Hiring the first ten engineers

The first ten engineers in a services company are not ten hires. They are the specification for every hire that follows, because everyone after them is assessed by someone from that group against a standard they set implicitly.

The decision that mattered most was refusing to lower the bar for speed. We had signed work and not enough people, which is precisely the pressure that produces a bad hire, and the cost of one at that size is not one person underperforming — it is the bar moving.

We replaced algorithm puzzles with a real task from a real codebase, time-boxed and paid. It predicts the job better and, more usefully, it tells the candidate what the work is actually like. Several people withdrew after seeing it, which saved everyone a difficult month.

The domain interview is run by the person who will work with the hire, never by a recruiter and never by me. If a practice lead has to live with the decision, they should own it, and they will not accept a marginal candidate to hit someone else's deadline.

The thing we would do differently is writing things down. For the first two years the standard lived in a handful of heads, and it worked until we needed to hire in parallel. We should have documented the assessment rubric a year earlier than we did.

The measurable outcome is retention. Most of that cohort is still here, and several now run practices. The uncomfortable version of the same fact is that a company hires its own future interviewers in the first year, whether or not it means to.