Summary

AI collapses the cost and latency of product discovery, but Fanton expects fewer total builders rather than unlimited headcount growth. Property Finder can put prototypes before users “in a day literally”; one agent page progressed from a Lovable design toward production in “a week or two.” Repetitive engineering should disappear while the best engineers move toward higher-order creative work.
The PM role survives because its real function is creative synthesis, not backlog administration or acting as connective tissue. Microsoft’s heuristic was that someone who could “play chess and play guitar” would make a great PM: analytical enough to decompose problems, creative enough to lift an engineering team. AI means smaller, more cross-functional teams—but not the death of that combination.
AI applications will need a defensible answer to “what are you actually doing?” The host frames Claude 4’s release as an external improvement to Lovable, without Lovable causing that model improvement. Fanton compares LLM platforms with iOS and cloud combined, calling OpenAI potentially “iOS plus Amazon Web Services.” The application-layer question becomes why users prefer Lovable, Cursor, or the next entrant—UX, learning, workflow, or something else.
Monzo’s product restraint is a trust strategy, although the host exposes the paternalism embedded in it. Fanton says the bank should sell only products it can back “a hundred percent,” including to a struggling mother with two children; meme coins fail that test. The host counters that allowing stocks while excluding crypto is “arrogant” when ordinary users may understand neither—an unresolved trade-off between customer protection and agency.
Portfolio expansion creates value only when every added experience improves retention and frequency across the whole. Rappi’s early super-app thesis disappointed because several unrelated experiences inside one interface lacked “the glue between them.” A taxi customer trying groceries should think, “Why didn’t I do this before?”—not merely observe that the company can offer both.
The strongest products combine practical utility with emotional reinforcement. Monzo’s escalating one-cent savings campaign reached £600 by year-end and vastly exceeded Fanton’s expectations because it joined a good habit with a push to overcome inertia. Product tone matters too: traditional-bank emails create anxiety because they feel designed “for them, not for me,” whereas supportive products show users how to get back on track.
High-performance operating systems should expose problems weekly without turning red metrics into instruments of fear. Fanton’s preferred WBR tracks the 10–20 things that matter, assigns each to the actual DRI, and asks what happened, why, and what the company will do. His disagreement with the host is explicit: exceptional people already apply more pressure to themselves than anyone else, so “hitting on the head” is a tool for average companies.
Product quality is partly a capital-allocation and culture outcome: scarcity sharpens invention, while comfort can erode urgency. Fanton recalls bootstrapped SkipTheDishes building the best partner-and-driver onboarding he had seen because “the need is what drives innovation”; excess money can introduce laziness. He estimates the energy at Rappi, iFood, or Mercado Libre at “two, three X” greater than in Europe, in his experience, while carefully conceding that this comparison is a generalization shaped by Latin America’s thinner safety net.

Deep dive

1. Daily compounding beats a constraining vision

Fanton played tennis six hours a day until 17, then quit because he would not become Federer. He now calls “if I’m not number one, I don’t wanna play” a mistake: the valuable discipline was finding “the wall” and practicing through it, not forecasting whether he would finish top 10 or top 100.
The host supplied the business analogue: his five-year vision was to become a Balderton associate, yet five years later he managed his own $200 million fund. Fanton’s alternative is Kaizen—make the company better every week and let improvement compound “53 times,” because the resulting destination will likely sit far beyond the original vision.
Losing his father at seven taught Fanton that only a small number of people are genuinely service-oriented: his aunts helped without money or transactional upside. It also produced his belief that “wisdom is much stronger on the pain than on the happy moments,” although he distinguishes useful adversity from childhood anxiety about food, rent, or security.
That hardship informs his founder model. Starting a company is like “choosing to have a kid, and you don’t know if it’s gonna live or die”; employees, friends, and investors have entrusted lives and capital to you. The irrational founder may be necessary at zero, but once substantial value exists, a steadier and somewhat more risk-averse operator can become the appropriate leader.

2. AI turns prototypes into testable artifacts

Fanton says the product-market-fit search has materially changed because teams that once failed even to assemble a meaningful test can now expose an idea to users “in a day literally.” That does not guarantee demand; it cheaply removes the inability to create something concrete enough for users to react to.
Property Finder’s agent-page redesign began as a Lovable prototype and moved toward implementation within “a week or two.” When engineers disliked its code, designers tried another tool whose output better matched the internal stack. The new selection criterion is therefore not merely visual fidelity but how closely generated code approaches a usable engineering artifact.
Some mock-up work is already being skipped, but Fanton found designers unusually enthusiastic. Mission-oriented designers care about making “a dent” in property and fintech, not preserving Figma as their formal output. The smaller group asking “Where is my role?” is questioning what happens when that artifact is no longer central.

