Seed Founders Need 150 Qualified Investor Targets in 2026

For Jason Calacanis, the mistake many early founders make is treating investor praise as if it were investor signal. A family office, a strategic prospect, or a wealthy individual can say the product is novel, say they have “never seen something like this before,” ask to stay updated, and still be a poor fit for a year-zero company. Calacanis’s advice is to separate what investors say socially from what they have historically done with money.

His rule is behavioral: look at the investor’s past investments, stage, check size, and process. If a family office has written seven checks in five years, all into private-equity deals brought by friends, and each check was around $30 million, a founder looking for $125,000 or $250,000 at seed should not infer much from a friendly meeting. “How on earth are they going to put 125k or 250k or what you're looking for in your seed?” he asked. “They're not.”

That does not make the meeting useless. It may create a future relationship. But in Calacanis’s framework, the founder’s immediate job is to find investors whose revealed preferences match the company’s stage. He named programs and funds that explicitly work closer to formation: Y Combinator, Techstars, Antler, a16z, Pear’s summer program, Sequoia Arc, Founder University, and LAUNCH Accelerator. In his own Founder University, he said, the intent is to meet founders at “year zero,” before incorporation or customers are necessarily in place; the accelerator can follow when the company is incorporated and investable.

The operating math, as he presented it, is blunt: contact 150 seed funds, get 50 first meetings, convert roughly 15 to 20 into second meetings, and close two term sheets. Calacanis framed that as the current field of play, not as an inspirational target. If a founder is not running that kind of process, he said, “you're just not playing the game on the field.”

The memorable stories — the founder who takes one meeting and receives a term sheet, or who was not even raising and gets funded — are memorable because they are exceptions. Calacanis told founders to judge second meetings more seriously than first meetings. A first meeting can happen because someone was polite, curious, or loosely referred. A second meeting is more meaningful because nobody is obligated to take it.

Lon Harris connected the point to other fields: many people imagine a content creator makes two or three videos, one goes viral, and a career appears. In practice, he said, it is daily repetition, hundreds of attempts, and refinement before growth begins. Calacanis’s fundraising view is the same kind of grind, translated into investor pipeline discipline.

A later question from Peridot, a dating app “for serious daters,” sharpened the same issue from another angle: founders are often told to find “the right investor,” but most advice reduces to stage, sector, or warm introductions. What tells an investor that there is alignment beyond category fit?

Jason Calacanis’s answer again returned to process. Investors use different filters, and founders have to discover those filters. He described a friend at True Ventures who reviewed decades of investing and concluded that his best investments came through trusted people in his network. That investor, Calacanis said, does not take cold emails as the basis for meetings; he takes meetings when another venture capitalist or portfolio founder recommends a company. For that investor, the founder’s job is not to write a better cold email. It is to reach the network that the investor actually trusts.

By contrast, Calacanis described LAUNCH as application-driven. Founders apply through launch.co/apply. LAUNCH meets with roughly 10% of applicants each week, with application volume varying from around 200 to 500 companies in a week. He said the firm now meets about 25 companies weekly, down from 140, which had become too many. Across a year, he described the funnel as thousands of meetings and roughly 100 investments, culled from 10,000 to 20,000 applications.

The resulting point is not that one path is better. It is that founders must qualify investors the way a sales organization qualifies leads.

For Calacanis, this is where the founder’s research becomes the pitch. If a founder is building a fintech marketplace, and Calacanis has previously invested in Thumbtack, Robinhood, and Wealthfront, the founder should name the relevant pattern: “you did Thumbtack and you also did Robinhood; we’re doing a marketplace for financial advisors.” That is not generic “alignment.” It is a clear statement that the investor has already shown interest in similar mechanics.

He called the fundraising market “bespoke and underground,” even with tools designed to make it easier. LAUNCH has a product called Whisper Network for its founders, where they can search investors, upload materials, and ask for introductions. Even there, he said, the work remains tedious: hunting, pecking, and researching.

The conclusion was organizational as much as tactical. Fundraising should be treated as a full-time job. In a three-founder company, Calacanis said, one founder can focus on product, one on sales, and one on corporate development — which he described as “another fancy way of saying raising money.” The company has to collect investors and build relationships over time, not spray a short batch of emails and wait.

