Anthropic recently asked its growth team to hire more product managers, not fewer. As explained in industry coverage, this was because Cloud Code had quietly transformed its engineering organization into a team that shipped nearly three times the number of its actual employees, and the disruption shifted from the integrated development environment (IDE) to the people deciding what to build.
It’s easy to overlook that detail in the noise of every AI productivity claim. This is also the structural change that the rest of the industry is now going through. The bottleneck in the software is no longer typing. It’s deciding what to type. And engineers who think of it as someone else’s problem are going to stagnate.
For most of the last decade, that decision rested with someone else. Software engineering used to be an art you gradually assimilated, then practiced in a long, predictable sequence: think deeply about the technology, write code, ask StackOverflow when stuck, go to a senior engineer when StackOverflow failed, send a ticket. The funnel was owned by the product manager. The construction was owned by the engineer. Both sides regarded this division as one of physics.
The funnel then collapsed in five stages.
A brief history of how the engineer’s day came to an end
StackOverflow era (from 2014 to late 2022): The way the engineers thought they stayed in one place. But new monthly questions on Stack Overflow are now down about 77% since November 2022, which coincidentally was not the case when ChatGPT launched. There is no referendum on the site this fall. It’s a referendum on the workflow it represents.
Browser-tab era (late 2022 to 2024): The first ChatGPT generation sat outside the IDE. Engineers ran the same loop they always ran, just with a faster Oracle: write a prompt in the browser, paste the answer back into VS Code, repeat. Work was still single-threaded and engineer-driven. The leverage was real but local.
IDE-Native Era (2024 to 2025): The cursor and cloud code moved the model inside the editor and gave it access to the full repository. The senior-engineer progression path was largely disrupted. For years, the prevailing wisdom among experienced engineers was that Bash had the longest shelf life of any tool in the stack. By 2026, for a meaningful share of working developers, the first command typed into a new terminal is cloud.
Niche-driven era (2025 to 2026): Larger context windows turned single-session work into something that previously required tickets, design documents, and sprints. Amazon’s Kiro IDE team reportedly took two weeks to two days to build the compressed feature using the same spec-driven workflow they were shipping. The AWS engineering team described an 18-month rearchitecture, originally intended for 30 engineers, was completed by 6 people in 76 days. The bottleneck was how much time it took to write the code. This began to reveal how clearly the team could describe what the right thing looks like.
Routine Era (2026): In April, Anthropic shipped cloud code routines: scheduled, persistent agents that run on a cadence, on a webhook, or overnight when the laptop is turned off. The crone is back. The hooks are back. The engineer’s job is now part of orchestration: spinning up a bunch before bed, reviewing a pile of pull requests in the morning. Third-party wrappers like OpenClaw, which was briefly suspended by Anthropic in April before a partial reinstatement, said the same from the open-source side.
The obstacle was removed; Most teams don’t have
Engineering has almost tripled. There has been no change in product management. The traditional 1:8 ratio of PMs to engineers, which was already stressful, is now closer to an effective 1:20 as each engineer ships more per day. For example, LinkedIn changed its associate product manager track "product manufacturer" Program that trains generalists in product, design, and engineering. Anthropic is appointing more PMs, not fewer. This pattern is consistent across companies that have actually deployed agentic workflows in production: the system is producing features faster than it is deciding what should be built.
For engineers, it’s the most important career signal of the decade, and it’s easiest to miss when productivity stories dominate the feed.
First principles matter more, not less
The tendency to declare fundamentals obsolete in the agent age goes completely wrong.
When production shuts down at 3 a.m. due to a memory leak, and the cause turns out to be a subtle proprietary bug inserted 4 years ago, no agent currently closes that loop from beginning to end. The operating system, network, concurrency, and query plans still determine who can resolve the actual incident. They also decide who can recognize those moments when an agent’s output looks right on the surface and is quietly, costlyly, wrong. The agent writing 70% of the code in a modern repo can’t reliably tell anyone where his assumptions about thread safety, memory ownership, or transaction isolation differ from those of the runtime. The engineer who can read and catch the difference is the engineer the rest of the team needs, and that engineer is built on fundamentals, not promotion skills.
The result is that fundamentals are now a leverage skill, not a hygiene skill. In 2014, after learning how TCP retransmit works, the debug ticket got closed rapidly. In 2026, the same knowledge prevents entire agent-driven release pipelines from shipping regressions at scale. The engineer who knows what’s going on down there has his blast radius go up, not down.
review is the new writing
Engineers in 2026 produce code at a rate that no one can read them carefully. The team that ships faster and survives is the team whose engineers review AI-generated code with at least the same rigor that they once reserved for writing it. The 2025 Stack Overflow developer survey found 84% of developers were skeptical of AI tools, with 46% saying they did not trust the output, up sharply from 31% a year earlier. That gap, heavy usage with low trust, is exactly where review skills matter most now. Coders who put in too much effort and review less are accumulating a debt that will come during the first real event, and the engineer who can pay it back is one who has combined his volume with deep first-principles knowledge of the systems involved.
The new differentiator is the product funnel
Both of these are necessary. Neither one is enough. The engineer who matters in 2026 is the one who has stopped waiting for the funnel to arrive in the form of a Jira ticket.
This means doing things that have historically been allowed without leaving the role.
Talk to customers. See how they actually use the product. Read the support queue. Sit in on sales calls. What used to send a product team signals through three layers of summary, an engineer can now get through in an afternoon.
Generate ideas, not just guesses. The product manager who used to get ideas for 8 engineers can’t get ideas for 20 with the same fidelity. The engineer who comes in with a valid, scoped opportunity is no longer doing the job of a PM. The engineer is doing the work that is necessary for the new ratio.
Work backwards from the customer. Amazon has been writing press releases head-to-head for two decades. Discipline extends well to one’s team and flock of agents. Both produce a large chunk of misdirected software without any clear explanation "the customer wins" Meaning before writing any code.
Stop hiding behind bandwidth. honest answer to "Do you have the potential for this idea?" ‘No’ Used to Be With Routines, Hooks, and a Cooperative Agent Stack, the Honest Answer Is Closer "What is the value of thought?" It’s a different conversation, and it’s much harder to do without a real perspective on the customer.
What will the next decade reward?
The history of the above five stages is not really the history of instruments. This is the history of what part of the work man was to do. The part that is still human, and will remain human for the foreseeable future, has moved up the funnel: from typing, to reviewing, to decision making, to selecting the customer for service and to problem solving.
The 2026 version of a great engineer is not the one who writes the most code. It’s one that knows what to build, can prove it’s worth building, and has the agent fleet as well as the review discipline to ship it without the system collapsing under its own velocity.
The engineers who grasp this will spend the next decade doing some of the most interesting work ever produced by software. Engineers who wait for tickets will spend it watching as the ticket is written by the agent next to them.
Ishan Gupta is a software engineer at Amazon.
<a href