The question that predicts whether you can scale

Most conversations about scaling a team fixate on headcount. How many engineers, what the org chart looks like, when to add a manager. There is a better question, and it predicts more: how many people on your team can take an initiative from a vague idea all the way to shipped, across function boundaries, without being carried? That number is almost always small, and it is the real ceiling on how much your company can take on at once.

The people who can do that are rare, and most organizations have far fewer of them than they think. Confuse them with senior titles or high output and you will keep hiring in a way that does not move the ceiling.

Barrels and ammunition

Keith Rabois has a framing for this that is worth borrowing: barrels and ammunition. Most people in a company are ammunition. That is not an insult. Ammunition is valuable, necessary, and often highly skilled. Give an ammunition-type engineer a well-defined problem and they will execute it well. What they do not do, by disposition or by role, is aim. A barrel is the person who takes an ambiguous goal, points it at reality, and fires: they turn "we need to solve X" into X actually shipped, dragging the ammunition along with them.

Rabois's observation is that a company's capacity to do new things is gated by its number of barrels, not its total headcount. Most companies have very few. Revisiting the idea on Lenny's Podcast in 2026, he estimated that even at a sizable company only a dozen or so people can carry a project from zero to launch, and one founder he asked put his own company at two. When you find one, the highest-leverage move is to load them with more ammunition, because a barrel with support multiplies, and a barrel starved of it is wasted.

Why founding engineers are barrels

If you have been a founding engineer, you have been a barrel by necessity. Nobody hands the first engineer a scoped ticket. The job is to take "we need a product" and turn it into a shipped product, crossing whatever function boundaries that requires, deciding what to build as much as building it. That is the barrel role in its purest form, which is exactly why early hires punch so far above their headcount. They are not adding capacity to a plan. They are the reason the plan reaches done.

This is also why counting barrels matters more as you scale. The scrappiness that felt automatic when the company was five people does not survive contact with process unless someone protects it. Growth adds ammunition quickly and barrels slowly, and the ratio determines whether the company still ships new things or just maintains old ones.

How to spot a barrel

Barrels are not reliably the loudest people, the most senior, or the highest-output. Volume of work is an ammunition signal. The barrel signal is completion across ambiguity: a track record of initiatives that reached done specifically because that person drove them, end to end. Look for the people who close open questions instead of escalating them, who cross into product or ops or design without asking permission because the goal required it, and whose involvement is the reason something stopped being stuck.

The tell is not "who is impressive in the room." It is "which initiatives would have died if this person had not carried them." Ask that about your last few real wins and the barrels become obvious, and they are often not the people the org chart would predict.

The mistakes that waste barrels

Two common moves destroy barrel capacity, and both look like normal management. The first is promoting a barrel into a management seat where the job becomes coordinating ammunition rather than driving initiatives. You take the person whose whole value was firing, and you reassign them to inventory. Some people can do both; many cannot, and nobody checks first. The second is hiring ammunition when the gap was a barrel. The team is stuck on an ambiguous, cross-cutting problem, so you add capable executors, and the problem stays stuck because it never needed more hands. It needed someone to aim.

Both mistakes come from counting headcount instead of barrels. If you cannot see the distinction, you cannot see why adding good people sometimes changes nothing.

The opposite failure: deciding only barrels matter

Overcorrect and you get a different problem. Treat barrels as the only people who count and you devalue the ammunition the barrels depend on, which is both wrong and self-defeating, because a barrel with nothing to fire is just a person with opinions. Push every engineer to "act like an owner" on work that genuinely needs focused, deep execution and you get a company full of people half-driving initiatives and nobody finishing anything at depth. A company of all barrels and no ammunition is not a dream team. It is a room full of started, unfinished things.

The point is the ratio and the fit, not a hierarchy of worth. Ammunition doing deep, skilled execution is not a lesser version of a barrel. It is a different role, and most of the actual work is theirs.

Growing barrels, not just hiring them

Barrels are not purely born, and you do not only have to hire them. You can make them, and most organizations never create the conditions. The way a barrel forms is by being handed a small but real initiative, end to end, with enough air cover to survive being wrong, early enough that they learn to close ambiguity instead of escalating it. That is the founding-engineer experience, compressed and offered on purpose. Give a promising ammunition-type engineer a contained problem that has no clear owner, protect them while they drive it, and you find out quickly whether they can aim. The ones who can are your next barrels, and you grew them instead of competing to hire them.

Where people will push back

"This just relabels senior or staff engineers."

Sometimes they overlap, often they do not. Plenty of staff engineers are exceptional ammunition, deep technical execution on scoped problems, and plenty of barrels are mid-level people who happen to drive things to done. Seniority measures depth and scope of technical skill. Barrel-ness measures whether someone closes ambiguity end to end. Different axes.

"You're devaluing individual contributors."

The opposite. Ammunition is where most of the real, skilled work happens, and the framing says so directly. The argument is against miscounting, not against ICs. Mistaking a barrel gap for a headcount gap wastes everyone's time, including the ICs you hire into a problem that was never theirs to fix.

"Not everyone can or should be driving initiatives."

Correct, and that is the point. A company needs mostly ammunition. The failure is not having too little of it. It is not knowing how few barrels you have, and staffing as if adding capacity were the same as adding drive.

When this does not apply

In a pure execution phase, a well-defined roadmap with clear scope and known problems, you may genuinely need ammunition more than barrels, and hunting for barrels is a distraction. Barrel density matters most in ambiguity: zero-to-one products, turnarounds, and any moment where the hard part is deciding what to do and dragging it to done rather than executing a known plan. Match the hire to the phase. The mistake is only assuming the answer is always more people.

The takeaway

Headcount is easy to count, which is why companies optimize it and then wonder why capacity did not follow. The number that actually gates what you can take on is smaller and harder to see. Your company's real constraint is not how many people you have. It is how many can take an idea all the way to done. Find those people, stop wasting them on seats that neutralize them, give them more to fire, and build the conditions that turn your best ammunition into the next barrels. Then count again in a year.

Further reading

• Rule-Breaking Is a Startup Virtue Until the Rule Is Load-Bearing The judgment a barrel needs when the domain is regulated.
• AI Isn't Replacing Engineers. It's Rewriting the Org Chart. Where judgment becomes the bottleneck when execution gets cheap.
• Keith Rabois, barrels and ammunition, via First Round Review and his 2014 "How to Operate" talk. Recent framing from his 2026 Lenny's Podcast appearance.