The Series A used to be $8 million to $10 million. Rudina Seseri says that is what a seed round raises now, and that plenty of first rounds come in at $30 million or $50 million.
That change has broken the job she does. Glasswing Ventures likes to write the first check and then help build the company, which is a business model that does not survive a $30 million opening round.
"How do you become the first check-in, roll your sleeves, help build the company when their first raise is 30 million? The economics don't work."
Seseri founded Glasswing and has spent the year splitting her portfolio in two — pre-seed companies valued under $15 million on one side, minority checks into $100 million-plus physical AI rounds on the other — while building an internal AI system that she says compresses two weeks of due diligence into half a day.
The full interview is covered here so you can skip it. 38 minutes of audio, 17 minutes of reading.
Here are the 16 insights that matter.
👤 Guest: Rudina Seseri, Founder and Managing Partner of Glasswing Ventures, an early-stage venture firm named one of America's top venture capital firms of 2026 by Time
🎙️ Host: Rana el Kaliouby, who ran Affectiva for 10 years, now invests herself, and is a limited partner in Seseri's fund
📰 Published: 16 September 2026 on YouTube (Pioneers of AI), cross-published on the Masters of Scale feed
🔴 YouTube | ⏱️ 38 min | ✅ Time saved: 21 min
Key Takeaways
A seed round now raises what a Series A used to, which leaves an unfunded gap in the middle
The old A was $8M to $10M; some companies now raise $30M or $50M as their first round
Glasswing's own AI does about two weeks of diligence work in half a day
Eight full-time AI engineers built it, and a supervisory agent coordinates the rest
She passed on Perplexity repeatedly because it looked like a thin wrapper with no moat
The easiest money in vertical AI is in workflows nobody has written down
30% of filled prescriptions are never collected, and the process to reshelve them is its own cost
She will not invest in AI governance yet because she cannot tell whether the labs absorb it
Her view of what survives commoditization is judgment, not models
1. Coachability Is the Filter
The interview opens on what Seseri screens for before writing a check, and it is not the product.
"We back a lot of founders. One of the questions we ask is are they open to advice?"
The reason it matters more now is that money has stopped being scarce. Seseri said "capital feels like the commodity of commodities", which turns a first meeting with a young founder into a one-term negotiation about valuation.
What she says she is selling instead is the relationship. "How do you build that trust up front so that they know this is not just a transaction?"
Founders who have raised before understand this and first-timers often do not. Her framing is that the choice of backer is about the person rather than the valuation or the brand of the firm, and about what that person does when things go wrong.
2. One Call, One Decision
El Kaliouby opened with a story about herself: in the fall of 2020, ten years into running Affectiva, she texted Seseri from a walk and asked whether to raise another round or look for an exit.
Seseri remembered the call and where she was sitting. She had turned a bedroom into an office during the pandemic.
The advice was one sentence. "I said at some point make a call. You can keep going and going, but you've got to take in so many words, make the tough decision and get it done."
What made it land, on el Kaliouby's account, was that it was not hedged. She said advice is usually wishy-washy and this was not.
Seseri's reasoning was that the work was already done. "But you hit the success. Time to call it." She added that she knows how hard letting go is, having started a firm herself.
3. Building Is Not the Moat
Asked what it is like to invest in the current market, Seseri reached for 2000 as the nearest comparison and said even that was less pronounced.
"It's an incredible environment." And: "So with AI, building is no longer the challenge."
Her evidence is her own weekend. She said she sat down two weeks earlier and built herself an agent to collect Boston audition listings in a format she wanted.
If anyone can build, the advantage moves. "the advantage is more around the data" and the access, plus the totality of building systems — models on their own, she said, can be commoditized, because you swap one out and put the next in.
What she is left with is a pairing, not a single asset. "it's a data and architecture combo" — and then how deep you can go in a given space.
The infrastructure layer is where the money is going, particularly around graphics processors and what runs on top of them, and those companies are raising very large rounds.
The problem for an investor is the evidence base. "You have to decide am I in for the journey or not with very little data" — and with rounds of $50 million and $100 million, the usual 18-month verdict takes much longer to arrive. "It's a very different model."
4. A Barbell Portfolio
Seseri described a portfolio that has split into two halves that barely resemble each other.
One half is the traditional business. "So the valuations are under 15 million. We come in early."