He offered a second-order version of the same funnel for firms that only take trusted referrals. If there are 10 venture firms a founder wants to reach, the best path may be to identify five founders connected to each firm, send 50 emails to those founders, meet 20 of them, and use those conversations to get on the radar of three or four firms. It is still a funnel. It just routes through founders instead of partner inboxes.

One example made the point physically. Lon Harris read from a LinkedIn post about a founder who put a sign on his laptop in a San Francisco cafe: “AI B2B startup raising pre-seed. Product is live. Come say hi.” The post claimed a few investors approached him without cold emails, pitch decks, or introductions, because he had made himself visible “where money already hangs out.” Calacanis’s reaction was simple: “Love it.”

He tied that to Bay Area startup culture more broadly. A clip from Paul Graham had recently circulated, he said, about San Francisco and the Bay Area having a distinctive culture of helping founders without expectation of immediate return. Calacanis described his own podcast as a version of that ethic: people he has never met tell him it inspired them to start companies or reach people they heard on it. In his view, that help-without-short-term-return norm is part of the tech industry at its best, and it is something founders can experience by going where the culture is dense.

When asked how early-stage evaluation has changed compared with 10 years ago, Jason Calacanis’s first answer was the cost and time required to reach the market. The amount of capital needed to get a product live, and the time needed to reach some version of product-market fit, have both “dramatically decreased,” he said.

He placed the current shift in a longer sequence. In the Web 1.0 era of the 1990s, he said, it could take $3 million and 12 months to get a product ready for market. A startup had to carry what were then treated as non-negotiable expenses: a PR firm at $10,000 a month for 18 months; an office at $25,000 a month; a $250,000 letter of credit; servers costing around $500,000; legal expenses around $100,000; and administrative hires such as HR, accounting, and legal. By the time a company had a couple of product people and developers, it had spent millions before learning whether it would win.

Cloud computing, outsourced services, WeWork-style office flexibility, and later no-code tools collapsed that burden. Calacanis described a Y Combinator and cloud era in which three founders could work from home, eat ramen, and stand up a startup for $300,000 in under six months. Then even that compressed: maybe the startup needed $30,000 for servers, or maybe cloud credits made the server cost effectively free.

Now, with AI-assisted development and “vibe coding,” he said founders are showing up having built a product, reached $100,000 in revenue, and then applied to Y Combinator or LAUNCH. In that version of the market, founders increasingly build first and use early revenue as the basis for acceptance into an accelerator or seed process, rather than raising purely to discover whether a product can exist.

Lon Harris said the recent pipeline included companies that had built something 14 days earlier and already had hundreds of customers. Calacanis pointed to TWiST’s own bounties program as an example of the compressed cycle. The show had posted a $5,000 bounty for a real-time fact checker and received 15 submissions in under 30 days, with three he considered “really good.” Another bounty was for an annotated product. Calacanis said his hope was not only to award prizes, but potentially to hire one or more builders and incubate the resulting idea into a startup.

The evaluation implication, in Calacanis’s view, is that early proof has become cheaper to obtain. A founder who once needed a seed round to discover whether a customer cared can now sometimes test that question in days or weeks. Calacanis called a first paying customer “light product-market fit” — not proof of a durable business, since people can be convinced to pay for many things, but still an important signal.

The old “must-haves” he described — office, PR firm, full administrative stack, owned infrastructure — have become less central. The expectation he emphasized is speed, proof, and the ability to exploit cheaper tooling before asking investors to fund discovery that the founder could have done alone.

Lon Harris posed a common AI-era fear: if a startup is building in a vertical where OpenAI, Anthropic, or another frontier lab releases a product, how should the startup differentiate? How does a founder avoid being “death-by-Clauded”?

Jason Calacanis’s answer was that large-language-model interfaces will not absorb every feature set. There will always be functionality the general model interface will not add because it would create clutter or confusion. That leaves room for products that build workflow, collaboration, services, communities, and marketplaces around the base model’s output.

His example was travel. LAUNCH had seen a startup called Roam Around, which he described as one of the best previews he had seen, built around AI-generated travel planning. A user could ask for a three-day trip to Italy or plan around a 19-day vacation. The company shut down and returned money to investors, he said, because much of the basic itinerary generation became available inside ChatGPT.

But for Calacanis, the basic itinerary is not the whole product surface. A travel startup can add multiplayer planning: several people researching a Tokyo trip together, watching each other’s work, voting on breakfast, lunch, dinner, shopping, and bullet-train destinations, and converting that into a finalized agenda. It can add community. It can add support. It can add a virtual travel agent who uses AI to plan with the customer by the hour. It can connect the traveler to a local guide on the ground.

