At 10x Science, the first engineer hired was brought in to find the product, not to scale code

Y Combinator published, this past Friday (10/30/2026), Andrew HunzekerHesed's account, a former Principal Engineer at Grindr, of how he became Employee #1 at 10x Science, a startup from the accelerator's W26 batch. The text is, admittedly, a recruitment piece. But it matters to anyone deciding on early-stage hiring precisely for what it reveals about the bet behind the role: giving a single engineer a mandate to define the product, not just to execute it.

Before Grindr, Andrew built software seen by millions of people: food delivery logistics, CPG restocking at Walmart, video streaming at Major League Baseball. According to him, the development cycle ended up looking the same everywhere big, even with AI speeding up the pipeline. It was that boredom, plus a conversation with 10x Science's founders at a talent summit, that led him to the startup.

Who is 10x Science

10x Science went through Y Combinator's W26 batch, according to the accelerator blog post's own categorization. The product is AI-native software for scientific data analysis, currently focused on proteomics and drug discovery, with a stated plan to expand into other life sciences areas.

The source doesn't give valuation, funding, or revenue, neither in the text nor in the post's metadata. The only stage-related data is qualitative: the company is, in the text's own words, making the transition from exploratory research to locking in features for paying customers. That suggests that, as of publication, 10x Science wasn't operating with consolidated recurring revenue.

On team size, the only thing that can be stated with confidence is what the job title itself already gives away: before Andrew, the operation was the two founders, David and Vishnu (last names not disclosed in the source). There's no public customer count, nor the name of the pharmaceutical lab mentioned in the account: only the description that it's a pharmaceutical company among the industry's top ten.

How engineering works day to day

The hiring process wasn't a battery of algorithm questions. It was a multi-day work trial with full mutual access: Andrew had access to the founders, and the founders had access to him, working together on a real problem before any offer.

The real test came in the first week. The chance came up to present, to a pharmaceutical company among the world's top ten, a solution to a problem that 10x Science's own team didn't yet know existed. They had three days to bring Andrew up to speed on the domain, design the solution, build it, and package it for a Friday demo. According to the account, that tool became a central piece of the product and has already found applications beyond the original presentation.

The stated toolset is deliberately simple: command line, VS Code, and whichever AI model is "least terrible" that week. Andrew deliberately rotates between providers, avoiding getting locked into a workflow imposed by a single vendor.

I don't want to get sucked into how some person at OpenAI or Anthropic thinks I should be working.

Andrew HunzekerHesed, Employee #1 at 10x Science

There's also a layer uncommon in software engineering: Andrew's work has already appeared on posters presented at scientific symposiums, and he has submitted work to the American Chemical Society. At this company, code also becomes signed scientific output, not just a commit in a repository.

What sets it apart from a conventional science company

The contrast Andrew himself draws, about his previous experience at large companies, sums up the point:

Exec to Design to PM to Engineer, with ever-increasing velocity due to AI.

Andrew HunzekerHesed, Employee #1 at 10x Science

At a conventional scientific software company, or at a big tech firm with a mature process, that funnel exists to reduce risk: the idea goes through product validation before becoming code, and the engineer rarely decides what to build. At 10x Science, the customer's problem landed directly on the desk of an engineer who had been hired three days earlier, with no PM layer between the pitch and the delivery.

The second contrast is about what counts as output. At a traditional bioinformatics or life-sciences software vendor, the engineer delivers features; it's the science team that signs papers and posters. At 10x Science, the two things blend together: the same professional who writes code appears as a co-author of scientific output submitted to entities like the American Chemical Society. That only works because the company treats the engineer as part of the discovery team, not as a technical service provider for scientists.

What to copy if you're building in Brazil

A multi-day work trial with mutual access replaces the algorithm interview when the role demands domain judgment, not just the ability to code. The trial tests whether the person can handle real ambiguity, something no CLRS question measures.

Throwing the newcomer into a real customer problem in the first week works as a filter and as onboarding at the same time. Early-stage Brazilian companies tend to shield new hires during the first few weeks; the 10x Science case suggests that this delays the most important signal, which is seeing the person produce under pressure with incomplete context.

Hiring for domain curiosity instead of AI hype is the criterion Andrew says he felt in the conversation with the founders: people talking about mathematical principles and the physics of sound, not about a commercial model wrapper. For Brazilian technical founders competing for talent with startups that only know how to sell "AI," this is a cheap-to-replicate recruiting edge.

Treating the AI model as a replaceable part, not a fixed stack keeps the company's engineering from becoming hostage to OpenAI's or Anthropic's product roadmap. This matters even more for those building in Brazil, where API cost and availability shift with exchange rates and each vendor's commercial policy.

Giving the first engineer a mandate to "find the thing," in Andrew's own words, is a bet that early stage isn't about technically scaling, but about discovering what's worth scaling. Founders who hire their first technical engineer just to "get work off my desk" are buying the wrong person for that stage.

Risks and what 10x Science doesn't say

The account carries an evident concentration risk: a tool built in three days by a single newly arrived person became, according to the text, a central piece of the product. There's no mention of documentation, review pairing, or knowledge redundancy, which is normal for a company of three or four people, but turns into a technical liability if 10x Science grows without fixing it.

Switching AI models every week, for the convenience of using "the least terrible" one, is a pragmatic practice at an exploratory stage. But proteomics and drug discovery usually require reproducibility and an audit trail once the product becomes a tool for regulatory or pharmaceutical decisions. The source doesn't say how the company plans to reconcile that model flexibility with the rigor that paying life-sciences customers normally demand.

The total absence of financial figures is the biggest hole in the account. Without disclosed funding, revenue, or customer count, it's impossible to assess whether the hiring pace and the willingness to bet everything on a single engineer are backed by real cash or by founder optimism from a recent Y Combinator batch.

It's worth noting that Andrew himself is the most honest point in the text regarding return expectations:

Will you get rich? Probably not. Will you gain more experience in six months than you can in years at a big company? Yes.

Andrew HunzekerHesed, Employee #1 at 10x Science

That's a necessary counterpoint to the romanticizing of the "founding engineer": the narrative of agency and scientific discovery is real in the account, but the protagonist himself dismisses the idea of financial return as a reason to join. Anyone reading this case as a hiring model needs to separate the two things: the product thesis (giving a discovery mandate to those who arrive early) can be copied; the expectation that this translates into valuable equity has no data in this source to support it.