The other half is minority positions in very large rounds — horizontal AI and physical AI infrastructure at valuations she put in the hundreds of millions to a billion.
The rationale is access rather than ownership. "So in the physical AI space, we've invested in unconventional AI and recursive AI" — she said the point is a seat at the table, for follow-on capital with her limited partners and for the ecosystem around the deal.
She is candid that the small check into a big round is a judgment call. "there's a lot of upside" is the reason she gave for doing them at all.
5. Seed Is the Old Series A
The structural claim underneath the barbell is about how the whole early-stage market has shifted up.
"I'm finding that I'm pretty convinced that the seed round now represents the old A."
The numbers she attaches to it. "The old A was 8 to 10 million. Many many of the seed rounds are in that range." Others skip that step entirely, "raising 30 and 50 right off the bat".
What has opened up is a gap nobody wants to fill. The earliest stage has not moved up to where seed now sits, "no founder wants to take capital in between".
Her rule for each half is different. On a larger check the company has to be "de-risked enough if you're writing a bigger check"; on a small one she expects to do what she called sweat, blood and tears alongside the founder.
And you have to do more of the small ones "because the failure rates will be higher".
What she says nobody has solved is how to be a first-check investor in this market. "How do you become the first check-in, roll your sleeves, help build the company when their first raise is 30 million? The economics don't work." Her working idea is to go earlier still, into incubation and formation.
6. Naivety as an Advantage
El Kaliouby raised Tech Trek, a program that sent students from MIT, Harvard and Princeton to the West Coast for a summer to build startups and pitch on their return, and asked whether undergraduates are backable.
Seseri had met one that day — a junior dropping out of a full merit scholarship to build a platform for analytical financial models.
Her first reaction is personal and she said so. She grew up in Albania, where private wealth was taken and education was the one thing nobody could confiscate, and she is wired to value it.
Her professional view splits in two. Some people, at any age, will figure it out without having the answers.
The second half is a regional argument with a track record. Boston venture firms used to back professors while "the west coast was backing students", and in mobile the transformation was "driven by the students rather than the professors".
"So, I always sort of go back to the Uber example." Her version: a student in Paris in the rain, unable to understand why he cannot summon a car from his phone, who then dealt with the barriers as they came.
"So, there is something about that naivity of I don't know too much to know that it's done this way."
The counterweight is distribution. She said the risk is whether young founders can get adopters to embrace them, and that in a vertical market domain expertise, trust and an understanding of workflows may be one of the few defenses left to an AI-native company.
7. 30% Never Get Picked Up
On where vertical AI actually creates value, Seseri gave what she called a one-two punch, starting with the easy half.
The first opportunity is productivity from workflows nobody has mapped. Her example is pharmacy fulfillment: "There are five different systems that have been stitched together", with a person touching the prescription at every step and retyping it between systems.
The hard part is not the model, it is the knowledge. "A lot of the challenges that we have in different industries is not that it can't be done is that the workflow is in the mind of somebody" — or in a binder nobody has opened, from which there have since been hundreds of deviations.
Access to genuinely vertical data is the other requirement, and together, she said, they make the return on investment obvious.
She named a portfolio company doing exactly that, and used it to introduce a statistic that reframes the whole category: "30% of prescriptions do not get picked up".
Her point is what that implies about the rest of the process. "So you've gone to the doctor you've bothered they've given you a diagnosis they've given you at a minimum a treatment if not a cure and you don't go to pick it up." Reshelving all of it compliantly is its own workload.
8. Deep vs Buy the Layer
El Kaliouby put the concern that has grown sharper over the last six months: the frontier labs are adding capabilities fast enough to make vertical companies obsolete.
Seseri framed it as two competing positions. "I think there are two schools of thought competing right now. One is you go deep and specialize."
The other is the platform pitch from the model and infrastructure companies — buy us as the layer, build your agents on top, and skip the specialized tools.
"I think the reality will be hybrid as is always the case in this world."
Her doubt about the do-it-yourself version is operational. "If everybody's coding internally and building their own agents, I have a lot of questions around the long-term sustainability of those agents, the quality, performance, compliance."
The defensible position, on her account, is not knowing more than the model. It is depth, trust and governance — the things a foundation model that has seen everything still does not have.
