Luca Ferrari runs Evernote with 20 people.
Bending Spoons agreed to buy the company in November 2022, when it had 350 team members and was roughly breakeven on $90 million in revenue. The deal closed a few months later. “Today, Evernote, we ran it with about 20 people,” Ferrari says, almost in passing.
He’s the co-founder and CEO of Bending Spoons, the company running Evernote, WeTransfer, Meetup, AOL, and more than 50 other businesses off one internal operating system. He walked through how his company hires, why it deletes almost everything on purpose, and what a Slack bot fixed in 3 minutes flat.
I watched the full interview so you don’t have to. Here are the 10 that matter.
together with upwork:
Keep the core small. Let your AI find the experts for the rest.
Upwork’s new MCP connector brings freelance experts into Claude and ChatGPT, so you hire the way section 1 argues for: on evidence.
▫️ Your AI drafts the job post
▫️ It ranks freelancers on your criteria
▫️ You decide from portfolios, past reviews and job success scores
▫️ Payments go out on milestones through Upwork’s trusted contract rails
“I found the right freelancer in 15 minutes.”
Upwork client
1. Why Bending Spoons Hires 22-Year-Olds Over VPs
Most companies pay for experience. Ferrari bets on hunger, even from people who have never held a job.
“I’d rather hire young graduates, find someone good, and then saturate their capacity.”
His reasoning starts with an uncomfortable idea: experience is worth less than most hiring managers assume. Software, design and customer expectations move fast enough that what someone learned five years ago has often gone stale by the time they need it again.
So Bending Spoons finds someone sharp early, piles on responsibility faster than any normal career ladder would, and expects them to keep up. It’s a demanding bet for everyone involved.
It only works if you can spot who’s sharp before they have a résumé to prove it. That’s the harder problem, and Bending Spoons spent years building a scoring system to solve it.
2. The 100-Signal Score Behind Every Hire
Without a track record to judge, you need another way in.
“We’ve built a science out of studying people’s track records. In someone’s application, we identify over 100 different signals, and through those, we predict their long-term potential.”
Each signal is weak on its own, and a GPA by itself predicts very little. Stack a hundred of them, though, and the weak signals add up to a strong one.
Some are obvious. The one Ferrari mentioned surprised even the interviewer: how a candidate treats the recruiting coordinator who schedules their interview predicts collaboration on the job better than how they come across to the interviewer.
You can steal that one without building the other 99. Watch how candidates treat the person with zero power over their outcome.
To build your own version of this scorecard before your next hiring round:
▫️ The Free AI Hiring Kit for Founders
▫️ Most Founders Hire Their First Engineer Too Early, Too Fast and for the Wrong Reasons
3. One Hiring Team for 50 Companies, and No Bonus for Closing Hires
Most companies let the hiring manager pick. Ferrari hands the decision to a central team with nothing to gain from rushing.
Asked what incentives his talent managers work under, he keeps it short:
“There’s none. Just trust.”
Nobody earns a bonus for headcount or for closed hires, and that’s deliberate. Left alone, Ferrari argues, a hiring manager takes the first adequate person to relieve their own workload. They also favor experienced candidates over greener ones who would often outperform them within two years.
A central team sees every hire across every Bending Spoons business, a sample size no single manager could match. And because waiting costs them nothing personally, they can hold out for the better candidate.
The version for a 15-person startup: whoever makes the hiring call should gain nothing from closing it fast.
4. The Math Behind Giving People Too Much Work
Ferrari’s case for overloading people rests on arithmetic.
“The bigger the universe of work item<s you can prioritize from, mathematically, the higher the returns on your time you’ll deliver.”
Picture a list of 10 tasks and time for only a few of them. Add an 11th. If you choose well, your best picks can only improve or stay the same, because the original 10 are still there, and the new task might turn out to be the best one on the list. More options raise the ceiling and leave the floor where it was.
The second effect matters just as much. People who only ever get work they can comfortably finish never practice choosing among too many good options. Ferrari wants that skill learned in month two on the job, while mistakes are still cheap, instead of in year ten.
5. Extreme Ownership Is the Only Hiring Filter That Matters Once You Pass the Bar
Bending Spoons tests for raw intelligence first. Past that point, Ferrari says something a lot less quantifiable decides everything.
“After you pass a certain threshold, it’s almost all about how badly you want it.”
He brings up Rafa Nadal, a player most experts didn’t rank inside the top 50 talents of his generation, who became one of the 3 best to ever play the sport. The gap wasn’t ability. Nadal wanted it more than the 49 players ranked above him.
