Jesse Middleton

The through-line

Almost everything Jesse Middleton says comes back to one conviction: groups of people organised around a shared mission produce outsized outcomes, and that dynamic can be measured rather than gestured at. He arrived at it as a self-described nerdy tech guy moderating IRC forums, then watched it scale commercially at WeWork, where he was on the founding team and where growth for the first five or six years was, in his account, entirely community driven [1][3]. What has shifted is the object of his attention. The community language now sits underneath a venture practice at Flybridge that writes the first $500,000 to $3 million into pre-seed and seed companies, and his day-to-day preoccupation is generative AI, which he insists is a genuine technological break rather than a cycle to be waited out [3][4]. The constant is a preference for the earliest, least legible moment: "I have zero interest in being a late stage or growth investor if I'm being honest" [3].

On community as a measurable growth engine

He is blunt that community at WeWork was not a slogan but a funnel, and he walks through the arithmetic. The company hosted events its members actually needed, such as bringing in designers to talk about designing software; every member invited a friend, not because anyone asked them to, and roughly half of those guests had a company of their own [1]. About half returned within a month for a tour, tours converted to paying customers 35 percent of the time, and each convert typically brought four or five colleagues because space sold by the desk at $500 to $750 a month, producing a lifetime value of at least $10,000 to $12,000 against a customer acquisition cost that was effectively zero, since the events were being hosted for existing members anyway [1]. He also names the point at which it stopped working: "when we stopped doing the things that didn't scale was about five or six years in", once the company was opening more buildings than the tour pipeline could fill, at which point sales teams, marketing plans and processes arrived and conversion fell [1]. He treats that as normal rather than tragic. "I can't imagine a community of 250 000 people all feeling as close as we did when they were 47" [1].

On what actually makes a community, and what it is not

Two things distinguish real communities in his telling. The first is authenticity concentrated in the early members: relationships must be tied to a common mission, and "it's not something where you can just throw a bunch of people in a room and hope they meet each other" [1]. A company can be built around a mission; a movement takes far more care [1]. From there "communities have this incredible flywheel where every member brings you know 10 other people into the fold", and the bonds outlast the institution, which is why he still meets New York Tech Meetup alumni around the world fifteen years on, and why he compares it to the shared culture people carry out of a company like Microsoft [1]. He draws the same lesson as an investor about earliest customers and earliest investors on a cap table [1].

The second is that communities are breeding grounds for ideas that arrive with a built-in user base. Companies get founded out of Slacks, Discords and events where people are chatting about a shared problem, and he offers open source as the original case: "ultimately open source software is community led software development that's what it is", developers building a product together for free, then dragging it into their employers and creating enterprises around it [1]. He is careful not to oversell it, noting that most startups fail regardless, but a community start gives you an early advocacy group willing to put you on a pedestal [1]. His exposure to this predates WeWork, through a fraternity brother who started Indy Hall in Philadelphia, one of the first coworking spaces in the country [1].

Asked whether community still gives WeWork a competitive edge against the many rivals claiming the same, he says no [1]. What survives is the brand, which he considers the most pervasive millennial real estate brand in the world, plus lunch and learns, demo days and recruiting events organised in every building by members rather than staff [1]. On the valuation history, from a $47 billion private company to a public company then worth under $6 billion, he offers only that "everybody's entitled to be wrong at least once" [1].

On designing rooms where people talk

His format preference is consistent across two decades. At WeWork Labs in 2010 and 2011 the observation was that New York had demo days at Techstars and similar programmes, but those were about pitching investors; there was nowhere to describe what you were building and get feedback from other builders, so they brought in people from Facebook, Google and Twitter and told founders to "just talk about what you're building", accepting responses ranging from we're going to crush you to we should work together [3]. He applies the same logic to Lumos House, which he attended and hosted in Austin, Miami and New York, praising the curation of people and the creation of a reason to be somewhere, and describing the appeal as "bringing people together giving them the space and the freedom to create a conversation that is their own" [2]. There is no predetermined topic; "it's a little bit like an unconference style that just happens in an evening", with lightning talks and the expectation that attendees leave having met someone they would never otherwise have crossed paths with [2]. His New York edition, The Melting Pot, deliberately mixed entertainment, tech and business, and he takes the recurrence of the same faces across three cities as evidence a family is forming [2]. He notes the smaller after-hours version is often the best part, as at WeWork parties that police shut down for noise, leaving a smaller group upstairs [2].

On backing founders at the idea stage

Flybridge invests pre-seed and seed in technology-forward, AI-heavy companies, has been at it since its first fund in 2008, manages about a billion dollars, and made much of its success in developer platforms, including being earliest investors in Firebase, Crashlytics and MongoDB, where his partnership still sits on the board fifteen years later [3]. He avoids hardware [3]. The stage he likes is illustrated by a founder just out of Snowflake he had worked with for three years across different ideas, to whom he offered terms when "he's got a a memo and a Google doc and and a name of a co-founder", with no company name yet [3]. He also gives unvarnished live feedback to early founders, including teams building an AI-driven API platform and an AI copilot for product marketing managers [5].