9. Judgment Stays Human
El Kaliouby raised the legal industry — "Companies like Harvey AI, right, have developed AI solutions for law firms" — and asked whether the opportunity is in tooling incumbents or in building an AI-native firm from nothing.
Seseri refused to pick. "I can make both arguments with the same level of conviction" — because, she said, nobody truly knows.
Her lean is toward the clean slate. "I'm increasingly believing that the company of your future should really start with a clean slate."
The version of the law firm she describes inverts the roles. "So as you look forward and you look to the future, why can't I have my legal agent be there and the lawyer just provides the judgment?"
"I'm really focused on this judgment piece."
Her argument for why models do not have it is about what they were trained on. A model has been trained on all web data, the best digital proxy for humanity's knowledge; a human with that much information would be far smarter than any of us, and the models are not.
Her explanation is a distinction between two things. They are neural networks "dealing with the brain, not with the mind", while people reach decisions from limited and imperfect information.
"I think for the safety of our species, I think we need to retain the judgment."
10. Governance Is Existential
El Kaliouby asked about AI governance and safety as an investable category, and said she has not put money into it because she cannot tell whether the labs absorb the function.
Seseri's answer is that the categories are merging. "So we invest actively in cybersecurity" — and security, safety, governance and even physical security are converging under one umbrella.
"Every enterprise is trying to figure it out." The open question is whether buyers go to their existing security vendor's new governance module or to someone else.
Her own use case is the argument. "I get 45 to 63 mails per hour." She uses an agent to sort, prioritize and pre-draft, which means the agent reads all of it.
Which creates the compliance problem she then had to buy a product for. "So we have a company called D2 that's basically providing that MCP security and we couldn't be in compliance if we didn't have that capability applied"
Her conclusion about pricing is the strongest claim in the section. Demand for governance is close to price-insensitive because "you must have it this is an existential question".
11. A Pipe That Shrinks
Asked about her investment in Liquid AI, a frontier lab based in Boston and founded out of MIT, Seseri explained the technical argument in plain terms.
The company computes at the edge rather than in a central data center, which changes the efficiency arithmetic.
The second feature is the one she spent longer on. In a conventional system "the pipe is hungry the same" whether the full compute is needed or not.
"So the lay person's analogy that I give is it's a concrete pipe and it doesn't change" — data and compute flow through it at the same width regardless of the job.
"So it shrinks if it doesn't need all that compute and it expands as it needs it." She said the advantage is not only performance but what it costs, in power and in emissions, to deliver that performance.
Her broader thesis is that the current approach runs out. "So I think there is a wave of opportunity with these big models" — at some point improvements to large language models stop, and the next step "will be a paradigm shift" that she thinks the frontier labs are positioned for.
She also named a world-models investment. "We've made an investment in a company called Odyssey which is a world models company", alongside physical AI and AI-native interfaces.
12. Boston's Own Density
On the argument that AI is happening in the Bay Area, Seseri conceded the premise before making her case.
"So I think we need to acknowledge and then we need to own our strength." San Francisco's ecosystem is better developed, she said, simply because the big technology companies are there.
What Boston has is people. "we have the highest quality students" and professors, and the density that comes with them.
Her instruction to the region is to stop comparing. "Let's just not sort of worry about what we are not. I want to speak to what we are."
The strategic problem is retention, not creation. The task is to capture and support teams "before they move to the west coast".
She thinks the talent pool is less competed for. "MIT has not been as picked over as some of the west coast schools", and she credits renewed institutional energy around entrepreneurship at MIT.
A funding change has made industry partnership more valuable. "Also, research funding has dried up. So, partnering with industry becomes a lot more important than it used to."
13. Not Rich and Ignorant
El Kaliouby asked what Seseri would say if her own daughter announced she was leaving to build a company.
"First of all, I would support it. But it cannot be at the expense of an education." Before, after or alongside — and in whatever form education takes in six years.
"But what I do not want my daughter to be is rich and ignorant."
What she wants instead. "I'd rather she be happy, well adjusted," and multidisciplinary: "I want her to be a full human in her fullest capacity."
Her worry about the alternative is about reinforcement. Narrow and algorithmically fed, she said, "we confuse knowledge with information" and lose the worldly view that holds people together.
14. Loving All Her Companies
Asked for her biggest mistake as an investor, Seseri named a temperament rather than a deal.
