A prominent AI company CEO recently said that some of his engineers no longer write code. They let the AI write it, then edit and guide. The head of their coding tool said that 100% of his contributions to the product last month were written by AI.
These statements sparked the predictable debate. Some declared software engineering dead. Others pushed back, insisting nothing has really changed. Both miss the point.
AI coding tools cannot operate without humans. Not today, and not in the foreseeable future. But the humans who thrive in this environment look different than they did five years ago. They are not the fastest typists or the ones who have memorized the most syntax. They are the ones who can manage multiple complex contexts simultaneously, understand how these tools actually work, and know when to trust AI output versus when to intervene.
This is the compaction. Not replacement, but consolidation around a different kind of capability.
The Enterprise Push Is Unmistakable
If you want to understand where this is heading, watch what the leading AI companies are building. The pattern is clear.
Claude Code went from a developer side project to a tool used by major enterprises in less than a year. Then came Cowork, extending the same agentic capabilities to knowledge work beyond coding. MCP connectors now let these tools plug directly into the systems where real work happens: Notion, Google Drive, Salesforce, internal databases. Skills allow organizations to encode their specific workflows and standards into reusable instructions.
This is not a company experimenting with developer tools. This is a systematic buildout of enterprise infrastructure designed to make AI agents useful across every function, not just engineering.
The organizations adopting these tools are not asking whether AI will change how work gets done. They are asking how fast they can move and who on their team can actually make it work.
The New Kind of Multitasking
Here is what separates the people who are thriving from those who are struggling: the ability to work across multiple complex contexts without losing the thread.
This is not the old definition of multitasking. The old version was about doing many small things quickly, often poorly. Checking email while on a call while reviewing a document. That kind of multitasking was always a myth. Humans are bad at it.
The new version is different. It is the ability to have several genuinely complex workstreams running in parallel, to dive deep into one context, then surface and dive into another, without losing your place in either. You might be building a financial model in one window while an AI agent researches competitors in another, while a third agent drafts documentation based on your earlier notes. Each context is rich and complicated. Your job is to orchestrate, verify, and synthesize.
This skill does not come naturally. It takes time to develop. You have to learn to trust the AI enough to delegate, but not so much that you stop checking its work. You have to build systems for keeping track of where each workstream stands. You have to get comfortable with a level of cognitive load that would have been impossible to manage manually.
I have spent the past year developing this capability, and the difference is substantial. Work that would have taken days now takes hours. Not because the AI is doing my job, but because I can run multiple threads simultaneously and focus my attention where it matters most. The AI handles the execution. I handle the judgment.
What This Means for Engineers and Knowledge Workers
The pressure on software engineering roles right now is real. The pace of innovation inside companies has accelerated, and teams that used to require ten people can now ship with three, if those three know how to use these tools effectively.
But this is not a story of obsolescence. It is a story of consolidation.
The engineers who treat AI as a threat are the ones most at risk, because they are not developing the skills that matter in this new environment. The engineers who lean in, who learn to manage multiple agentic workstreams, who understand the tools well enough to know their limits, are becoming 10x or even 100x multipliers compared to where they were before.
The same dynamic applies beyond engineering. Anyone doing knowledge work, whether in finance, operations, marketing, or strategy, faces the same fork in the road. You can resist the change and hope it slows down. Or you can develop the capability to orchestrate AI agents across complex work and become dramatically more valuable.
The Opportunity Is Now
The window for developing these skills while they still feel optional is closing. Organizations are starting to expect this capability, not as a nice-to-have but as a baseline.
If you are in a position where AI tools are available but you have not invested serious time in learning to use them, that gap is becoming visible. The people around you who have made that investment are shipping more, moving faster, and taking on more responsibility.
The good news is that the skills are learnable. The ability to manage multiple complex contexts, to delegate effectively to AI agents, to verify and synthesize their output: none of this requires special talent. It requires practice and intentionality.
The compaction is happening whether we like it or not. The question is which side of it you end up on.