Those are the areas he does not expect ChatGPT itself to handle. The model can provide output, but it is unlikely to create the full marketplace of local guides, ground support, itinerary collaboration, translation, booking, queueing, and contextual service. Harris, drawing from his own trip planning, agreed that he already uses AI instead of search for some questions, but still uses Viator, Tripadvisor, and other platforms. In practice, he said, the result is layered tools, not one tool replacing all others.

Calacanis described the opportunity as finding “your spot in the stack.” A founder can assume that the base model will produce the generic answer for free. The startup must decide what happens before and after that answer. In travel, that could mean the product connects a user with someone who gets in line at a restaurant, translates, explains the backstory, arranges behind-the-scenes access, or takes the traveler to comparable but less influencer-saturated places.

The broader claim was not limited to travel. Calacanis mentioned TaxGPT as another example of a product that could layer around base model capability. His differentiators were consistent: community, multiplayer mode, support, services, and marketplaces. The frontier lab may be strong at generalized output; the startup can be strong at the workflow, trust, coordination, and human service layer.

A live viewer asked about the best path for a bootstrapped hardware founder whose prototype is real and nearing profitability: stay lean and grind revenue, or start bringing in strategic people before scaling?

Jason Calacanis’s practical answer came first: at that point, the founder should consider investors. His reason was that investor attitudes toward hardware have “totally 180’d.” The old line was “hardware is hard” and therefore unattractive. The new view, as he framed it, is that hardware can be one of the few remaining moats.

He gave examples of hardware or hardware-adjacent products with lock-in or differentiated utility: Plaud, Whoop, Eight Sleep, Terra Kaffe, Cafe X, and a bookmark product discussed by the team. He has invested in hardware companies himself, including Terra Kaffe and Cafe X, as well as companies that did not work out, such as a camera company he compared to Dropcam or Ring and a smoke detector startup before other smoke detectors emerged. The lesson was not that hardware has become easy. It is still hard and takes more capital. But it is no longer automatically disqualifying.

Hardware can create lock-in because the product lives in the customer’s physical routine. Whoop has a wearable relationship with the user. Eight Sleep is embedded in the bed. Hardware can also increase the fidelity and usefulness of the software experience, because the device can collect data, perform an action, or create a recurring interface that pure software does not own.

For financing, Calacanis said hardware founders need investors comfortable with the capital requirements, but they can also work backward from hardware companies and use preorder dynamics. Kickstarter remains viable in his view. So do viral ads aimed at the first thousand customers, especially if those customers are willing to pay three or four times the eventual retail price to be first. He put himself in that category: someone willing to overpay for early access, whether for a Tesla Roadster or various Kickstarter products that may later sit unopened in a closet.

Lon Harris extended the hardware discussion to robotics, suggesting that consumer appetite for “weird little new robot designs” could become its own category as humanoid robots become cheaper. Calacanis’s response was explicitly prospective: he thinks businesses will figure out what to do with commoditized robots, and he guessed they could ultimately be priced at “a dollar or two dollars an hour.”

He sketched a robotics-as-a-service model for warehouse work. In his hypothetical, a company making Optimus or Figure robots could approach Amazon and compare the fully loaded cost of a human sorting packages — wages, insurance, workers’ compensation, and overhead — to a robot offered at $4 an hour, maintained on site, and deployed in fleets. If a human fully loaded costs roughly $30 to $35 an hour, the robot subscription could be compelling even before the robot reaches very low hourly costs.

For consumers, he imagined a lease-like model: $299 a month for 500 hours of robot time, with overages charged by the hour, analogous to mileage overages on a car lease. Harris supplied the acronym “AaaS,” for “Automatons-as-a-Service,” and Calacanis adopted it. These were projections and pricing sketches, not claims about a market that already exists at those rates.

The same logic applied to security robots. Calacanis recalled Knightscope, a security robot company he associated with early TechCrunch 50 or LAUNCH events. Security guards may cost $20 an hour and fall asleep or play Candy Crush, he said; a robot can patrol continuously, record with 360-degree cameras, and work 24 hours a day. Even at the same price, he argued, it may do some parts of the job better because it never stops circulating.

