AI Makes Code Abundant—Now “Absorption Capacity” Is the Real Constraint for CTOs
AI is making code cheaper and faster to produce, but organizations are hitting a new constraint: their capacity to absorb, validate, secure, and ship the resulting change.
AI’s impact on engineering is moving from experimentation to operating-model change. In the last 48 hours, the signal isn’t just “AI can write code”—it’s that AI is changing what limits delivery, how companies staff teams, and which metrics are safe to use.
Two stories frame the shift. Snap is cutting 1,000 jobs and explicitly citing AI advancements as a driver, a blunt indicator that some firms now view automation as a structural productivity lever rather than a feature add-on (TechCrunch). Meanwhile, Zendesk argues that GenAI makes code “abundant” and moves the bottleneck to “absorption capacity”: the organization’s ability to define problems clearly, integrate changes into complex systems, and safely operationalize what’s produced (InfoQ).
This is why the emerging fascination with token-based metrics is risky. Reid Hoffman’s “tokenmaxxing” commentary acknowledges that token usage can indicate adoption, but warns it’s not a productivity metric without context (TechCrunch). For CTOs, that’s a governance warning: if you measure the easiest thing (tokens, PR counts, lines changed), you’ll optimize the wrong behavior—more generated output—while your true constraint is review bandwidth, test reliability, release safety, and system coherence.
The tooling direction reinforces the same theme: assistants are becoming cross-tool operators, not just copilots. Adobe’s Firefly assistant can act across Creative Cloud apps to complete tasks end-to-end (TechCrunch). Translate that pattern to engineering and you get agentic workflows that can propose changes across repos, configs, dashboards, and runbooks. That amplifies the need for strong change-control primitives (policy-as-code, environment protections, auditable pipelines) because “what changed” will matter more than “who typed it.”
What CTOs should do now is treat absorption capacity as an explicit platform capability. Invest in the systems that turn AI output into safe production change: higher-fidelity automated testing, dependency and vulnerability prioritization, and clear ownership boundaries for review and release. InfoQ’s session on open-source dependency risk management is a reminder that as change volume rises, supply-chain risk management must become more automated and more targeted (InfoQ). The goal is not to slow down AI-generated change, but to raise the organization’s throughput for validation, integration, and rollback.
Actionable takeaways: (1) Rebalance metrics from “output” to “integration throughput” (lead time to safe deploy, change failure rate, rollback time) and avoid token-count proxies without context. (2) Build a thin layer of governance for agentic tooling—permissions, audit trails, and policy checks—before assistants can touch production-facing assets. (3) Increase absorption capacity deliberately: invest in tests, dependency controls, and clearer problem definition so “abundant code” becomes reliable shipped value, not an unreviewable backlog.
Sources
- https://www.infoq.com/news/2026/04/zendesk-absorption-capacity/
- https://techcrunch.com/2026/04/15/reid-hoffman-weighs-in-on-the-tokenmaxxing-debate/
- https://techcrunch.com/2026/04/15/snap-is-cutting-1000-jobs-16-of-its-workforce/
- https://techcrunch.com/2026/04/15/adobes-new-firefly-ai-assistant-can-use-creative-cloud-apps-to-complete-tasks/
- https://www.infoq.com/presentations/open-source-dependencies/