The Skunk Works Renaissance
The Skunk Works Renaissance
Days are numbered.
In the summer of 1943, Lockheed had no room left for its most important project. So a young engineer named Kelly Johnson cleared a patch of ground next to the wind tunnel in Burbank, built walls out of old engine crates and scrap lumber, and threw a rented circus tent over the top. Inside, a hand-picked crew built America’s first jet fighter in 143 days.
We’re about to need a lot more circus tents.
The Original Circus Tent
The setup matters more than the airplane. Process over product.
German jets had shown up over Europe and the Army Air Forces needed an answer fast. Lockheed won the job to build a fighter around a British jet engine, with a deadline of 150 days. Kelly Johnson had wanted for years to run an experimental group under his own direct control, and this was his shot. According to Lockheed’s own history, work started on a handshake. The formal contract didn’t arrive until four months after the team had already begun.
There was no factory space, so the team worked in that makeshift shed under a circus tent, drafty and unheated, next to a plastics plant that made the whole place stink. Johnson hung a big red countdown sign on the back wall that read “Our Days Are Numbered.”
The National Academies memoir of Johnson puts the crew at 23 engineers and 103 shop mechanics, and notes that Lockheed’s top management gave him a free hand. The XP-80 was delivered in 143 days, seven days early.
The name came from the smell. An engineer named Irv Culver was a fan of the “Li’l Abner” comic strip, which had a foul-smelling backwoods still called the “Skonk Works.” One day he answered the phone “Skonk Works, inside man Culver speaking,” and it stuck. The comic’s lawyers objected, so it became Skunk Works. That same group went on to build the U-2 and the SR-71 Blackbird.
Johnson later wrote down how he ran it in Kelly’s 14 Rules. The manager gets “practically complete control” and reports to a division president or higher. Keep reports to a minimum. Use a small number of good people, 10% to 25% of what a normal program would use. And my favorite: “The number of people having any connection with the project must be restricted in an almost vicious manner.”
A small team, real authority, very little paperwork, and serious air cover from top brass.
The Bigger Version, on a Mesa
Months earlier and a long way east, the same pattern played out (cinema history may have made this the more culturally referenced version). Oppenheimer won Best Picture at the 2024 Oscars, and it told the story of J. Robert Oppenheimer, General Leslie Groves, and an isolated mesa in New Mexico.
In October 1942, Groves asked Oppenheimer to run the new weapons lab, even though, as the Atomic Heritage Foundation notes, Oppenheimer had little managerial experience. In November they picked Los Alamos. Less than three years later came the Trinity test. Groves brought the authority, the budget, and the secrecy. Oppenheimer brought the scientists, and pulled physicists, chemists, and engineers into one place to work on one problem.
Again, process over product. I’m not trying to make Los Alamos (or what came out of it) a feel-good business story. What I am pointing at is the operating system: a cross-disciplinary group, cut off from the normal machinery, given a clear mission and backing from the very top, moving faster than anyone thought possible.
That’s the model AI is going to demand of us, and it’s going to look like a Skunk Works Renaissance.
Fast > Slow
Corporations are likely to get bifurcated, especially the ones that aren’t net-new.
Think of any business that’s been around for more than five or six years and is doing eight or nine figures in revenue a year. They’re in a weird pinch. They already have a chassis built, and it was built for (and in) the pre-agentic era: the org chart, the approval chains, the tool stack, the SOPs, the vendor contracts. All of it made sense when it was designed. And a lot of it is about to be dead weight.
For a long time the rule was that the big beat the small. Then the startup world told us the small beat the big. I honestly don’t think either is right anymore. The fast and nimble are going to beat the slow and hardened.
Newer companies, smaller companies, with fewer employees, less baggage, and a CEO who can decide on tooling on a Tuesday and have it rolled out by Wednesday - it’s a dynamic OS that minimizes procurement cycles and skips twelve-month security reviews and decisions by committee. That agentic-enabled pace, in the very near future - if it hasn’t already taken a bite out of a moat - is going to take a bite out of yours.
And it’s not going to be measured in buy-cycles or years: it’s going to be measured in days. Numbered days.
The Tremor in Enterprise Software
Back in December 2024, Microsoft CEO Satya Nadella told the BG2 podcast that the very idea of business applications could “collapse” in the agent era, because they’re “essentially CRUD databases with a bunch of business logic,” and that logic is moving into AI agents.
Then 2026 happened. Markets and investors coined a name for it, the “SaaSpocalypse.” CNBC reported in August that Salesforce had lost more than 40% of its value since the end of 2024, despite accelerating revenue growth and consistent margins, and that a major software ETF had its worst quarter since 2008 in Q1. Airtable, once valued at nearly $12 billion, agreed to sell for less than $1.3 billion. The worry is that tools like Grok Bot, Claude Code, and OpenAI Codex will hollow out the economics of expensive software over time.
To be fair, there’s been a big rebound since. Hargreaves Lansdown’s analysis this month found the average S&P 500 software stock fell 27% to its April low and has since won back most of it. But they also point out that profits were broadly fine during the selloff.
