Anthropic recently told its growth team to hire more product managers, not fewer. The reason, as reported in industry circles, was that Claude Code had quietly expanded his engineering organization to a team roughly three times the size of the actual workforce, and the bottleneck had shifted from the integrated development environment (IDE) to the people deciding what to build.
It’s easy to miss this detail in all the noise AI productivity claim. It is also the structural change that the rest of the industry is now experiencing. The bottleneck in the software is no longer writing. He decides what to write. Engineers who treat it as someone else’s problem fall into a plateau.
For most of the last decade, that decision rested with someone else. Software engineering it was an art you slowly absorbed, then practiced in a long, predictable sequence: Dive deep into the technology, write the code, ask Stack Overflow when you get stuck, ask the lead engineer when Stack Overflow fails, submit a ticket. The product manager owned the funnel. The owner of the construction was an engineer. Both sides treated this division as physics.
Then the funnel collapsed in five steps.
A brief history of how an engineer’s day is compressed
Stack Overflow cycle (2014-end of 2022): The mindset of engineers lived together. But new monthly questions on Stack Overflow are now down about 77% Since November 2022, the launch of ChatGPT was no accident. The fall is not a referendum on the site. This is a referendum on the workflow it represents.
Browser icon cycle (late 2022 – 2024): The first generation of ChatGPT sat outside the IDE. The engineers ran the same loop they always used with a faster oracle: Type a command in the browser, paste the response back into VS Code, repeat. The work was still single-threaded and engineer-driven. Leverage was real, but local.
IDE’s native period (2024-2025): Cursor and Claude Code moved the model inside the editor and gave it access to the full repository. The chief engineer escalation path has largely been resolved. For years, the prevailing wisdom among veteran engineers was that Bash had the longest shelf life of any tool in the stack. By 2026, the first command typed in a fresh terminal for a meaningful share of developers is claude.
Specifically managed period (2025-2026): Larger context windows have transformed single-session work into something that used to require tickets, design documents, and sprints. Amazon’s Kiro IDE team was reported to build compressed features in two weeks to two days using the same specification-based workflow they were shipped with. The AWS engineering team completed the 18-month re-architecture, originally designed for 30 engineers, in 76 days by 6 people. The bottleneck is how long it takes to write code. It started with how clearly the team could describe what it looked like properly.
Cycle of Routines (2026): In April, Anthropic shipped Claude Code Routines: scheduled, persistent agents that run on a cadence, webhook, or overnight while the laptop is closed. Crone is back. The hooks are back. The engineer’s job now is to organize in parts: Spin a bunch before bed, review a bunch of pull requests in the morning. Third-party wrappers like OpenClaw, which was briefly discontinued by Anthropic in April before being partially reinstated, made the same point from the open source side.
The bottleneck moved; most teams don’t
Engineering has nearly tripled. Product management hasn’t budged. The traditional 1:8 ratio of already stressed PMs to engineers is now effectively closer to 1:20 as each engineer ships more per day. For example, LinkedIn replaced the associate product manager path with a "Product builder" program that trains generalists in product, design and engineering. Anthropic is hiring more PMs, not fewer. The pattern is consistent among companies that have actually implemented agent workflows in production: The system produces built features faster than making decisions about what to build.
For engineers, this is the most important career signal of the decade, and the easiest to miss when productivity stories dominate the feed.
First principles are not less, but more important
The instinct to declare fundamentals obsolete in the age of the agent gets the trend completely wrong.
When a memory leak halts production at 3 a.m. and the cause turns out to be a subtle ownership error brought forward 4 years ago, no agent currently in the wild closes the loop. Operating systems, networks, concurrency, and query plans still decide who can handle a real incident. They also decide who sees the moments agent output it looks right on the surface and is silent, expensive, wrong underneath. An agent who writes 70% of the code in a modern repo can’t reliably tell anyone where their assumptions about thread safety, memory ownership, or transaction isolation diverge from the runtime. An engineer who can read the difference and grab the engineer the rest of the team needs in the room, and that engineer is based on fundamentals, not motivational skills.
The bottom line is that fundamentals are now leverage skills rather than hygiene skills. By knowing how TCP retransmission works in 2014, the debug ticket was closed more quickly. In 2026, the same knowledge keeps the entire agent-driven release pipeline from sending regressions to scale. Knowing what was happening below, the blast radius of the engineer went up, not down.
Review is a new post
In 2026, engineers are creating code at a speed that exceeds what anyone can read carefully. A team that ships fast and survives is one whose engineers take reviewing AI-generated code at least once with the same seriousness that they save writing it. 2025 Stack Overflow developer poll 84% of developers have deployed AI tools, and 46% say they don’t trust the approach, up from 31% last year. This gap, heavy usage combined with low confidence, is where review skills are now most important. Coders who push the many and look at the few are racking up debt at the first real event, and the engineer who can pay it back is one who combines his volume with a deep first-principles knowledge of the systems involved.
The new differentiator is the product funnel
Both of these are necessary. Neither is enough. The engineer who matters in 2026 is the one who stops waiting for the funnel to arrive in the form of a Jira ticket.
This means doing things that the role has historically been allowed to skip.
Talk to customers. See how they actually use the product. Read the support queue. Sit on the sales call. An engineer can now get the signal that a product team gets through a trifold summary in an afternoon.
Create ideas, not just guesses. A product manager who used to be a source of ideas for 8 engineers may not source ideas for 20 engineers with the same devotion. An engineer who emerges with a proven, comprehensive opportunity is no longer doing the job of a Prime Minister. The engineer does the work required by the new ratio.
Work backwards from the customer. Amazon is writing a press release for the first time in two decades. Discipline goes well with one-man teams and squads of agents. Both produce a lot of software that goes in the wrong direction without a clear statement of what "the customer wins" means before any code is written.
Stop hiding behind bandwidth. Honest answer "Do you have the ability to implement this idea?" it used to be ‘No’. With routines, hooks, and a stack of cooperative agents, an honest answer is closer "What is the idea worth?" It’s a different conversation and harder to talk about without a real customer perspective.
What the next decade holds
The five-phase history above is not really a history of tools. It is the date on which part of the work the person has to perform. The part that is still human and will remain human for the foreseeable future has progressed through the funnel: From writing, to reviewing, to making decisions, to choosing a client to serve, to a problem to solve.
2026 version a senior engineer not the one who writes the most code. He’s the one who knows what to build, can prove it’s worth building, and has the agent fleet and review discipline to ship it before the system collapses at its own pace.
Engineers who master this will spend the next decade working on some of the most exciting business software ever produced. Engineers waiting for a ticket will spend watching the ticket being written by the agent next to them.
Ishan Gupta is a software engineer at Amazon.
