Inside Bending Spoons: Finding Talent, Leveraging AI & Driving Operational Excellence | Luca Ferrari

Summary

• Bending Spoons is a buy-to-hold-forever operator, not private equity — Luca Ferrari runs ~$3B in run-rate revenue at 54–55% adjusted operating income margins with over $4M of revenue per core employee. The model: acquire proven consumer software brands, transform them “sometimes beyond recognition,” and integrate them onto a shared proprietary operating system of 50+ internal tools. “We have never sold a material business. We buy to hold and operate forever.”
• The Evernote case is the template: bought for ~$200M on a little under $100M of roughly breakeven revenue, headcount cut from ~350 to 50–60 within a year to a year and a half — and today it runs on about 20 people, “very very profitable,” with product velocity Ferrari calls “at least three times as fast.” He attributes the speed to the smaller team, not despite it: “Instagram was built by… like 12 people.”
• Hiring and firing are centralized company-wide and run as a quant science, because “the team is almost all that matters” once strategy is set. A dedicated talent team scores 100+ signals in some applications — like a quant fund stacking marginally predictive factors — and Bending Spoons received 800,000 applications last year while hiring fewer than 300. Hiring managers get no bonuses or KPIs: “There’s none. Just trust.”
• Ferrari favors raw talent over experience because “experience gets stale relatively fast” and “you can actually get worse through experience” — exposure to low standards or politics normalizes them. The insight dates to his failed startup Evertale (2010–13), where the best of 12 people delivered “easily 10 times” the median contributor’s output and was among the least experienced. “Experience, we can give it to you.”
• The cultural core is “extreme ownership” (name borrowed from Jocko, definition his own): past an admittedly high intelligence threshold, “it’s almost all how badly you want it.” His exhibit is Rafael Nadal — “not even probably a top 50 talent in his generation” per experts, yet one of the three best ever. Paired with “relentless simplification”: complexity compounds nonlinearly, the burden of proof sits on whoever adds it, and job titles were deleted entirely about a decade ago.
• AI has produced “close to breakthroughs in productivity”: an in-house design tool (Diagram) that does design work “in maybe 1% of the time,” and Alt Spooner, a Slack agent with the same level of system access as the employee it mimics. Evernote’s GM used it to verify, root-cause, and propose a fix for a bug in roughly 3 minutes — human-to-human coordination “would have taken weeks maybe.”
• The M&A edge is operational, not analytical: “It’s not that we see things in businesses that nobody else saw” — being the best operator means you can mathematically outbid everyone and still earn high returns. Ferrari says he doesn’t believe they’ve been outbid in five years; offers are firm (raised “almost never by more than 5 or 10%”) and he tells sellers to shop them. Deals are getting fewer and bigger: Vimeo (~$400M revenue, 1,000+ people) was transformed by the same ~50–60-person task force that did Evernote at a quarter the scale.
• Capital allocation is deliberately patient: buybacks “could be an appealing way of creating shareholder value” in the very long run, Singleton-style, but acquisitions returns are “way too appealing” for now. Decision philosophy: “procrastination is awesome if it doesn’t come from laziness,” and numbers serve logic, never replace it — he raised salaries despite a pricing experiment showing only modest uplift, because compounding reputation effects “will never show up immediately.”

Deep dive

1. Aspiring to “the best company there ever was” — and banishing the founder cult

• Ferrari’s operating philosophy is deliberate polarization: pick one or very few pursuits and go all out, keep everything else at “minimum commitment,” because “all sorts of rewards emotional and material are at the extremes.” A 10-out-of-10 marriage is “worth 100 times more than having an adequate relationship with my wife” — same logic applied to the company.
• The cathedral framing: “we’re not building a nice church, we’re trying to build the greatest cathedral that anybody has ever built… it’s more fun, energizing, better people will want to work with you” — offered aspirationally, “I know we have a very slim chance.”
• Ironic given Senra’s podcast, Ferrari rejects founder mythology: “we try to eliminate the idea of founder from this company as much as possible… all that matters is your contribution, your trajectory.” The ambition is shared, “but it’s not necessarily a founder thing, it’s more of a Bending Spoons thing.”
• The episode opens with Ferrari’s read on AppLovin’s Adam Foroughi — “uniquely focused, ruthless, and I mean it in a positive way… 10 out of 10” — the founder Senra says Ferrari most resembles in fanaticism.

