We turned a Slackbot into an $11M Series A in synthetic biology

The hardest year was 2024. We had two products effectively — a free-feeling academic tier and a paid commercial tier — and they were starting to diverge in ways that hurt both. Academic users were our best evangelists. They'd take the product to their next industry job. But the academic version was eating about 35% of engineering hours and bringing in maybe 8% of revenue.

In August 2024 our biggest commercial customer threatened to churn because we'd shipped a feature their compliance team didn't like — a Slack notification that included sample IDs by default. They were doing GLP-adjacent work and any leak of sample identifiers was a real problem. They gave us thirty days.

§The advisor sentence that ended an argument

Joon and I had a hard fight about whether to kill the academic tier. He wanted to. I felt like we'd be killing the thing that had made us. We didn't speak for four days. Our advisor — a former director of operations at Ginkgo Bioworks — finally told us, on a video call, "You're not killing the academic tier. You're moving it to a different binary, with different defaults, that doesn't get features the commercial customers don't ask for. Stop arguing about your feelings."

It was one of the most useful sentences anyone has ever said to me. I've since recommended the same approach to several other technical-founder cofounder pairs, including the team at Drift Studio when they spun their SaaS out of an agency.

The Series A happened almost by accident. In late 2024 a partner at the deep-tech firm joined a customer's Slack to look at how we'd built the equipment scheduler. He spent two hours in the channel reading old messages. At the end of those two hours he emailed me: "I want to lead your A. Can we meet tomorrow."

The term sheet came eleven days later. $11M at a $54M post. We took it because the partner understood what most VCs hadn't: the moat wasn't the scheduling product, it was the run-log graph. Every experiment our customers logged got linked to the equipment, the reagents, the protocol, and the result. Three years of that data, even from a few hundred labs, was worth dramatically more than the SaaS itself.

As of April 2026 we have 220 commercial customers, $4.7M ARR, and a team of 22. We use Linear for issues, Notion for the wiki, GitHub for code, Vercel for the front end, and AWS RDS for the run-log database. The academic tier is still free, still has a separate codebase with stricter defaults, and is still where roughly 40% of our pipeline originates. The original Slackbot is still in production and still answers to the slash command we wrote in 2022. We've never rewritten it.

§What the Series A actually bought us

The four months after the Series A closed were when the math compounded. We replaced the Slackbot front end with a proper Next.js dashboard, but kept the bot itself running as a parallel ingest path because forty-three of our fifty wet-lab customers still preferred typing the slash command "/onyx run" over clicking buttons. Average plate-throughput per customer climbed from eleven plates per week to twenty-eight after we shipped the Opentrons OT-2 protocol generator. Net revenue retention hit 142%.

We made three hires that mattered: a former Ginkgo Bioworks application scientist named Pratiksha who built our protocol library, a platform engineer from Benchling who rewrote our queue on AWS Batch, and a head of customer success poached from Twist Bioscience. None of them were ML people. Our investors had pushed us to hire more ML engineers; we hired the people who understood the lab. Trial conversion at the wet-lab level — actual bench scientists, not procurement — went from 18% to 54%. The lesson echoed something I'd read in the AI underwriting story from this cohort: the AI is the easy part, the domain credibility is the moat.

1. 1
   Boring infrastructure is a great place to start.

   "Equipment scheduling" sounds like a feature. For our customers it was a daily, painful, multi-person problem. Pick the boring problem people already have a workaround for.

2. 2
   Free academic adoption is a top-of-funnel, not a market.

   It funds your product instincts and your recruiting pipeline. Don't confuse it with the company.

3. 3
   The data graph is the moat. The product is the gateway.

   Investors who understood our run-log graph wrote the cleanest term sheet. The ones who saw us as a scheduling tool offered half as much.

4. 4
   A good advisor breaks ties faster than you can.

   Joon and I might have argued for another month. One sentence from someone who'd been in our shoes ended it.

5. 5
   Don't rewrite your origin story.

   The original Slackbot still works. Rewriting it would have cost six engineer-months for no customer-visible benefit. Founders who scale-cosplay rewrite working code; founders who scale leave it alone.

What this looked like in practice.

The thing nobody tells you about *Slackbots as wedges* is that the medium teaches you the maximum useful complexity of a feature. If a workflow does not fit in a slash command, it probably does not belong in v1. We carried that constraint into the Next.js rewrite and shipped a dashboard that was 60% smaller than what the design team initially proposed.

The thing nobody tells you about *biotech distribution* is that the buyer's calendar runs on grant cycles, not fiscal quarters. NIH R01 timelines, NSF deadlines, and DOE funding windows determined our pipeline shape more than any sales motion we ran.

The thing nobody tells you about *the Series A* is that the partner who led your round expects you to grow into a different company than the one you raised on. We had three difficult conversations in the first ninety days about whether to chase pharma logos or stay with academic and small-biotech labs. We stayed. NRR proved it right.

The thing nobody tells you about *hiring domain people over ML people* is that the ML stack is increasingly commoditized — Modal, Anyscale, Hugging Face — but a former Ginkgo scientist is irreplaceable. We could rebuild the model in a weekend; we could not rebuild Pratiksha's protocol library in six months.

The thing nobody tells you about *replacing your wedge* is that you do not have to. Customers love the Slackbot. Maintaining both interfaces costs us roughly ninety engineering hours a quarter. It is the cheapest retention investment we make.