3. Smaller teams still need the “chess and guitar” PM

Fanton remembers being well paid at Microsoft for copying code and unit testing while Bjarne Stroustrup operated at the level of languages, metalanguages, and runtime compilers—“like seeing Van Gogh painting.” AI should remove more mundane work and elevate creative engineering, leaving more exceptional engineers doing consequential work even if the profession’s absolute headcount falls.
Team composition should follow the problem: sometimes only engineers, sometimes data specialists, sometimes 10 engineers with one PM and one designer. When coding is not a requirement for everyone, hiring PMs can relieve that constraint and let them spend more time with customers and explore ideas while engineers handle implementation.
Microsoft’s PM heuristic was that anyone who could “play chess and play guitar” would likely excel: chess proxies analytical ability, instruments creativity. Fanton rejects the host’s suggested “connective fabric” definition and what he calls a European backlog-manager version; a PM should raise the creativity of the technical team while decomposing ideas into tractable steps.
In hiring, he first looks for “light in their eyes,” then tests abstraction and quantitative breakdown. AI might favor more extreme combinations—a “crazy creative” beside a mathematician—because tools could make their work easier to glue together. PMs still need to articulate a one-page account of what, why, and plan, even if ChatGPT now helps write it.

4. Great products are never done because somebody keeps caring

Fanton’s blunt diagnosis is that most products are “mainly crap.” Some teams lack talent or capital, but the common failure is that people stop caring and treat a product as something fixed once. That also defeats the idea that companies will replace SaaS with 170 internal products: anything valuable but non-core still requires continuous maintenance and improvement.
Nothing Phone is his proof that apparently mature categories remain open: the distinctive back, clean design, and his son’s enthusiasm signaled that “somebody cared about the phone users again.” Simplicity is not itself the goal—Whoop is complex “with a purpose,” drawing users into how their bodies work—while unnecessary complexity has no comparable defense.

5. Trust raises Monzo’s product bar—and creates a real dispute

Fanton calls Monzo “the most principled company I ever seen.” Rather than simply shipping and iterating, it emphasizes trust, emotion, and releasing only what it can back “a hundred percent.” That restraint carries unusual value in banking because balances, payments, and financially vulnerable customers make failure more consequential than ordinary feature disappointment.
His test is whether the bank could proudly cross-sell a product to every user, including “a mother with two kids” struggling to find work. He would not offer her meme coins; somebody else can provide those tools, while Monzo holds a deliberately higher standard for products offered under its trusted brand.
The host’s pushback—worth keeping—is that this makes Monzo an “arbiter of justice.” A mother might knowingly want the meme coin and profit, while ordinary customers may understand stocks no better than crypto. Fanton concedes that this is “one way of looking at it” but retains the view that somebody should maintain the higher bar.
Product trust varies by category. A bank fundamentally manages risk and must align its values with customers; Fanton sees Monzo’s win-win framing as the original innovation. Food delivery is simpler—food should arrive hot and on time, with refunds when it does not—though hygiene, fulfilment, and remediation still create lower-level trust requirements.

6. Mission narrows the roadmap; exposure to users resolves it

Property Finder’s mission is “to make living great for the region.” Fanton first asks whether an idea, if it works, materially advances that mission; ease, scalability, and financial sustainability follow. He prefers starting small and letting observed customer impact determine whether the company should scale and eventually monetize the idea.
He rejects dictatorship versus discussion as a false choice. Three hours—or an entire weekend—of disagreement costs almost nothing beside developers spending months building the wrong product. A decision still needs an owner, but its quality improves when at least two opposing views are fully heard and participants generate intelligence through the argument.
“Let users guide you” does not mean asking them to invent the product. The Henry Ford answer is to put the car in their hands as quickly as possible and observe whether it changes their lives. Fanton warns against abandoning imperfect first iterations, but calls a product a flop when the team delivers precisely what it intended and users still do not care.

7. New features count only when they improve the whole

Rappi’s super-app ambition produced memorable misses: the team assumed doing “this and that and that” inside one app would be inherently compelling, while customers were content using three separate apps. Shared UX and customer support were insufficient because the experiences lacked “the glue between them.”
Fanton’s revised rule is that every new category must improve retention and frequency for the platform as a whole. When a taxi product adds food or groceries, the new experience must exploit customer knowledge, drivers, loyalty, or another genuine advantage. “Nobody gives you a prize because you do cars and groceries at the same time.”
Monzo’s one-cent savings campaign showed the opposite dynamic. Users saved one cent on day one, two on day two, three on day three, reaching £600 by year-end. Fanton expected solid performance, not the “ridiculous” response produced by combining a visibly beneficial habit with an extra incentive he describes as a kind of lottery, helping users overcome inertia.
The larger insight was emotional: acting as the trainer who both prescribes the exercise and gets someone to do it can outperform numerical forecasts. Immediate feedback, including Monzo’s pop-ups, makes progress felt; Fanton uses haptic feedback as an analogy. By contrast, traditional-bank emails trigger anxiety because their tone suggests “they want me to do something for them, not for me.”