2. An independent path to entrepreneurship

• Ferrari built without podcasts, business books, or studying other entrepreneurs. Starting in Copenhagen, then Milan, “not exactly a center of entrepreneurial pursuit… we were absolute nobodies, so it’s not that we could pick up the phone and call Jeff Bezos.”
• The logic chain worth keeping: “If you emulate what most people are doing, you’re pretty much guaranteed to at best be mediocre plus.” First-principles experimentation costs more mistakes and time, but yields “a few insights, a few new ways that will set us apart” — and geographic isolation shielded them from “the mantras that the big startup hubs were preaching.”

3. Evertale’s failure taught the 10x talent lesson

• Evertale (2010–13) used AI to write diaries automatically — “actually pretty cool but never managed to make it scale. Commercial failure.” The crucial datapoint: on a team of ~12, the best person contributed “easily 10 times” the median — “literally worlds apart” — and was one of the least experienced.
• The generalization: most work “is not rocket science,” and in technology both customer expectations and methods evolve so fast that “whatever people learned back then… a lot of that experience basically you can throw it away.” Software from the early 2000s is “primitive and almost unacceptably bad” by today’s standards.
• The sharper claim — experience can be negative: “If you’re exposed to low standards… you’ll normalize those over time,” and years spent succeeding through politics won’t transfer to what Ferrari calls “a radical meritocracy.” Talent — “a good brain and a massive eagerness to excel” — never fails to be valuable, and “experience, we can give it to you.”
• Team quality dictates organizational design: mediocre talent makes process and procedures “the lesser evil”; a strong team wants “as few rules as possible.” His analogy: you coach the Dream Team with Jordan and Barkley differently from a team of modest talents — “you can probably win with both. It’s a lot easier to win with Jordan.”

4. Hiring as quant science: 100+ signals, and the customer-support tell

• Ferrari’s framework for evaluating people: inputs (controllable) plus boundary conditions (not) produce outputs, and outputs alone can give “a massively distorted picture” — luck, or an unseen contributor. Track-record length fixes this: “nobody would question that Warren Buffett is almost certainly insanely good at investing… you could have been lucky for decades, but that’s astronomically unlikely” — while a hedge fund’s 50% year “could easily be two lucky stock picks.”
• Since new graduates have no sample to judge, Bending Spoons built testing plus a signal science: in some applications, 100+ signals — GPA, email exchanges with recruiters, task performance — explicitly modeled on quant funds testing “hundreds or even thousands of signals,” each marginally predictive, collectively decisive.
• The best specimen of a non-obvious signal: interviewer impressions of whether a candidate seemed “open to criticism and reasonably pleasant” don’t predict on-the-job collaboration — smart candidates perform niceness. But how candidates treat the logistics/support staff scheduling their interviews “predicted poor behavior in a social context much better than how they interacted in an interview.”

5. Centralize hiring and firing — because hiring managers have all the wrong incentives

• Across 50+ acquired businesses, hiring and firing are centralized under “talent managers.” The reasoning: a team lead who feels the pain “will try to fill the position as quickly as possible” with someone adequate rather than someone who can be amazing, and will favor “very experienced candidates over green, immature but potentially much better contributors.”
• Asked what incentive structure replaces this, Ferrari’s answer is stark: “There’s none. Just trust.” Specific KPI-tied objectives “almost invariably lead to bad outcomes” — optimizing for box-checking over the company. “How frequently were there super mechanistic KPIs with our pay tied to it in projects where everybody was pushing in the same direction? I’ve never seen it.”
• Centralization also delivers scale advantages: the talent team sees the whole pipeline, learns statistically whether waiting yields someone better, and can reroute candidates — “maybe someone applied as a product manager but they could actually be amazing as a growth manager.”
• Senra’s connection: Brad Jacobs’s maxim “an empty seat is less damaging than a poor fit.” Ferrari agrees — “completely” — then goes further below.

