Average Is Not Good Enough

Deep-tech companies buy equipment and supplies constantly, so giving everyone a card or approving each purchase can slow work and obscure costs. Hodak argues for setting budgets and actively managing procurement, then tracking shared materials and experiment costs. Science built internal software called Helix to connect purchasing, manufacturing, and cost data; one wafer iteration in his example cost $40,000.

Hodak says early hiring should draw on the network around a company’s founding scene, then expand beyond it. At Science, employees vote on applicants, followed by a companywide phone screen, homework, and a full interview. The process spreads review across the team to reduce bottlenecks, and the company uses regular “would you hire this person again?” votes as an ongoing signal about hiring outcomes.

Hodak’s central operating lesson is that iteration speed shapes a company’s chances of success, and infrastructure sets that speed. Purchasing, recruiting, budgeting, and quality systems may sound routine, but weak processes can make a growing organization unwieldy. He also says founders must make their own judgment calls, especially when a high-stakes decision has no clear adviser.

On choosing industry or academia, Hodak says the answer depends on the field: a PhD can make sense where the work remains primarily academic, while a strong company can teach founders how organizations execute. He recommends evidence of concrete achievement and says hands-on work helps people build broad understanding across fields such as neural interfaces. For long-horizon startups, he advises pricing the next value inflection point’s experiments and raising about twice that amount.

Hodak says Science puts company context into Helix so people and AI agents can connect information across purchasing, quality, and manufacturing. He describes AI as a team multiplier, especially for coding and regulatory work such as identifying standards and assembling evidence tables. He argues that internal software can fit a company’s processes more closely than commercial tools, particularly when agents make custom development more accessible.