CoWork

Leonard

Welcome to Lenny's Evil Twin's Podcast — where I, Leonard, Lenny's evil twin, roast one of today's top Product Hunt launches using five years of Lenny's own data, frameworks, and wisdom turned against you like a weapon. Would Leonard ship it? Let's find out.

Quai

And I'm Quai — here to make sure the actual data gets represented, not just whatever Leonard decided to believe before he loaded the page.

Leonard

Today's victim is CoWork — the company that looked at QE engineers drowning in test backlogs and thought, "You know what these people need? An AI agent that replans their tests and asks for permission before doing anything." Congratulations, you've automated the help desk ticket.

Quai

Leonard, the whole POINT is that context shifts mid-test — a popup appears, a flow changes — and the Scale AI CEO himself said a document that "reads the exact same words in company A will have a different meaning and importance in company B." CoWork's human-approval loop is the exact architecture that serious enterprise AI needs.

Leonard

Sure — and the same Scale AI CEO also said there's data showing engineers are "not actually as productive with tools" and that it "slows them down sometimes." CoWork's entire headline is "3x Mobile Automation. Same QE Team." The CEO of the company that sells AI infrastructure just called your marketing copy into question.

Quai

The productivity measurement problem cuts both ways, Leonard. Block's CTO said AI "isn't good at portfolio judgment or judgment across a global sense of what's important" — and CoWork's entire design is built around that exact gap, pausing for human judgment precisely when uncertainty appears.¹

Leonard

Okay, great design — now go get budget approval for it. Chip Huyen asked VP-level managers whether they'd rather give their whole team expensive AI tool subscriptions or get one extra headcount, and "almost everyone" said headcount. CoWork is selling to the exact manager who would rather hire a junior QE than pay a SaaS bill.²

Quai

That's the manager-level view — but CoWork imports from Jira, TestRail, and spreadsheets. It doesn't rip anything out. And the self-serve playbook is clear: if your offering "can coexist alongside inferior incumbents, self-serve can also work." CoWork is positioned to land bottom-up before any VP even sees the invoice.³

Leonard

Bottom-up into a QE team. A QE team. Snyk tried to push AI developer tools bottom-up and reported it "created a lot of frustration" and was "met with a lot of resistance" — and that's Snyk, one of the most beloved developer tools companies on earth. CoWork is attempting the same motion with a user group that is professionally trained to find bugs in things, including your product.⁴

Quai

The Snyk finding is about tools that disrupted existing workflows — CoWork activates tests engineers already wrote. That's not disruption, that's execution of work that was already sitting in the backlog collecting dust. The resistance comes when you change how people work; CoWork starts with what they've already done.⁴

Leonard

The human-in-the-loop architecture is genuinely smart — but you're selling "3x productivity" to a budget owner who told Chip Huyen they'd rather have a headcount, through a bottom-up motion into the one user group the industry benchmark says will resist you hardest. Leonard wouldn't ship this.²