6. There are no seats: “we all have the same job”

• Ferrari rejects the seat concept entirely: “having sharp job descriptions is bad.” Work is a blob prioritized by ROI, and his line to new hires: “we all have the same job at Bending Spoons, all of us starting with me — helping the company succeed on a daily basis.” Titles like software engineer exist “just so we don’t step on each other’s toes too much.”
• The underlying claim: “very little work in a company, especially a digital business, is strictly necessary. Almost all of it is elective.” Of 100 things a company could do, maybe 10 are ROI-positive; “many companies are doing 40 things… in some cases tragically they’re not doing some of the ROI-positive things despite doing so many other things.”

7. Why previous owners left value on the table: perverse incentives and talent arbitrage

• Ferrari says many of the reasons those opportunities were not seized “frankly lay outside of their control.” Public-market incentives punish value creation — raise subscription prices, lose 30% of customers who each pay twice as much, and “often the markets will punish you dramatically” for the subscriber decline even as revenue rises. Inside Bending Spoons, where each business is “a source of cash for further deployment,” those unpopular decisions get made.
• A second cause is talent: matured businesses “have long stopped attracting some of the most hungry, ambitious talent” — executives work with “perfectly valid talent, but maybe not standout talent.” Senra’s realization mid-conversation: someone working on AOL inside Bending Spoons doesn’t think they work for AOL. “Exactly.”
• Senra’s parallel from Ramp’s Karim — on Brex’s acquisition: “best talent’s not going to go to Virginia and work for Capital One… if there was a war, it’s definitely over now.” Ferrari: “I don’t know their industry well, but it sounds possible.”

8. The McKinsey story: honesty, midnight shifts, and the consulting parallel

• The origin: three engineer friends from northeast Italy (Ferrari’s hometown under 1,000 people; parents cut hair) agreed whoever got the most lucrative offer would fund the other two building the prototype. Ferrari got McKinsey — and, “close to incapable of lying,” told the partner upfront he planned to quit for the startup as soon as possible. The partner enthusiastically hired him anyway.
• He worked on the startup “from midnight to 2 in the morning” — “not like when people say during the night they mean 7 to 9 p.m.” — spent vacations on it, raised ~$500K a year later, and left.
• The structural insight he took: McKinsey’s best graduates ended up serving telcos and banks “to which you would never have sent your CV” — and Bending Spoons replicates this for engineers, who’d never apply to Evernote or AOL for five years but are “incredibly excited, and rightfully so,” to spend 12–18 months rebuilding them. “You get the best of a startup and a big corp”: large user bases and resources, plus tiny teams making big changes.

9. Saturate capacity: the “immensity of the possible” is a trainable superpower

• The math: everyone should have “way more on their plate than feels even remotely comfortable,” because, assuming people select well, adding an 11th possible task to a list of 10 cannot lower the ROI of what they choose — “but it’s possible that it happens to be the highest ROI of all.” A manager’s number one coaching job is helping reports select well.
• The deeper purpose is training people to handle “the immensity of the possible” early: “you need to completely eradicate the concept of ‘I’m only good at my job if I exhaust my task list.’ There’s no such thing.” Better to test someone at day seven than discover at year ten, running a billion-dollar unit, that abundance overwhelms them.
• Constant rotation across businesses serves four goals: fresh eyes (“when you have looked at the same thing for a long time you stop having good ideas”), preventing subcultures (“we’re highly opinionated on what the optimal culture looks like — we want it uniform”), learning and motivation, and redeploying toward new acquisitions where expected returns run highest.

