scale. Massive interconnected codebases. Compliance requirements. Security audits. Long-running maintenance tasks that crush developer productivity.

If Codex can handle that complexity reliably, we're looking at a fundamental shift in software development economics. Not just for Cisco - for every company managing large-scale infrastructure.

The timing matters too. Enterprise AI security is exploding, and Cisco's positioning themselves as the infrastructure backbone. Faster development cycles become a competitive advantage when the market moves this quickly.

My Bet

This partnership signals the death of AI coding assistants as we know them. The future belongs to autonomous engineering agents that complete tasks, not just suggest code. Cisco's production deployment proves enterprise readiness, and OpenAI gets validation for their agentic approach.

Within 18 months, every major enterprise will have AI agents handling routine development work. The companies that adapt fastest - like Cisco appears to be doing - will ship features while competitors are still debating AI strategy.