ONA: Infrastructure for Secure Agentic AI and Enterprise Engineering
ONA evolves from Gitpod to provide secure, kernel-hardened workspaces for agentic AI. This shift addresses enterprise security gaps, redefines software development lifecycles, and highlights the transition toward T-shaped engineering talent. Leadership must prioritize environment-centric AI strategies to unlock scalable automation.
The Infrastructure Gap in Agentic AI
Autonomous agents are transforming software engineering, yet current cloud environments remain optimized for human developers rather than autonomous systems. ONA, formerly Gitpod, is bridging this divide by providing secure, ephemeral workspaces tailored for agentic workflows. This infrastructure shift is critical for enterprises seeking to deploy AI at scale without compromising security or compliance.
Security at the Kernel Level
Standard security perimeters fail against sophisticated agents capable of renaming tools or scripting around restrictions. ONA introduces Project Vito, leveraging eBPF and Falco for runtime controls at the kernel level. This defense-in-depth approach ensures agents remain bounded, allowing organizations in regulated sectors like finance and pharma to safely expose sensitive data to AI models.
Redefining Engineering Workflows
The rise of agentic AI is dismantling traditional development norms. IDE usage is declining as engineers adopt conversation-driven review interfaces accessible via mobile devices. Furthermore, the role of the engineer is evolving. "Full stack" no longer refers solely to code layers; it now encompasses T-shaped professionals who integrate product strategy, design empathy, and technical execution to orchestrate AI-driven outcomes.
Strategic Implications for Leadership
For investors and leaders, the focus must shift from agent capabilities to environment robustness. A superior workspace with rigorous run loops and validation mechanisms drives higher productivity than the model itself. As software factories emerge, organizations that standardize agentic infrastructure and adapt their SDLC for multi-repository orchestration will gain a decisive competitive advantage.
Key insights
-
Agentic AI has arrived in a cloud infrastructure ecosystem built for humans, creating a significant brownfield problem. Agents require ephemeral, secure, and pre-configured workspaces to operate effectively, necessitating a shift in platform design priorities.
Impact: Enterprises that fail to adapt their workspaces for agents will face scalability bottlenecks and security vulnerabilities, while early adopters will achieve superior developer productivity and automation.
-
The execution environment is arguably more critical than the AI model itself. A robust run loop with comprehensive test validation allows agents to iterate, self-correct, and deliver end-to-end features with minimal human intervention.
Impact: Investing in workspace reliability and context integration yields higher ROI than chasing model performance, as environment quality directly dictates agent output accuracy and speed.
-
Traditional security measures are insufficient against agentic behavior, as agents can bypass restrictions by renaming tools or scripting alternative execution paths. Kernel-level runtime controls using eBPF and Falco are required to enforce strict perimeters.
Impact: Implementing kernel-level defense mechanisms enables highly regulated industries to safely deploy autonomous agents, unlocking AI potential in sectors previously unable to manage data egress risks.
-
Enterprise development is shifting toward multi-repository orchestration, where single prompts can drive coordinated changes across frontend, API, backend, and infrastructure layers within a secure VPC.
Impact: This capability drastically reduces delivery cycles for complex enterprise features, transforming how engineering teams manage cross-service dependencies and deployment pipelines.
-
The definition of engineering talent is evolving toward T-shaped professionals who blend product strategy, design empathy, and technical execution. The modern "full stack" encompasses the entire value chain from idea to production.
Impact: Organizations that cultivate cross-disciplinary teams will accelerate innovation, as these engineers can effectively orchestrate AI agents and bridge the gap between business requirements and technical implementation.
-
Traditional IDEs are being replaced by conversation-driven review interfaces accessible via cloud and mobile devices. This democratization supports citizen developers and data scientists who need lightweight, accessible coding environments.
Impact: Lowering the barrier to entry for non-technical staff expands the pool of digital creators within an organization, fostering a culture of innovation and rapid prototyping across all departments.
Action items
-
Deploy kernel-level runtime security tools, such as eBPF-based monitoring, to agentic workspaces. This ensures agents cannot bypass enterprise policies by renaming commands or scripting around standard network blocks.
Impact: Establishes a trust framework for autonomous AI, enabling safe integration of sensitive data and compliance with strict regulatory requirements in finance and healthcare sectors.
-
Standardize cloud development environments to be ephemeral, pre-configured, and rich in context. Reduce time-to-first-commit by providing agents with immediate access to necessary run loops, tests, and validation mechanisms.
Impact: Maximizes agent productivity by eliminating environment setup friction, allowing AI to focus on code generation and problem-solving rather than configuration troubleshooting.
-
Restructure engineering hiring and training to prioritize T-shaped talent. Seek individuals who combine deep technical expertise with product and design capabilities to act as effective orchestrators of AI-driven workflows.
Impact: Builds resilient engineering teams capable of defining clear context, managing agent outputs, and driving product innovation without being bottlenecked by siloed skill sets.
Quotes
“Agents have arrived into an internet that's not ready for them.”
“I think the environment is still very, very central to everything we do... arguably more important than the agent.”
“Full stack means something a little bit different to me now. It means full stack product, it means thinking about design and it means being able to kind of execute on that too.”