10. Extreme ownership: past the IQ threshold, “it’s almost all how badly you want it”

• The definition — name taken from Jocko’s book (“so immediately evocative… we need to use that”), concept redefined: “caring in your belly tremendously about being the best at what you do, about helping the team and the company succeed.” Ferrari has watched “people who were probably close to genius-level IQ fail here because ultimately they saw their job as a way to make ends meet.”
• The Nadal exhibit: “Most experts I’ve talked to believe he wasn’t even probably a top 50 talent in his generation, but went down as one of the three best for sure” — pure extreme ownership. Ferrari tells new joiners there’s “less than 1% chance you fail to have an amazing career” if they bring it, because intelligence was already tested.
• Senra’s Singleton-on-Apple story lands the same point: Singleton picked Apple among computer startups partly because Jobs “had to make it work — they had nothing else.” Ferrari’s read of Jobs: the eye for detail “maybe wasn’t even his main thing” — deep down it was extreme ownership. Senra adds Josh Kushner’s rule: smartest, most experienced, or wants it more — “you always pick the person that wants it more.” Ferrari: “Oh yeah. Completely.”
• The density requirement: extreme ownership is contagious and creates “an escalation of positive reinforcement” — but dilute it and the extreme owners “either leave or lose it.” “You can never have a high-performance team where more than a small fraction of people lack extreme ownership.”

11. Relentless simplification: burden of proof on whoever adds complexity

• The mechanism: humans add parts — features, processes, rules, headcount — and rarely remove them, and each new part adds interdependencies, so complexity compounds nonlinearly: going from three to four pieces makes the system “maybe 40% or 50% more complicated,” not 33%. Left unattended, any organization drifts toward “our modern society with all the bureaucracy.”
• Two operating rules: whoever proposes adding complexity carries the burden of proof — objectors “don’t need to prove it, they’re done” — and everyone must hunt existing complexity, fighting what he calls consistency/inertia bias: “the longer it’s been there, the more we should be questioning it.”
• Senra’s connection: Elon’s “go ultra hardcore on deletion” and Tobi Lütke’s line that “the world belongs to the fast,” with the Raptor engine’s evolution as “the modern day Picasso” of getting ahead by reduction. Ferrari’s caveat: even people who simplify “tend to be incremental in it. But often by far the biggest wins is complete removal” — kill the feature 2% use rather than run “a million migrations.”

12. The titles deletion: a 10x win from asking “why do we even have these?”

• The specimen case: a decade ago, drowning in defining who qualifies as “director” and managing the emotional drain of who got “senior,” someone asked why titles exist at all. Root cause: bragging rights and CV signaling — “probably true, but also not something the company needs to be involved with.” So they deleted them: pick whatever title you want for LinkedIn, “just don’t embarrass us.”
• Ten-plus years on: “hundreds if not thousands of person hours” saved, and “not a single instance of someone complaining” across thousands of people. The org chart is algorithmic — manage product managers and a tool automatically labels you “product management lead,” whether you lead two people or 200. Incremental simplification gets you 5%; “if you get rid of it completely, it’s liberating — a 10x improvement potentially.”

13. The operating system: 50+ proprietary tools acquisitions get “installed” onto

• Over a decade of investment produced what Ferrari calls an operating system: one system each for payments, A/B testing, lifetime-value prediction, recruiting, AI-model orchestration, credentials, data processing — 50+ proprietary tools. Buying a company “is almost like installing them on this operating system.”
• Improvements flow like internal open source: platform teams own tools, but each business fixes bugs and adds features that propagate portfolio-wide — so “adding businesses actually makes us better as a whole… another entry point for innovation.”
• Why not sell the tools? Competitive advantage, focus, and honest market assessment: most digital businesses “don’t want the most sophisticated A/B testing platform — overwhelming,” and the tools are “doubly powerful because they’re fully natively integrated with one another” — value that fades if you adopt only one.

