How Plastic Labs hired their platform engineer in 5 weeks with Superposition

Plastic Labs is building Honcho, a memory and identity layer for AI agents. The product helps agents track people, tasks and entities over time so they can act with context instead of treating every interaction as a blank slate. That work sits close to the metal of how modern AI products will be built.

Vince is a data-focused founder. Before Plastic Labs he spent years building data products and infrastructure at Penn State and Passage Labs, working across research, data science and engineering. For Honcho to be more than a research project, he needed a platform engineer who could treat reliability and scale as first-class problems, not afterthoughts.

The plan at first was simple: find someone through their own network or through high-signal online communities. Vince and his co-founders met on Twitter, so they trusted that world more than generic job boards. They did not expect LinkedIn or Indeed to produce the kind of person who would be excited about identity infrastructure for AI agents, and qualified.

They posted the role on Wellfound mostly to get a basic ATS. What they got instead was a firehose of irrelevant inbound. They realised they needed someone who had run serious infrastructure at scale and who also fit a very particular culture of self-taught, high-agency work.

The candidates were few and far between and there was no realistic way Vince could sift the internet himself while trying to build Honcho.

Vince kicked off the search with Superposition by getting on a call with the voice agent and talking through the role in his own language. Plastic Labs uses deliberate, sometimes unconventional terms to describe what they look for, including a strong preference for autodidacts.

The agent pushed back on that language in a useful way. When Vince said "autodidact," it didn't just accept the word. It asked what that meant in practice for this role. Did he want people who write up their learning in public, who replicate papers, who ship side projects that show how they think? That process helped Vince to turn taste into criteria the agent could actually search for.

"The agent did a great job of drilling down into what we really mean by those terms for this specific job."

From that call, Superposition built a search pattern that combined the obvious signals with the weird ones. It went beyond LinkedIn. The agent crawled the wider web, connected profiles, and looked for signs of the traits Vince cared about, not just titles and employers. For the candidate Plastic Labs ultimately hired, Superposition did exactly that: it linked a conventional profile to a personal site where he was documenting learning a new language and writing about music. None of that context lived on his resume.

That cross-linking mattered. Those side projects and essays were exactly the kind of breadcrumbs that Plastic Labs uses as evidence of autodidactism. Vince is clear that he might have landed on that person's LinkedIn eventually, but he would not have searched in a way that surfaced the personal site as a first-class signal.

"Being able to synthesize across all those disparate data sources was ultimately how we built conviction around this person fitting our desired profile."

Under the hood, Superposition did this at serious scale. Over the five-week search it scanned 43,802 candidates and surfaced only 15 to Vince. Those 15 were not a random sample. Each came with a clear explanation of why they matched the traits Plastic Labs cared about, and had already been reached out to and replied with interest.

Vince runs a tight hiring loop, so once those candidates started coming in, the process moved quickly. They had structured interviews, saw the match between the person and the "autodidact infrastructure hacker" profile they had defined with the agent, and made an offer.

"The results have been fantastic. We certainly wouldn't have found this person on our own."

For Vince and Plastic Labs the biggest change is not one hire. It is how he now thinks about the act of hiring. Instead of blocking off days to trawl LinkedIn, tweak job posts and send cold messages, he leans Superposition for leverage. He invests upfront in describing the role clearly, then lets the agent run.

The role at Plastic Labs is not standing still. As Honcho grows, they will need more people who can work at the intersection of infrastructure, data and applied research. Vince expects those searches to look more like this one than like the old way. The agent already understands the company's taste, the bar for technical depth, and the kinds of public work that matter to them.

Any time a new gap opens up, the assumption is simple: start with Superposition, treat it as the default hiring surface, and let it bring back the kind of candidates a small, focused team would never have time to uncover on its own.

"When other founders tell me they're hiring I tell them it's just a no brainer. You should use Superposition."