My read? It’s not about what’s happened, but what is liable to happen. The markdowns are focused on a future where AI is easiest to see as a substitute, including design and creative tools, internal processing systems, tools with nice interfaces and basic tasks that aren’t worth the bloat - the agentic vice is tightening on them with each new model update.
So no, I’m not saying Salesforce or Oracle or Adobe disappear next year. My contention is that the threat is real, it’s early, and it applies far beyond software vendors. Every established company running on a pre-agentic chassis has the same exposure. The software stocks are just where it showed up first, because that’s where people are paid to price the future.
How the Big Adapt
So how does a company with eight, or ten, or thirty-plus years of history and aging infrastructure fight back? You can’t rip out the chassis overnight. The board won’t let you, the customers won’t tolerate it, and honestly, you shouldn’t.
My answer: it’s the Skunk Works model.
Small SWAT teams, cross-functional, pulling from different domain expertise, set up next to the main business rather than inside it. Their job is to find the processes, products, workflows, and SOPs that no longer make sense in an agentic world, and prove a better way fast enough that the rest of the company can’t ignore it.
Kelly’s rules translate almost one for one: practically complete control, a small number of good people, minimal reports, and reporting to the top.
Two Non-Negotiables
There are two roles that are mission critical. The first is the obvious one - the CEO, who brings their authority and weight, full stop. The second is more interesting - it’s a role I’d make sure is on every one of these teams, and the one I think is going to become the superstar in this period - it’s the turbo-charged traditional business development executive.
Why? A great BD exec understands both ends of the business and is a ridiculously good translator. They know the customers you serve and the value props you actually need to deliver. They also know the internal mechanics, where deals get stuck, which handoffs break, and why the contract takes six weeks to get through legal. And the really good ones can speak some of the engineering lingo, helping developers see what’s needed today, and down the road. And the really really good ones know the AI toolkit, see what’s happening, and can already foreshadow what it means to the business.
That combination is rare and it’s exactly what a skunk works team needs.
Engineers can build almost anything now. Golly, the average Joe can build almost anything now. The hard part? The hard part is knowing what’s worth building and where the friction really lives. The rockstar BD exec understands the vibes at a revenue level and can speak to the points of friction both internally and externally.
The ones with a real bent toward AI, the ones who are actually playing with these tools at night and seeing their horsepower firsthand, are going to be the new MVP of the skunk works ecosystem. I’ve spent a lot of my career in and around that seat, across startups and big enterprise, and I’ve never seen a moment where that profile mattered more.
What To Do…
If you’re overseeing one of these companies, if you’re feeling the new guys clipping your heels, if you’re seeing forecasts be rewritten... there’s a playbook. And it might feel painful or unorthodox, but it’s necessary and I think this is what it looks like:
✱CEO endorsement and oversight. Not a memo, real sponsorship. Kelly Johnson reported to the top and Groves had Washington behind him. If this team reports into a middle-management committee, it’s dead on arrival.
✱Cross-functional team members. Engineering, operations, finance, customer-facing people, sitting together and working on the same problem. A small number of really good people* (see below).
✱A secure sandbox. A walled garden where the team can do playful (and at times even dangerous) tinkering with the toys, without putting customer data or production systems at risk.
✱A supercharged BD exec with an AI bent. Someone who knows the customer, knows the internal plumbing, and has been getting their hands dirty with the tools.
✱Permission to disintermediate. An explicit license to go after aged processes, workflows, product lines, and SOPs that no longer make sense, including the ones somebody senior is attached to.
The Atrophy Apocalypse
The sandbox is the part that will feel nerve-wracking for a lot of executives and boards. I get it. Letting people be a little dangerous is the opposite of what most risk frameworks are designed to do. But that’s the point of the walled garden. You contain the risk so you don’t have to kill the creativity. If you get the right CEO buy-in and let teams be playful, creative, and even a little dangerous, I firmly believe you’ll be surprised how much of the old way of working simply falls away.
Kelly Johnson’s team worked under a sign that said “Our Days Are Numbered.”
Of course, he meant their looming deadline. For a lot of established companies right now, however, I think it can be read just as literally. The fast ones will find their circus tent. The slow ones run the risk of going extinct by atrophy.
*So what does ‘good people’ actually mean? You need ones that know the systems, but aren’t beholden to them. They are creative. They are passionate. They - as David Epstein puts it - have range.
✱David Epstein’s “Range” highlights kind vs. wicked environments. Specialists excel where the rules are stable and feedback is fast, like chess or golf. Generalists win where the rules are shifting and past patterns mislead. The agentic transition is 100% a wicked environment, and that explains why the generalist’s value is rising now.
✱Want to get even more scientific? If your company uses a DISC profile (or similar) tool, you’re looking for a mix of Pioneers and Skeptics. The Di, which DISC calls “Pioneering,” is bold, comfortable taking chances, and persuasive enough to bring people along. The DC is the blunt problem solver who looks at an SOP and asks why it exists. Those are the people who aren’t beholden to the status quo.
✱However, a tent full of pioneers starts fast, argues constantly, and finishes little. So add at least one high C (Conscientiousness), who is analytical, precise, and keeps the sandbox secure, and one high S (Steadiness), who is patient and supportive enough to hand the work back to the main business without scaring it.