Why the discipline exists now
Software engineering had this moment already. One developer with a coding agent and a harness ships what a team shipped two years ago — not because typing got faster, but because the machine took the plumbing and left the human the judgment.
Growth is having the same moment, later and messier. The work is still split across a tool per channel, a freelancer per craft, and a founder doing the routing by hand. The tools describe; a human does; nothing learns. Distribution engineering is what it looks like when that splits gets rebuilt the way code was: agents own the execution, memory compounds across channels, and the human moves to the one place they can't be replaced — deciding what ships.
What a distribution engineer does
Not a growth hacker, not a channel specialist, not an agency manager. A distribution engineer designs and runs the system that distributes a product:
- Picks the channels deliberately and treats each one as an asset with a measured yield, not a place to post
- Runs the disciplines through agents — the sweep, the draft, the page, the variant batch — instead of doing each by hand
- Holds the approval gate: everything that goes out is reviewed, priced and signed off by a human
- Feeds the memory: every result, reply and ranking lands back in one place, so the next run starts smarter
- Measures per run, not per month — what did this send, this page, this batch actually earn
What it is not
It is not automation. A workflow moves data on a schedule; it cannot weigh whether the thread is worth replying to. Distribution engineering keeps judgment in the loop twice — inside the run, where the agent reads before it writes, and at the gate, where the human decides.
It is not growth hacking. Hacks are withdrawals from channels; engineering is deposits. The practice is boring on purpose: pick, run, measure, keep what earns, repeat weekly.
And it is not a rebrand of marketing ops. Ops keeps the tools running; engineering replaces the tools with a system that does the work.
Where the term comes from
The pattern borrows its name from what happened next door. When enrichment and automation tools got programmable, the people who built machines instead of buying seats became GTM engineers — a real title now, on real job posts. Distribution engineering is the next iteration: the machine-building extends past the pipeline into every discipline a product needs to be found, and the plumbing moves to agents.
The moat logic drives it. When any product can be rebuilt in a weekend, distribution is the asset that compounds — and an asset that compounds deserves engineering, not improvisation. The full argument lives in /blog/distribution-is-the-moat.
The harness
A practice needs its tool. Software engineering got harnesses — the loop that carries the agent, the memory, the review gate. Moatt is that harness for distribution: one agent, every discipline, one shared memory, one metered key, and your approval on everything that ships. Code has Claude Code. Distribution has Moatt.