The quality he screens hardest for he calls the Pied Piper effect, after the fairy tale rather than the television show. A founder must attract three constituencies early, customers, potential hires and investors, and "you have to get them to follow you irrationally", before there is any proof: an enterprise client signing on an app alone, employees taking equity promises, investors taking the next meeting [3]. Alongside that sit grit, perseverance, intelligence and category knowledge, tested through heavy reference checking and a look for people who have lived a certain journey and carry a chip on their shoulder [3]. The reason he needs irrational followership rather than evidence is structural: "if I could tell the story of how successful you are already as a seed investor you're too late for me" [3].

On how to pitch him, and what kills it

Too many slides is his first complaint, with "14 or 15" the maximum an early-stage deck should run; founders try to unload everything they have ever learned about their industry in the first email [3]. The corrective is compression, since one of the best qualities he sees is that founders "are able to articulate a complicated thing their business in a very succinct way in a way that I could then retell it", and whether that arrives as a deck or a Notion page is immaterial [3]. His second complaint is exit talk. Venture math means "I invest in 40 or 50 companies and I invest early I need each one to be 50 to 100x post dilution", and there are easier ways to make a small IRR, including debt at high rates, so a founder describing a $50 million sale or promising to make him money signals they are not building anything of value [3]. What he wants is the long game, because "our best returning companies create massive value not just for shareholders but for their constituents" and the world around them [3]. He is also clear about the economics of his own job: his LPs enjoy working with him as long as he makes them money [4].

On why AI is not a fad

The most common LP question he fields is whether generative AI is a flash in the pan, and Flybridge built a public-market AI index partly to answer it [4]. His verdict: "it's not a fad like this AI Trend this movement every Global 2000 company is heavily investing in it", and more tellingly, they are being rewarded for it, with the index returning two or three times the cloud index or the S&P over the prior year and commanding higher enterprise-value-to-revenue multiples [4]. He contrasts this with web3, where he still believes blockchain will prove valuable in many applications but concedes there turned out not to be many great enterprise use cases yet, whereas with AI "when you can replace a lot of the menial human task labor things with with the technology that's just as good or better for on Tenth the cost it's a real thing" [4].

Inclusion in the index requires launching generative products or services placed at the core of the business, investing significant dollars, actually earning revenue from those advances rather than spending ahead of product, and being active in acquiring or partnering, as with Apple's OpenAI announcement, which had kept Apple out of the published version until then [4]. The index is updated monthly and carries a list of potential future additions such as Samsung, including private companies that have said they plan to go public, and he is candid that he hopes his own portfolio companies eventually appear on it [4]. The underlying advice to founders is older than the index: "when you're building your company even when you're four people you should look to the public market equivalence", because most early founders have no idea who their public comparables are and will be judged against them eventually [4].

On what AI is actually doing inside companies

He answers the buzzword objection with specifics. Every portfolio company, B2B and B2C alike, is already implementing AI through open source projects or enterprise tools across customer service and financial analysis, and he points to digits as AI-powered bookkeeping and accounting [4]. Melody Arc, founded by people out of Amazon and Walmart and now selling to retailers including some of his own portfolio companies, resolves 80 to 90 percent of inbound support requests entirely through AI, and his point about the consequence matters as much as the number: none of those companies laid off support staff, they redirected humans to the larger, more challenging and angrier customers [4]. Another seed investment, an enterprise generative AI management platform for the Global 2000 with customers including Thompson Reuters and Ferrari, is live in the Ferrari app, telling drivers in real time whether a warning light is a genuine problem and what to do about it [4]. He uses tastic 60 to 70 times a day to write content [4]. And at splice, a platform for music producers he compares to GitHub for music production, the positioning is deliberately restrained: the product suggests sounds and samples to match a hummed idea, style or vibe, but "we're not generating music we stay far away from that we are the artist friend" [4]. His own usage is the summary argument: "I use it every day to make me a better investor team member partner board member and even husband" [4].

Takeaways

• Community-led growth at WeWork was a measurable funnel: member-invited guests converting on tours at 35 percent, each convert taking four or five desks at $500 to $750 a month, roughly $10,000 to $12,000 of LTV at near-zero CAC [1].
• The unscalable tactics stop working at a predictable point; at WeWork that was five or six years in, when building openings outran the tour pipeline and processes had to replace them [1].
• Focus disproportionately on the earliest members, customers and investors, because the flywheel of each person bringing ten more depends on those first relationships being authentic [1].
• Open source is the template for community-built enterprise value: developers build a product together for free, then bring it into their employers [1].
• Design gatherings without a predetermined topic and without pitching, so builders describe what they are making and get real feedback rather than performing for investors [2][3].
• The single hardest quality to measure and the most predictive is the Pied Piper effect: getting customers, hires and investors to follow you irrationally, before any proof exists [3].
• Cap the early-stage deck at 14 or 15 slides and be able to explain the business so succinctly that the investor can retell it; never lead with exit scenarios, because seed math requires 50 to 100x post dilution [3].
• Judge AI adoption by whether public companies are placing generative products at the core, spending real money and earning revenue from it, which is why the AI index has outperformed the cloud index and the S&P by two to three times [4].