“I would bet there’s less than a 1% chance you fail to have an amazing career because we know you’re smart.”
Ferrari borrowed the term “extreme ownership” from Jocko Willink, though he defines it slightly differently: caring, in your belly, about being the best at what you do. He’d rather work with a smaller team of people who feel that way than a larger one that doesn’t, because the feeling is contagious, and its absence is too. It’s a close cousin of the high-agency trait that shows up in almost every founder story worth reading.
6. Complexity Needs a Permission Slip at Bending Spoons (Just Ask What Happened to Job Titles)
Left alone, organizations get more complicated over time. People rarely remove a rule, a step, or a headcount line.
“The burden of proof is on those who want to add complexity, which helps reduce the addition of complexity dramatically.”
If you want to add a process, a tool, or a new layer of approval, you have to argue for it. If you think the company should stay as it is, you don’t have to prove anything.
You just raise a flag. That asymmetry is the whole mechanism.
It’s not a hypothetical. Bending Spoons doesn’t have job titles. When someone asked why the company needed them, nobody could name a benefit beyond bragging rights and a convenient line for a resume.
So they cut titles entirely and told employees to write whatever they wanted on LinkedIn. The company has never brought them back.
It’s the same instinct behind Elon Musk’s approach to deleting parts: the fastest way to simplify a system is to remove a piece of it completely, not trim around its edges. Ferrari makes the same case for what a business can strip out and still hold its moat.
7. The Same 50-Person Team Handled a Business 4 Times the Size (Here’s the Math Behind It)
In 2023, roughly 50 Bending Spoons employees spent a year rebuilding Evernote’s codebase, infrastructure, and monetization. In 2026, a team of the same size did it again on a business worth 4 times as much.
“We did broadly speaking the same thing with Gmail, with roughly the same number of people, 50 to 60. But Gmail is roughly $400 million in revenue, so approximately 4 times as large.”
Evernote was generating a little under $100 million when Bending Spoons bought it. Gmail, one of the newer names in their portfolio, brought in around $400 million. Same headcount, same timeline, 4 times the scale.
The reason isn’t magic. The technology platform underneath every acquisition keeps getting better, so the marginal team gets more capable with every deal, not less. Complexity of a business and the headcount needed to run it don’t move together the way most operators assume, which is exactly the assumption an acquisition readiness model forces you to test before you sign anything.
Before you assume a bigger acquisition automatically means proportionally more headcount, run the actual math:
▫️ The M&A Accretion/Dilution Model Every Founder and Investor Should Run Before Any Deal
▫️ Accretion, Dilution, and the Cold Hard Truth of M&A Math
▫️ The Anatomy of an M&A Deal Inside an Excel Template
8. 800,000 People Applied to Work at Bending Spoons Last Year (Fewer Than 300 Got In)
Talent density doesn’t happen by accident, and it isn’t cheap to build. But it compounds in a way few operators account for.
“We got 800,000 job applications last year. We hired fewer than 300 people.”
That ratio is the actual explanation for why Bending Spoons can run Evernote with 20 people while the previous owners needed 350. It’s not that the old team was lazy or incompetent.
They simply couldn’t attract the applicant pool Bending Spoons can. Nobody applies to run a legacy note-taking app the way they’ll apply to a company known for hiring the best engineers in the market. It’s the same dynamic behind Lovable’s climb to $400 million in ARR with 146 people, or Hyperliquid running a live exchange with 11: small teams don’t win because they’re scrappy, they win because talent density beats headcount.
Ferrari calls it an arbitrage: access to talent that a standalone business, even a well-run one, structurally can’t match. The acquisition doesn’t create the advantage. The employer brand built years before the acquisition does.
9. Bending Spoons Says It Hasn’t Been Outbid in 5 Years
Ferrari’s approach to negotiation sounds almost boring, and he thinks that’s the point. One number, stated plainly, rarely moved by more than 5% to 10%.
“We have been able to deliver the extremely high returns we have while winning essentially all winnable sales processes because of that massive advantage as an operator.”
The confidence isn’t posture. It’s downstream of everything else in this piece: better talent access, a shared technology platform, and the ability to run a business better than almost any other buyer on the table.
“I don’t believe to this day that we have ever been outbid.”
