Anthropic confirmed it runs its own wet biology lab. The build-versus-buy question behind it.
On September 18, 2026, TechCrunch reported that Anthropic has a wet biology lab in the Bay Area where it can use its AI models to run physical experiments, which the company confirmed to the outlet. TechCrunch quoted Anthropic's head of life sciences, Eric Kauderer-Abrams, telling Reuters: "We believe that to do biology, the final test is still, and will be for a while, in real lab work." He added, "We absolutely are doing that today." According to TechCrunch, he said the lab operates like most biotech labs: Anthropic conducts some research there while also working with external partners. The company declined to give specifics on what the lab is working on, though it did say the main focus was fundamental biology, not drug discovery. TechCrunch also noted that Anthropic bought Coefficient Bio, a stealth AI biotech, in April.
Set aside the biology, and the louder argument TechCrunch records about whether a company that warns about AI risk should run such a lab at all. What remains is a decision about what to own. By its own account, Anthropic built the piece where the final test happens and kept external partners for the rest. TechCrunch's reading is that the boundary is also commercial: Anthropic doesn't want to give the appearance of competing with the pharma industry, where it has numerous major customers and partners, and it has already faced backlash for launching products perceived to compete with those of its customers. So the company had to decide what it needed to control, what it could leave to others, and who would be affected by building. Hiring panels probe for that same behaviour when they ask one of their most common technical-strategy questions.
Why they ask it
The question, as reported in the L8 Loop corpus, is: How do you choose when to build in-house vs. using a third-party solution? On the surface it's asking for a framework. What the panel is really checking is whether the candidate has made this decision with real money and real people attached.
A panel listens for three things. The first is whether the candidate can tell work that differentiates the business from work that merely has to exist. The second is whether they count the full cost of each path. Building means a permanent maintenance burden and the opportunity cost of the engineers who carry it. Buying means integration work, a dependency on someone else's roadmap, and less leverage at every renewal. The third is whether they noticed the people the decision touched: the team that wanted to build, the vendor relationship, the partner who might read an internal build as a threat.
The trap
The usual weak answer is a list of criteria. The candidate names cost, time to market, core competency and vendor lock-in, maybe adds a weighted scoring matrix, and stops. All of it is correct, and none of it shows that the candidate has ever had to choose. Panels hear this answer constantly, and it tells them the candidate has read about the decision.
The second trap is a slogan. Buy everything that isn't core sounds decisive, but the follow-up is always how the candidate decided what was core, and the slogan has no answer ready. The mirror-image failure is the builder's story in which the vendors were all inadequate and the team's own system was a triumph. The interviewer is left wondering whether the alternatives were ever seriously evaluated.
A strong answer starts from one specific decision and lets the framework come out of how that decision was made, so the principle is backed by something that happened.
Applying STAR-T
Situation. Set the stakes and the constraint in two sentences. An illustrative version: the platform team's homegrown feature-flag service had become a source of incidents, the engineer who wrote it had moved on, and a product launch depended on safer rollouts within the quarter. Include whatever made the choice contested. Here, several senior engineers wanted to rewrite the service properly, and finance was wary of another vendor contract.
Task. State what was personally owned. That wasn't the evaluation of vendors. It was the recommendation and its consequences. A strong candidate says plainly that the call was theirs to make or to bring to a named decision-maker, and that they would be accountable for the outcome either way.
Action. Most of the answer belongs here, and it should show the test that was applied. In the illustrative case, the manager asked where the company's advantage actually lived. Customers didn't choose the product for its flag evaluation. They chose it for the experimentation logic layered on top, which encoded how the business made decisions. So the team bought the commodity layer and kept building the part where the advantage lived. The manager priced both paths over several years, including on-call load for the in-house option and a negotiated exit clause for the vendor. They also sat down with the engineers who had wanted to build and gave them ownership of the experimentation layer, so that the engineers didn't read the decision as a loss of interesting work.
Result. Report the outcome against what was promised: the launch went out on schedule, flag-related incidents stopped, and the engineers who were freed up shipped the experimentation work that had been stalled. If a number exists, give the real one. If it doesn't, say what was observed.
Trade-off. This question rewards naming the cost. The vendor's roadmap now set the pace for certain features, the contract became a recurring line that someone had to defend, and one engineer who had wanted the rewrite stayed unconvinced. A candidate who says this shows that the decision was weighed honestly.
The follow-up that breaks weak answers
The follow-up is usually some version of: What would have to change for you to reverse that decision? Candidates who recited a framework have nothing to reach for here, because a checklist doesn't include conditions for reversing itself. Candidates who made a real decision know those conditions already. They might be a pricing change that crosses a threshold, a product direction that turns the commodity into a differentiator, or a vendor acquisition that puts the dependency in a competitor's hands.
A second version asks about the decision that went the other way: Tell me about a time you built something you should have bought. The strongest candidates keep that story ready alongside the first, since a candidate who has only ever been right about build versus buy probably hasn't made the decision often.
Score your answer against the director’s bar
Q: How do you choose when to build in-house vs. using a third-party solution?
Bank one build-versus-buy decision as a story and rehearse it until the trade-off comes out unprompted. Try it free →