Anthropic recently told its growth team to hire more product managers — not fewer. The reason: Claude Code had quietly turned its engineering org into one that ships at roughly three times its actual headcount. The bottleneck had migrated from the IDE to the people deciding what to build.

This is easy to miss amid the noise of AI productivity claims. But it represents the central structural shift the rest of the industry is now living through. The constraint in software is no longer typing — it's deciding what to type.

A Five-Phase Compression of the Engineer's Day

To understand where we are, it helps to map how fast the workflow has changed:

  1. The Stack Overflow era (2014–late 2022): Engineering knowledge lived in forums. New monthly questions on Stack Overflow dropped roughly 77% after ChatGPT launched in November 2022 — a verdict on the workflow, not the site.
  2. The browser-tab era (late 2022–2024): ChatGPT sat outside the IDE. Engineers pasted answers back into VS Code. Leverage was real but local.
  3. The IDE-native era (2024–2025): Cursor and Claude Code moved models inside the editor with full repo access. The senior-engineer escalation path largely dissolved.
  4. The spec-driven era (2025–2026): Larger context windows compressed multi-sprint work. Amazon's Kiro IDE team reportedly cut feature builds from two weeks to two days. An AWS engineering team completed an 18-month rearchitecture — originally scoped for 30 engineers — with 6 people in 76 days. The bottleneck shifted from writing to specifying.
  5. The routines era (2026): Anthropic shipped Claude Code Routines — scheduled, persistent agents that run overnight or on webhooks. The engineer's job now includes orchestration: spin up a swarm before bed, review a stack of pull requests in the morning.

The Bottleneck Moved. Most Teams Haven't.

Engineering throughput has roughly tripled. Product capacity hasn't moved. The traditional 1:8 PM-to-engineer ratio now plays out closer to 1:20 in effective output terms.

The pattern is consistent across companies running agentic workflows in production: the system produces built features faster than it produces decisions about what should be built.

  • LinkedIn replaced its associate product manager track with a "Product Builder" program training generalists across product, design, and engineering
  • Anthropic is actively expanding its PM headcount
  • Teams that have deployed agents at scale consistently cite decision bandwidth — not engineering bandwidth — as the constraint

For engineers, this is the most important career signal of the decade. And it's the easiest one to miss while productivity stories dominate the feed.

First Principles Matter More, Not Less

The instinct to declare fundamentals obsolete in the agent era gets the trend exactly backwards.

When a memory leak takes down production at 3 a.m. — traced to a subtle ownership bug pushed four years ago — no current agent closes that loop end-to-end. Operating systems, networks, concurrency, and query plans still determine who can resolve a real incident. They also determine who can spot when an agent's output looks correct on the surface and is quietly, expensively wrong underneath.

The agent that wrote 70% of the code in a modern repo cannot reliably tell anyone where its assumptions about thread safety, memory ownership, or transaction isolation diverged from the runtime.

Fundamentals are no longer a hygiene skill — they're a leverage skill. The blast radius of the engineer who understands the stack has grown, not shrunk.

Review Is the New Writing

Engineers in 2026 generate code at a rate that exceeds what any team can read carefully. The 2025 Stack Overflow developer survey found 84% of developers using AI tools — but 46% saying they don't trust the output, up sharply from 31% the previous year.

That gap — heavy use paired with low trust — is exactly where review skills now matter most. Engineers who push volume without rigorous review are accumulating technical debt that will come due in the first real incident.

The New Differentiator: Owning the Product Funnel

Both fundamentals and review discipline are necessary. Neither is sufficient. The engineer who matters in 2026 is one who has stopped waiting for the funnel to arrive as a Jira ticket.

That means taking on work the role historically skipped:

  • Talk to customers. Watch how they actually use the product. Read the support queue. The signal a PM gets through three layers of summary, an engineer can now get firsthand in an afternoon.
  • Generate ideas, not just estimates. The PM who used to source ideas for 8 engineers can't do it at the same fidelity for 20. An engineer who shows up with a validated, scoped opportunity isn't doing the PM's job — they're doing the job the new ratio requires.
  • Work backwards from the customer. Amazon has written the press release first for two decades. The discipline applies equally to solo contributors and agent swarms — both produce working software in the wrong direction without a clear definition of what "customer wins" means.
  • Stop hiding behind bandwidth. The honest answer to "Do you have capacity?" used to be no. With routines and a cooperative agent stack, it's closer to "What is the idea worth?" — a harder conversation that requires a genuine point of view.

What the Next Decade Rewards

The five phases above aren't really a history of tools. They're a history of which parts of the job required a human. The parts that remain human — and will for the foreseeable future — have moved up the funnel: from typing, to reviewing, to deciding.

The engineers who internalize that shift early will define the next generation of technical leadership. The ones who don't will find their tripled throughput pointed in the wrong direction.