14. Fewer, bigger deals: the same 50-person task force at 4x the scale

• The Evernote transformation (early 2023) took a task force of ~50 Spooners rewriting the codebase, re-architecting cloud infrastructure, rethinking monetization — on a business a little under $100M in revenue. In the first half of this year, “broadly speaking the same thing” was done to Vimeo — roughly $400M revenue, a team originally over 1,000 people — with the same 50–60 Spooners.
• The lesson: “the complexity of transforming a business doesn’t scale linearly with its revenue,” and the operating system keeps improving. At ~$3B run-rate revenue, “the business that moves the needle for us today needs to be a lot bigger than when we acquired Evernote” — so the trade-off between deal count (more OS entry points) and size resolves toward fewer and larger.

15. “Not another PE play”: the Evernote deal math

• Ferrari’s three differences from private equity: “we’re not a fund. We don’t buy to sell. We have never sold a material business”; interventions go product-deep (“I’ve never seen a private equity reinvent a product or rebuild the technological infrastructure”); and deep integration onto a shared platform, which PE can’t do because integrated businesses can’t be sold piecemeal. “We are almost as different as it gets — other than we acquire stuff for a living.”
• What he saw in Evernote: a quarter-billion lifetime users, several million active users at acquisition, a widely known brand with generally positive or neutral associations, and predictability — Ferrari estimated that paying customers had been on the platform for roughly 5–10 years, though he said he did not remember the exact figure. Subscription revenue made it more predictable, and value lay in existing users rather than volatile new acquisition. Price: ~$200M against a little under $90–100M of roughly breakeven revenue.
• Today: revenue “more than 100, less than 200” million, “very very profitable” (group adjusted operating income margin runs 54–55%, with tenured businesses higher), run by about 20 people — down from ~350 at acquisition via 50–60.
• Why 20 people can outperform 350: talent arbitrage (800,000 applications last year, fewer than 300 hires — “if you’re running Evernote, even if you’re Steve Jobs, 800,000 people are not applying”), portfolio risk capacity (a standalone CEO who cuts headcount and fails “is done — that’s the blemish on your CV you can never clear”), and the amortized technology platform no standalone could justify.

16. AI breakthroughs: Diagram and the alter-ego agent

• Diagram, built in-house, lets anyone — designer, PM, growth manager — pull up an app’s screens, iterate new interfaces that automatically follow the head designer’s guidelines and the codebase’s functional constraints, generate the code, and create a segment for A/B testing. For someone with design skills it’s “sometimes 1% of the time”; more interestingly, “it enables doing design work for people who before couldn’t.”
• The deeper diagnosis: “we humans are insanely inefficient at exchanging information… quite efficient at absorbing it” — the PM-to-designer loop of explaining, waiting three days, getting back the wrong thing collapses when the machine iterates instantly.
• Alt Spooner (“alter ego Spooner”) is a Slack agent with, by design, the same tool access as the individual it mimics. The live example: Evernote’s GM tagged it to check the support tool (Morus) for whether a bug was widespread, root-cause it in the codebase, propose a fix, and ping the lead engineer to review — “in maybe 3 minutes she essentially identified and fixed a bug” that human coordination would have taken “weeks maybe.” Ferrari himself pulled Meetup MAU cuts by geography and device in minutes instead of interrupting a data analyst for days.
• Ferrari’s clarification under Senra’s probing: headcount minimization is not the objective — “if more people create more value we would certainly deploy them.” But shrinking acquired bureaucracies back to startup mode is causal, not incidental: “despite is the wrong word — in many ways it’s because it’s a lot smaller.” Evernote’s post-acquisition feature velocity is “at least three times as fast under almost any frame of measurement.”

