"That is the nature of early-stage recruiting," Jian says. "We need volume."
"Job postings," Alex says. "LinkedIn, AngelList, university job boards. We cast a wider net."
Jian nods. "Agreed. But the description must be precise. We cannot afford to waste time on unqualified applicants."
They open a shared document. The blank page stares back at them.
"Version one," Alex types. "Inferthermic is a fast-growing AI safety startup seeking talented engineers to join our mission of building safer artificial intelligence."
Jian reads it. "Generic. Every company says this."
"Version two." Alex deletes and retypes. "Inferthermic develops post-processing alignment systems for large language models. Our entropy-reduction technology reduces risky model outputs by 98% without retraining or keyword filtering."
"Better," Jian says. "Technical accuracy. But lacks mission."
"Version three." Alex edits again. "Inferthermic builds the safety layer for enterprise AI deployment. Our systems process millions of LLM outputs daily, preventing harmful generations while maintaining model capabilities. We reduce risky interactions by two orders of magnitude through mathematical alignment, not heuristic filtering. Backed by leading investors. Growing from 3 to 15 people. Seeking founding team members in engineering, research, and operations."
Jian reviews. "Specific. Credible. Emphasizes the technical differentiation." He pauses. "Add something about the team. People join teams."
Alex adds: "Founded by UCSB and Berkeley researchers with backgrounds in statistical mechanics, distributed systems, and enterprise sales. Alumni of Google, Andreessen Horowitz portfolio."
"Accurate," Jian says. "Maddy will verify the sales claim."