Engineer Argues Cloud-Native Stack Needs Fundamental Rethink, Not More Layers

During a recent podcast discussion, Alex Zenla, CTO and co-founder of container isolation company Edera, challenged the software industry's acceptance of flawed infrastructure abstractions, arguing instead for what she calls "spite-driven development"—designing systems to solve genuine technical pain points rather than layering complexity atop broken foundations.

The Problem with Accepting Broken Abstractions#

Zenla stressed that engineers should refuse to work around fundamental design problems. Rather than patching inherent flaws in existing systems, developers ought to redesign architecture from the ground up. She cited Linux as a prime example: while it remains indispensable, the kernel has real limitations that ripple through every layer of software built above it.

The core issue surfaces acutely in security. A kernel vulnerability immediately compromises an entire node's security posture, regardless of how well-designed overlying Kubernetes or container systems might be. According to Zenla, the industry has grown too comfortable stacking abstractions—Kubernetes, containers, orchestration layers, and applications—without addressing root-level weaknesses below.

Multi-Tenancy and the Linux Kernel Bottleneck#

Contemporary cloud environments rely heavily on Linux namespaces and cgroups to isolate workloads, but Zenla argues these mechanisms are insufficient for true multi-tenant security. Kernel memory is shared across all containers on a node, meaning isolation remains fundamentally incomplete. She advocates for pivoting toward stronger virtualization and hardware-level isolation rather than continuing to patch namespace-based solutions.

AI as a Tool, Not a Replacement for Expertise#

Zenla endorses using large language models as "symbiotic assistants" for rapid prototyping and learning—having Claude explain specifications, for example—but warns against blindly trusting AI-generated code. Engineers must retain deep technical knowledge to debug failures and understand what their systems actually do. Vibe-coding (building without understanding) conflicts with spite-driven development, which demands genuine comprehension of one's own software.

Why Hardware Matters#

Zenla also flagged a practical inefficiency: the industry forces consumer-grade GPUs optimized for graphics rendering to double as AI accelerators. Long-term, specialized hardware like TPUs or custom kernel drivers are essential for secure, efficient AI workloads.