How I run 15 experiments at a time
and still ship one launch a month, by myself
a launch every month, by myself.
closer to two if you count the work i hand off to L3s and fellow L4s and shepherd through.
most PRs in my org under my VP. most LOCs landed.
and the dominant narrative in 2026 is that this IC archetype is dead.
here’s what i’m going to argue:
the “coding machine is dead” narrative doesn’t survive contact with the actual data
the new coding machine doesn’t type fast. they run a portfolio
taste is the moat: taste in picking what to start, what to drop, what to hand off
your time stops being about typing and becomes about review, deciding what to work on, design docs, cross-org alignment and mentoring
none of this works without the infra to get signal back fast
let’s go.
the narrative doesn’t survive contact with the data
every six months a new piece ships about how the IC is dead and the future is “engineers who think in specs” or “PMs who code with cursor.” these takes have legs because they’re flattering to a lot of people who want them to be true.
but the dashboards inside actual product orgs tell a different story.
the engineers landing the most code are still getting promoted, still getting the staff offers, still getting the call when something’s on fire. that hasn’t changed.
what changed is how they’re shipping it.
the new coding machine runs a portfolio
a year ago, a high-output engineer might run two or three things at once. one main project, a side rock, maybe a refactor in flight.
i run 15 experiments at any given time.
yeah. 15.
every morning, the first thing i do (before standup, before slack, before anything) is open my experiments dashboard and walk the list. for each one i’m asking three questions:
is this trending toward launch? if yes, what data am i still waiting on?
is this dead? if yes, kill it today, free the slot.
is this borderline? if yes, what’s the cheapest variant change that would make it land?
this takes 20 minutes. by the time standup starts, i’ve already made the most important decisions of my day.
bad answer for what i do: “i’m a strong IC who ships a lot of code.”
good answer: “i run a portfolio of small bets in parallel, kill the bad ones on day 2, and i’m aggressive about handing off the next wave to teammates so my own time stays on the highest-leverage thing in the queue.”
the portfolio is the unit. the launches are the output.
Paid subscribers continue below — the rest covers: the taste muscle that makes this work, the two kinds of work i hand off and why, where my time actually goes now, the infra requirement nobody talks about, and the failure mode that kills this operating model.