8. Weekly reviews turn red metrics into shared work

Fanton’s WBR model, learned through DoorDash’s influence on Rappi, combines transparency, data, and weekly rhythm. He dislikes OKRs and elaborate prioritization spreadsheets when they become management proxies; the review should directly expose the 10–20 things that determine whether the company will meet its objectives.
Each item belongs to a DRI—the person actually doing and understanding the work, not an executive three levels above. That person writes what is happening, why it is happening, and what the team will do. The format makes operating knowledge explicit while keeping ownership close to reality.
When an item turns red, leadership’s job is to help rather than “hit them in the head.” The DRI feels particular ownership, but Fanton compares failure to the All Blacks losing a World Cup: the team lost. Blaming one person while the broader system remains unchanged can consume years without producing genuine accountability.
The host presses the case for fear and pressure, citing harder-driving leadership cultures; Fanton answers that “truly super people don’t need fear” because they are already their harshest critics. Both Monzo and Revolut are, in his rough formulation, “0.111% on the success.” Standards and ambition should be high, but imposed fear is “more for average companies.”

9. Time to value can be immediate—or deliberately earned

Users do not obsess over a product as its builders do; they arrive with a purpose and offer very little time for a first impression. Fanton therefore prioritizes rapid value, then asks whether the company keeps investing after winning attention. The common error is not merely slow onboarding but failing to deepen the relationship after the initial payoff.
The host complicates that principle with Noom’s roughly 50 pre-sign-up questions: terrible immediate time to value, yet potentially a filter for intense customer need. Fanton agrees the funnel likely loses more people but retains highly engaged users; like a thorough doctor, the depth of questioning can itself demonstrate care and become part of the product.
Whoop makes users wait around seven days for a baseline, but Fanton thinks it manages the journey well enough that the wait feels worthwhile. It also delivers supportive recovery language—“Let’s get back on the horse”—instead of punishing someone who slept badly or stopped exercising.
Serving novices and experts requires progressive experience, not one interface flattened for both. Property Finder can give a first-time Dubai buyer a taste of the journey taken by someone managing 50 properties, just as Duolingo lets beginners visualize speaking five languages. Complexity should unfold as users enter and progress through the journey.

10. AI makes platforms stronger and application moats harder to explain

Scale creates risk aversion because a startup has nothing to lose while a $40 billion or $100 billion company has substantial value to protect. Fanton says leaders must define acceptable failure—banks cannot fail in the ordinary sense—while preserving the obligation to create a future that differs from the present.
The host identifies a new dependency: Claude 4 can improve Lovable’s output significantly through an external release. Fanton sees precedent in iOS elevating app experiences and cloud accelerating developers, but LLMs are more “full stack.” His bullish OpenAI framing is that it could combine B2C distribution with the roles of “iOS plus Amazon Web Services.”
Building atop that platform is rational because individual companies cannot reproduce the underlying model. The harder question is what remains at the application layer: “What are you actually doing?” Fanton tests possibilities—user experience, learning, perhaps fine-tuning, though that “doesn’t seem to be a thing anymore”—and asks why users will prefer Lovable over Cursor or numerous successors.
Banking similarly involves scale, regulatory expertise, process, and capital; Fanton doubts there will be 100 institutions at Monzo, Revolut, or Nubank’s level. He hopes Monzo can become a $100 billion company and sees no categorical reason it cannot, but explicitly leaves the call hedged: “Too many variables” and “time will tell.”

11. Ambition and scarcity shape product quality as much as process

Comparing regions cautiously, Fanton says the energy at Rappi, iFood, or Mercado Libre can feel “two, three X” greater than in Europe, in his experience. Latin American employees often wear the company identity proudly because elite opportunities and safety nets are scarcer. He stresses that Europe still has exceptional people and companies; this is a broad comparison, not a claim about every individual or national character.
SkipTheDishes is his capital-efficiency specimen. Its Manitoba location was inconvenient enough that others avoided the turboprop journey, yet Just Eat visited and found an unfunded team that had optimized every corner of partner and driver onboarding. “The need is what drives innovation”; abundant funding can create laziness because almost every constraint appears purchasable.
After only two months in Dubai, Fanton offers a strong caveat but remains “very bullish.” He describes it as a city or country behaving like a startup, treating residents as customers and even naming the licensing office a “happiness center.” Its differentiator is institutional optimism: people continually speak as if the future will be better than the past.
His sharpest missed-opportunity story is declining Mercado Libre roughly 14 years earlier despite recognizing “the best team I’ve ever seen.” He also says he bought into it for an uncertain amount—“$15,000,” “$50,000,” or “some number”—and held it; the investment rose 40X. He wishes he had invested more. The broader correction to milestone thinking: prestigious arrivals fade quickly; health, family, friendships, and growth through difficulty endure.

2025 Silicon Valley Tech Mid-Year Review with Fusion Fund’s 张璐 Liability for AI Harms: How Ancient Law Can Govern Frontier Technology Risk, with Prof Gabriel Weil