"I love all my companies and anything you know about investing tells you that you need to focus on the winners and getting sort of the middle performers to move up."
It cuts against how she is built. "So, that is a constant struggle. I've gotten better at it" — she said she wants to save companies rather than triage them.
The discipline argument she makes to herself is about whose money it is. "You got to put the effort where you can generate the biggest returns because I want to support my founders."
"But we're also managing money for endowments for pension funds." Her objection to the caricature of that: "It's also the scholarships that need to be funded for families that can't afford them out of these endowments", plus teachers' and firefighters' pension funds.
Which is why she treats founder-friendliness as something that can be overdone. It is "very easy to say I'm founder friendly" and lose the other perspective.
15. The One That Got Away
The miss she names is Perplexity, and she said the answer depends on the day.
Glasswing saw it repeatedly — "many many times and passed on it" — and she credited her partner Clay with seeing it as often.
The reason was that it looked like a thin wrapper and the firm wanted a moat.
El Kaliouby's reading, which Seseri accepted, is that some days the call still looks right and some days it does not.
What the miss taught her is about distribution rather than product. In mobile, she said, "your app was so much better than mine" was rarely the difference — virality was, and once a product reached consumers a second or third entrant could not replicate it.
16. 105,000 Founders Ranked
Glasswing has built AI into its own operations, and this is where Seseri is most specific.
"we have gosh like eight AI engineers" working full time.
"So, over the last few years, we've actually built out what I call the brain due diligence platform." It runs several agents that interact, coordinated by a supervisory agent trained on large volumes of data.
What it buys is time without a quality trade. It maintains "the depth of diligence but moves much faster".
"I would estimate about two two and a half weeks worth of work in half a day."
The second effect is that the firm can ramp on a new thesis quickly and then line up human experts and customers to check the conclusion.
The sourcing system is separate and larger. It pulls from research archives and other sources to predict who is starting a company — "about 104,000 founders" last week, over 105,000 this week, each with a rating.
The purpose is to reach them first. The aim is to "discover them before they tell the world", which she called a work in progress.
Bonus Insights
The one founder characteristic the data has surfaced is an old Union Square Ventures observation about "founders that have known each other since childhood". Seseri was careful about how much weight to put on it: "I mean that's one data point. But I think it's probably a much more complex formula."
Asked what a venture capitalist's job becomes, her answer was one word. "Judgment at the end." Research, diligence and analysis get automated; "It's about how we weigh the analysis. Again I think analysis will be automated. It's about the judgment and the patterns that may not be captured in the numbers that are hard to quantify."
On what excites her, she went past the industry. "I mean I think this is bigger than the industrial revolution." The questions she raised were social rather than technical: "What will the social norms be in the future? Will we need to work in the same manner?"
El Kaliouby's own closing summary named coachability as the thing that stayed with her. "I thought it was really cool how she looks for founders that are coachable", particularly when the investment is very early and the investor expects to be a working partner.
Seseri's bottom line is that in a market where anyone can build and capital is abundant, the things that still compound are proprietary vertical data, an understanding of workflows nobody has written down, and human judgment — and that the venture business itself is being reshaped by the same forces, with seed rounds the size of old Series As and diligence work collapsing from weeks into hours.
Products, Companies & Tools Mentioned
Glasswing Ventures (Seseri's firm: early-stage AI investing, eight in-house AI engineers, and a diligence platform that ranks more than 105,000 potential founders)
Perplexity (The pass she regrets on some days — seen many times and turned down as a thin wrapper without a moat)
Liquid AI (The Boston frontier lab she backed, which computes at the edge and scales its compute up and down instead of running a fixed pipe)
Odyssey (Her world-models investment)
Harvey (Raised by the host as the example of AI sold into existing law firms, against the idea of an AI-native firm)
Affectiva (The company el Kaliouby ran for 10 years, and the subject of the phone call that opens the episode)
Microsoft, OpenAI and Anthropic (Named as the platform layer arguing that customers can build agents on top of them and skip specialized tools)
Uber (Her standing example of founders too inexperienced to know a thing could not be done)
Union Square Ventures (Source of the observation that the strongest founding teams have known each other since childhood)
Watch the full episode:
If this was worth your time, send it to someone closer to the industry than you are.
Get the latest market chatter as it happens: