Managers as ICs

Sources#

• Running an AI-native engineering org

Summary#

Fiona Fung's "spicy" org change on Claude Code: every manager starts as an IC first and stays directly responsible for a part of the product. The org is kept "agile and as flat as possible." What made this newly viable is that agentic coding collapsed the onboarding cost that used to make manager-into-codebase re-entry daunting — so managers can reclaim maker hours without a painful ramp.

The change#

When Fung joined Claude Code, one of the first things she did was work with recruiting so that every manager begins as an IC before taking on people-support responsibility. Rationale: it lets a manager "roll up your sleeves, get into the codebase, and learn what it's like to be an engineer on the team" before owning the responsibility of supporting people. She frames flat-org + managers-with-IC-ownership as one of the three things she prioritized that "worked really well."

Why now: onboarding stopped being daunting#

Fung has switched between manager and IC roles across her career (Meta, Microsoft, Anthropic) and always felt "a slight hesitation" returning to IC work — "it could be daunting if you haven't done it in a while." Agentic coding removed that friction: onboarding to Claude was "such a different experience," because she could ask Claude to teach her the surface area instead of taxing engineers (Code as Source of Truth). The cost that historically pushed managers out of the codebase — ramp time and the social cost of asking for help — fell sharply.

Dogfooding rationale#

The deeper reason is dogfooding (see Dogfooding as Product Discipline): "on Claude Code we use Claude to build Claude, and Claude Code to build Claude Code" — so a manager who isn't in the codebase isn't experiencing the product they're shipping. Managers-as-ICs is how leadership keeps "feeling the product in their bones." It's the structural enforcement of the dogfooding norm at the management layer.

Connection to engineer–PM convergence#

This is the management-layer instance of Engineer PM Convergence: Boris Cherny notes every member of the Claude Code team writes code (EM, PM, designers, data scientist, finance, user researcher). Managers-as-ICs extends "everyone codes" up the org chart and is consistent with the lighter, flatter AI Native Product Cadence Anthropic runs.

Connections#

• Role Averaging, Not Role Elimination — the mirror-image both/and: Anthropic keeps managers coding; Ambrosino turns ICs into managers-of-agents; managers persist in both
• Fiona Fung — instituted the change
• Code as Source of Truth — Claude-as-onboarding-doc is what makes manager IC re-entry cheap
• Dogfooding as Product Discipline — the rationale: managers must use/build the product to keep product sense
• Engineer PM Convergence — managers-as-ICs is "everyone codes" extended up the org chart
• AI Native Product Cadence — the flatter, faster org this org-shape supports
• Boris Cherny — the IC archetype (logs 150-PR days) the management norm points toward
• Harness Shrinkage as Models Improve — falling onboarding/coding cost is what makes a flatter org affordable
• Compounding Loop Optimization — "build your own tooling" presumes everyone can code; the flat, IC-heavy org is the substrate that makes afternoon-built internal tools the norm
• Parallel Agent Orchestration — running a fleet of concurrent agents is the IC-becomes-manager shift in literal form: the intensive Codex user manages, delegates to, and reviews a team of agent-workers

Open Questions#

• Fung's own open question: "Do you still need separate iOS and Android orgs?" — if engineers flex across platforms via Claude, the traditional platform-split org may dissolve too. How far does flattening go?
• Does manager-as-IC scale past a certain org size, or only work while Claude Code is small and the codebase is Claude-legible?

Sources#

• Running an AI-native engineering org