17. Singleton, not a conglomerate, and the long-run buyback option

• Influences are few by design — “at the risk of reinventing the wheel, also coming up with some real powerful innovation” — with one exception: Netflix’s culture deck, whose “beat complexity with talent, not process” ethos “rubbed off on us.”
• On the conglomerate label: he’s never used it. Berkshire and Teledyne are “relatively distinct and separate parts”; Bending Spoons makes everything “as homogeneous and integrated as possible.” Senra’s read on Singleton and Buffett is that they were exceptional at selecting companies and management teams but generally avoided businesses requiring extensive operational fixes — the inverse of Bending Spoons’ stated edge. Ferrari’s Singleton tribute: achievement divided by popularity, “he would be at the top of the rank… if you ask 100 people even in business, 95 will not know who he was.”
• On following Singleton into buybacks: “not imminently” — acquisition returns are “way too appealing” — but “in the very long run that could be an appealing way of creating shareholder value.”
• On the IPO’s timing, the decision philosophy: “procrastination is awesome if it doesn’t come from laziness,” because delayed decisions get made with more information, and sometimes boundary conditions shift and “you just don’t need to make it any longer.” They prepared from the first half of 2025 and kept the trigger decision open. The transcript later states that the definitive decision was made in “spring of 2016,” which conflicts with the surrounding 2025–26 timeline; it does not confirm that an IPO occurred in early July.

18. Strategy and negotiation: the best operator is mathematically the highest bidder

• The strategy in one sentence: achieve maximum operational excellence — structural integration eliminating redundancies, plus investment in talent and technology — so that “a business is better off with you than with almost anybody else.” Then “you’re almost guaranteed to be able to compound capital very efficiently through acquisitions, because by definition, mathematically, you should be the highest bidder” while both sides earn excellent returns. “99% of our resources” go to operating, “remarkably little” to deal-making; on Evernote he believes his bid was “way, way higher” than the next — “which by the way, in hindsight, we should have negotiated better.”
• Negotiation posture: fair, competitive first numbers, raised “almost never by more than 5 or 10%” — with one recent exception where a seller forced a range before data was available and the offer rose substantially once it was. The confidence claim: “I don’t believe to this day that we have ever been outbid… not in the last five years” — sellers who walked sometimes returned 6–9 months later to transact at the original price.
• The tell of a strong offer: weak bidders demand five-day exclusivity “because their best chance is to win on timing.” Bending Spoons says the opposite — “we encourage you to go and shop it around… once you convince yourself this is the best offer, we’ll sign faster.” Senra’s parallel: Buffett bidding Clayton Homes — $12.50 against every counter, closing with “if every capital market in the world closed tomorrow, you can still rely on this offer.” Ferrari, naming no names: “we have had one or two situations a little bit like that.”

19. Numbers are dangerous; logic and rationality never fail

• The setup is Senra’s favorite Ferrari anecdote: Ferrari is a stats geek who can recite jiu-jitsu and CrossFit records without knowing what the athletes look like. Senra says Tia-Clair Toomey probably won about eight CrossFit Games titles and missed one while pregnant; Ferrari responds that he has probably never seen her.
• Yet the business lesson inverts the expectation: “numbers can be very dangerous because they are an approximation of reality… they need to be handled with care,” whereas “logic and rationality properly defined are perfect — there’s never a decision where being logical and rational isn’t the optimal strategy.”
• The specimen: Bending Spoons generates over $4M revenue per Spooner and pays “some of the highest compensation in the markets where we operate” — “pay needs to be high enough that nobody forgets about it, but it’s not front and center.” A two-month experiment posting higher salaries showed only modest applicant uplift, but Ferrari raised pay anyway: the compounding reputation of being an extremely high-paying employer — word spreading at universities and on job boards — “will never show up immediately… that will take multiple years.” Decide on numbers alone and “you’re very likely to miss out on a lot of opportunity.”

Verification Notes

• The transcript gives an internally inconsistent IPO decision year: it discusses preparing during 2025–26 but later says the definitive decision was made in spring 2016; the transcript also does not establish that an IPO occurred in early July.