Hire for Spikes

Solve for one thing

Eric Glyman hires teenagers known in the Minecraft community for running paid private servers, "the kind of code that wouldn't be in a code base at Google, but they could get things done and self-teach," alongside championship hockey players and competitive mathematicians and programmers, on the theory that excellence in anything hard, however unrelated to the job, is the tell. The framework behind the hires is his hiring model at Ramp, which he explicitly likens to Moneyball: find distinct skill sets and solve just for those instead of buying well-rounded generalists. The mechanics start with very short job descriptions ("if there's one thing that matters, go just for that") and interviews centered on that one thing.¹ Candidates may be raw or have real deficiencies elsewhere so long as they carry high drive and high capability in the spike. Glyman treats a spike as a proxy for drive: "It's so hard to be excellent at anything. If someone is excellent at something, even distinct, weird, or a hobby, it signals they probably have what you're really looking for: the drive to spend not 10 or 100 but 10,000 hours developing mastery."

The composition rule

The non-obvious half is that spiky people have weaknesses, and the framework resolves them at the team level rather than the individual. In Glyman's words: "Jack of all trades, master of none. If you add too many requirements you find average-ish people and can't select for spikes. Across the company, well balanced; but you might want very imbalanced people. You're not looking for the perfect person, you're building the most capable team."¹ Because every function ends up spiky, the heights across all functions are higher than a roster of generalists would produce, and high talent density then compounds.

Interviewed with Keith Rabois, Glyman sharpens the same idea with an F1-racer frame: an early-stage startup is "an F1 racer, it looks weird, it can break down after the race, but it's built to spec and goes very fast," not a finished all-weather product, and taking on a candidate's defects is the price of the speed.² He also frames spiky talent as a mispricing: the market pays up for well-rounded people and underprices extreme spikes bundled with defects, and that gap can be a company's edge. His inversion of the standard loop is the shortest possible job description, since long descriptions and many-skill vetting select for the average, and most of hiring happens as hunting before the interview. A second spike example from that interview, alongside Calvin Lee: Veral Patel, who sold a company to Apple at around fourteen and now works on Shortcuts.

The CTO's version

Karim Atiyeh gives the same philosophy from the engineering-recruiting seat and adds mechanics Glyman leaves implicit.³ He describes hiring for slope as well as spikiness and wanting to "assemble the Avengers," each person with a clear superpower. He also describes exploiting asymmetric information: competing for talent against Facebook and Google with less money and brand at Paribus, the founders leaned on knowing which MIT and Harvard freshman-year classes were brutal, recruited freshmen where rivals chased juniors, and used a great grade in a known-hard first-year class as a high-signal spike. His interview technique is to probe the single thing a candidate claims to be great at until it reveals both how good they are and how well-calibrated a judge they are of their own ability. The framework's payoff shows up as versatility as much as competence: an early hire named Calvin Lee arrived as a one-month intern who had left high school early for the informatics olympiad and finished college in two and a half years, then went from engineer to sales to running forward-deployed engineering.

Relationships and tensions

The framework converges with the a-player framework in rejecting well-rounded hires for inch-wide, mile-deep specialists, and it pairs with hire for give-a-damn on drive over credentials. It runs in open tension with hire junior generalists for potential, the opposite intake prescribed for an early growth function, a tension the archive holds without a stated reconciliation beyond the common stage-and-role distinction. Ramp's own hiring deep cut adds unreasonably short job descriptions, five mispriced talent pools, and strong-yes-or-bust interview design.⁴ Glyman and Atiyeh both note limits: spiky teams require strong composition and glue, and hobby-based spike signals can encode bias in who gets to spend 10,000 hours on a given pursuit.