The App That Changed How Engineers Ship Code

A $22M Raise and a Very Specific Product Thesis

The episode opens with Conductor fresh off a $22 million Series A co-led by Spark and Matrix. Charlie and Jackson explain the product simply: it’s a Mac app for running a bunch of coding agents at the same time, each in an isolated copy of your repo, so you can review and merge the results instead of doing all the typing yourself.

Why Cloud Matters More Than Just Convenience

Their launch announcement is Conductor Cloud, which fixes a very practical pain point: until now, everything ran locally, so if you shut your laptop, the agents stopped too. Charlie says the real limit today isn’t compute so much as human cognition — most people can only keep 3 to 5 agents in their head at once — so the next unlock is partly infrastructure, partly interface design.

From Manual Repo Clones to a Real Product

Before Conductor existed, they were doing this the ugly way: multiple clones of the same repo, one Claude instance per clone, then eventually discovering worktrees and realizing they should probably stop suffering. Jackson says the app was born from that friction, while Charlie frames it as a timing story too: early on, the models just weren’t good enough to move beyond the IDE, but eventually they hit a sweet spot where orchestration became useful instead of chaotic.

College Teammates, Startup Co-Founders

The origin story is charmingly specific: they met in college, where Charlie was a fifth-year and Jackson was a freshman — “the biggest age gap possible,” as they joke. They reconnected later while Jackson was on Netflix’s machine learning infrastructure team and Charlie was at Replicate doing growth and engineering, and their shared love of building projects together made a startup feel natural.

YC: A Graveyard of Ideas Before the Right One

Their YC application was for an AI that books reservations, including tennis courts, using browser control with Sonnet 3.5 — a classic “solution in search of a problem,” they admit. During YC they tried idea after idea, shipping prototypes every few days, and Charlie says the turning point came when Aaron told them two things: devtools would suit them, and they should literally put a poster of themselves on the wall saying, “Make something these guys want.”

The False Start Before Conductor: Build the Thing After the IDE

That advice pushed them toward devtools and toward a more ambitious belief: coding tools weren’t going far enough. They first tried building a post-IDE interface where humans just tell AI what to do and review the output, but Sonnet 3.5 still required too much handholding, so they backed off and shipped Chorus, a multi-model chat app, while building internal tooling that eventually led them to Conductor.

The Moment It Felt Like Magic

Conductor went from first line of code to a testable product in about two to three weeks. Jackson says the “whoa” moment was assigning one task, hitting Command-N, assigning another, then watching the unread dot appear as the first agent finished; Charlie’s personal litmus test was whether Jackson, who usually rejected his shiny demos, actually found this one useful.

What They’ve Learned From Top Engineers — and Where This Goes Next

Charlie’s viral-demo instincts came partly from Replicate, where attention was part of the job; his advice is to be direct, skip corporate speak, and make the kind of post you’d actually want to like yourself. On product usage, they say the best engineers don’t have crazy setups — they have well-crafted markdown “skills files,” strong judgment about constraints, and “slop-free zones” where AI can roam versus parts of the codebase where humans stay firmly in charge. Looking ahead, they expect models to get 10 to 100 times smarter, agents to run much longer, and interfaces to evolve from juggling a few bots to managing an entire organization of AI workers, with code review itself still stuck in what Charlie calls the “2010 GitHub PR review era.”