His answer to the hardware founder ended with timing: hardware remains difficult, but the investor climate has improved. “You're in... it's good timing,” he said.

The practical problem behind “Founder Community College” is that many people now want startup skills without the full apparatus of a traditional degree. A question from an X user proposed a companion to Founder University for a coming wave of $500,000 to $5 million AI-driven, single-person, owner-operated companies: less fundraising, fewer board-seat and cap-table concerns, more focus on tools and building products.

Jason Calacanis said he had been thinking along similar lines, though his immediate exploration was around venture education. He had recently asked about accreditation laws and whether he could offer a degree. His conclusion: he could give a certificate, but not a degree, unless he partnered with an accredited university. The word “degree,” he said, is what creates legal trouble.

The practical foundation is LAUNCH’s Associate in Training program. Calacanis said the firm was hiring six people in June, beginning them as researchers, then moving them to analysts, then associates over roughly three years. He described a salary path in the range of $60,000, $70,000, $80,000, and $90,000. His pitch is that instead of paying for a venture-capital degree, the firm pays trainees to learn the work.

He also described the hiring design as intentionally overinclusive. LAUNCH hires two or three times as many people as it expects to retain, assuming some will opt out, some will not meet the firm’s bar, and roughly one-third will become long-term “all-stars.” He said some parents had asked whether they could pay him $60,000 a year for two years to teach their children venture capital, because no obvious course exists for that path.

That demand led him to compare the possibility with Kauffman Fellows. A screenshot of Kauffman Fellows’ “Our Model” page described a two-year education program for venture professionals built around self-reflection, peer learning, a structured curriculum, and a network. Harris reported that producer Jacob found the two-year tuition to be $80,000, with a $100 early application fee and a $300 later fee. Calacanis characterized Kauffman Fellows as diverse and valuable in some respects, but also “a bit of an elitist group” and “incredibly overpriced.”

His proposed alternative was apprenticeship-heavy. LAUNCH could create a practical program by charging perhaps $50,000 per person, enrolling 10 people, and using the $500,000 to hire two full-time staff. Participants could travel with him to Saudi Arabia and Tokyo, see programs in operation, interview companies, and work inside business units. He suggested a Monday morning professional training session and a Friday end-of-week session, with operational work in between.

Lon Harris added a media-training dimension. He argued that many strong founders and investors are poor podcast guests or communicators, and that a LAUNCH-style program could teach newsletter writing, social media, podcasting, and other public communication skills. Calacanis joked that because “some idiots started a podcast about startups 16 years ago,” every venture capitalist now has to have one.

The bigger structure Calacanis imagined combines founder and venture education: a company-building program with two tracks, one for Founder University and one for “Venture University,” where participants major in one and minor in the other. They would learn both sides of the table. He said he would be open to partnering with a university on a two-year Founder University degree, if a university wanted to do it, with students embedded in companies and trained to build.

The sharpest reputational aside came when Jason Calacanis was asked who in Silicon Valley he most dislikes. He named Mark Zuckerberg and Sam Altman, while framing both criticisms as objections to behavior rather than personal animus. On Zuckerberg, he objected to what he sees as a pattern of choosing company self-interest over broader human consequences, citing privacy, teenage girls’ well-being, partner treatment, copying, and platform optimization. On Altman, his criticism was relational: Calacanis said Altman “screwed Elon,” and Elon Musk is his friend.

For founders, the more transferable point was Calacanis’s insistence on qualifiers when facts are uncertain. He referenced a New Yorker profile of Altman, described it as brutal, and said press pieces may be partly true while also shaped by agendas. In another situation involving allegations against a founder backed by Y Combinator and General Catalyst, he said he tried to frame his comments carefully: if the allegations were true, the behavior might be unethical but not illegal; maybe investors did not know; maybe they were trying to fix it.

That caution came from experience. Calacanis said he had apologized after being too harsh about a military-tech figure, concluding that he had “kicked him when he was down.” His current practice, especially when discussing viral clips or breaking claims, is to say “allegedly,” “perhaps,” and “not confirmed.” Harris agreed that in 2026 people need to pause before sharing stories that satisfy their biases, asking whether something is AI-generated, community-noted, framed to provoke, or missing context.

For a founder raising money, that is not a media-theory aside. Investor trust is built partly through judgment under uncertainty: how a founder talks about competitors, facts, allegations, market claims, and other people’s behavior when the full record is not available.