If a business will be worth more under your ownership than under anyone else’s, you can offer more for it and still come out ahead. Sellers who try to shop the offer around usually come back to the same number months later, which is the same logic that shapes every term sheet negotiation once both sides know who actually needs the deal more.
Before you walk into your own negotiation assuming a lower opening number wins, run the numbers on what winning the deal is actually worth:
▫️ IRR vs Return Multiple Explained: The Venture Math Every Founder and VC Must Master
▫️ The Venture Capital Method: How Investors Really Value Startups
10. A Slack Bot Fixed an Evernote Bug in 3 Minutes (While the Ticket Was Still Being Written)
Bending Spoons built an internal AI agent called Old Spooner that lives in Slack and has the exact same system access as the employee using it.
“In maybe 3 minutes, our general manager for Evernote identified and fixed a bug.”
She flagged a bug in a feedback channel, then asked Old Spooner to check whether it was a one-off or a widespread issue, search the codebase for the root cause, draft a fix, and ping the engineer who’d need to approve it. All of that happened before she’d normally have finished writing the ticket.
Ferrari doesn’t frame this as replacing engineers. He frames it as removing the translation tax between people: the days lost explaining an idea, waiting for someone else’s calendar, and re-explaining it when the first attempt misses.
Machines don’t have a calendar to wait on. It’s the same shift behind giving an agent its own computer instead of a chat box, or Brian Armstrong running 1,200 agents inside Coinbase: once an agent has the same access a person would, the bottleneck stops being how fast a person can explain the problem.
This is a fourth grouped list rather than the usual three: the piece closes on a hands-on AI section, and that’s the single highest-intent build-this-now moment in the whole draft. If you haven’t shipped one of these yet, start here:
▫️ Ship Your First AI Agent in a Day
▫️ 100 AI Agent Ideas You Can Build This Weekend
▫️ The 5-Agent Sales Team You Can Build This Weekend
▫️ Claude Cowork Now Runs a $10,000/Month SEO Agency
What This Means for You
Bending Spoons built the operating system first and let the acquisitions follow it, not the other way around.
▫️ Founders: Stop assuming experience is the safest hire. Build a working scoring system for the signals that actually predict performance in your context, then trust it more than a resume. A free AI hiring kit is a faster starting point than building one from scratch.
▫️ Investors: When you’re evaluating a serial acquirer, ask what happens to headcount and revenue on the second and third deal, not just the first. Run it through the same lens as an M&A accretion and dilution model: that ratio tells you whether the advantage is structural or a one-time integration story.
▫️ People in tech: If your company has more process than it did 2 years ago and nobody can say why, that’s the drift Ferrari is describing. Ask what would happen if you removed one piece of it entirely instead of trimming around the edges.
The 5 Principles to Steal
Hire for hunger. A resume tells you what someone has already done. It tells you almost nothing about what they’ll do once nobody is checking.
Score like a hedge fund. No single signal proves anything. A hundred weak signals, tracked honestly over time, become a genuine edge.
Give people too much. The math only fails if you’re bad at picking what to add. Assuming you’re not, more options can only raise your ceiling.
Force complexity to earn it. Removing a rule should be free. Adding one should require an argument. Most companies run that backwards.
Build platform, then empire. The acquisitions look easy once the operating advantage exists. They look impossible before it does.
Ferrari is chasing one answer: is this business better off with us than with anyone else on earth? Once that answer is consistently yes, the rest of what Bending Spoons does stops looking unusual. It starts looking inevitable.
Ferrari covers Henry Singleton, IPO timing, and Elon Musk's deletion obsession in the parts we cut. Listen to the full episode.
If this breakdown saved you an hour, send it to one founder or investor who needs it.
Keep Reading: More Founder Playbooks Worth Your Time
▫️ Travis Kalanick: 10 Lessons on Stealth Building, Physical AI, and His New $70B Bet
▫️ Alex Karp: 12 Lessons From Palantir’s $500B Run
▫️ How Tech’s Quietest Super-Angel Built a Multi-Billion One-Person Venture Platform
Keep Reading: Build the Operating System Behind Your Own Company
▫️ The Board Meeting Operating System
▫️ The Free Founder Operating System
▫️ Inside OpenAI: The Operating Model That Makes 2 Engineers Beat 200
Keep Reading: Put AI Agents to Work This Week
▫️ How Jensen Huang Turned a Green Van and 30 Years of Rejection Into Nvidia
▫️ Andrej Karpathy: The AI Workflow Shift Explained
▫️ The one-person company stopped being a meme. This is the operating system that runs it