This week at TechCrunch’s StrictlyVC event in Athens — part of the Panathenea festival taking place in the city — I sat down with Niko Bonatsos of Verdict Capital, Andreas Stavropoulos of Threshold Ventures, and Ben Blume of Atomico to ask about the current state of venture investing, the wave of mega-IPOs that SpaceX is about to kick off, and where they still see an ocean of opportunity. Our conversation, following, has been edited for length and clarity. You can check out the full discussion at page bottom.

With SpaceX reportedly eyeing a $1.75 trillion valuation at IPO, and OpenAI and Anthropic potentially not far behind, what will the impacts be on the broader market?

Andreas Stavropoulos: I remember how exciting the Google IPO was, and how it ushered in a reopening of a market that had been very pessimistic about tech in the early 2000s — how it was an enabling event that brought in a whole new generation of entrepreneurs. The same thing is happening now. With every subsequent wave of paradigm shifts, the scale changes by orders of magnitude, and that’s to be expected. What business today in the information age is not a technology business?

Ben Blume: These are phenomenal companies, and with each one of these scale liquidity events, they generate wealth and returns that go back into the next generation of companies.

Niko Bonatsos: My co-founder at Verdict was the first-ever investor in what is now known as Cursor. So if Elon feels like he’s [having] a good moment, maybe Cursor [which Musk revealed recently that he has the option to acquire for $60 billion] will have some good news too. But more broadly, for the next next generation of companies, as Andreas mentioned, they could be going after much larger markets, and immigrant founders, as we know, they’re the ones who dream really big, they have nothing to lose, and they can go the distance, and Elon Musk is an immigrant founder himself. So, for those of us who come from Greece or other smaller markets, wow, you know, that’s a great example.

Some have suggested SpaceX at that valuation could soak up so much public market capital that it hurts companies going out in its wake. Is that a real concern?

Stavropoulos: You can choose to see most things as optimistic or pessimistic and make very good arguments for both. Something like a SpaceX, macro-wise, is going to end up bringing more people into the market than the short-term impact of soaking up some liquidity. Consumer involvement in markets in the last 30 years has gone from something that wasn’t really a thing to something people trade on their phones every day. Those numbers add up.

Blume: SpaceX is such a one-of-one company. For a long time, space has been a government and public sector domain. To give investors real financial access to it — I think that’s going to capture a widespread imagination. It may mentally draw from longer-tail allocations that might otherwise have gone into the next 20 or 30 software businesses, but I think the interest it generates more than compensates.

Is the current flood of capital into AI justified by future earnings, or is this a case of extreme FOMO?

Bonatsos: If you’re an AI-native founder or a company in the American dynamism space right now, you can live life in the fast lane. If you’re not in one of those two buckets, it’s really tough. In 17 years in Silicon Valley, I’ve never seen more groupthink. Three quarters of all venture capital raised over the last year went into five companies. Today, if you’re a 40-year-old tenured professor at Stanford not building something in AI, no one wants to meet you.

That said, something real is changing. Two founders with today’s AI tools can make more progress in two months with one round of funding than they could a year ago with ten people, two rounds, and a full year of work. This is changing how companies get started and how they’ll capitalize themselves — potentially going straight from pre-seed to Series B.

Stavropoulos: There will be a correction that pushes some capital back out of the market. The promise and the optimism is still significantly ahead of the short- to medium-term ability to show results. But on a long-term, macro scale, I don’t think we’re being over-optimistic. The problem is that shouldn’t be mistaken for thinking every 19-year-old with an idea is the next big thing.

How do you actually price deals when things are moving this fast?

Blume: The best founders have no shortage of capital options. You have to think about what’s a meaningful ownership stake for your fund, and walk away when you can’t get there. The interesting dynamic is that we’re a $500 million fund looking at the same opportunities as people investing from a $10 or $15 billion fund. The incremental value of a dollar to us versus them is very different. That distorts round sizes and makes it difficult for offers to stack up like-for-like.

Bonatsos: We do first-money investing — basically instead of friends and family, instead of angels. We invest in what I’d call “freaks” — individuals where, like in professional sports, a few people break all the records. One day goes by and they learn and mature and make the progress that takes the average smart founder a whole week. Most of the founders we’ve backed so far are working on markets that don’t have a name yet — which is exactly why the valuations are low. Larger asset managers can’t tell their teams to go find companies in a market that doesn’t exist yet.

There’s a lot of talk about very young founders getting term sheets almost on arrival. Is age really a proxy for anything meaningful right now?

Stavropoulos: At times of disruption, when the world seems to be changing in some fundamental way, it especially favors lack of experience. Experience can actually steer you the wrong way. That doesn’t mean it’s changed forever — we’re going through a phase where things haven’t settled down yet, and that creates fertile ground for new ideas, and typically younger entrepreneurs. But I don’t want to over-generalize.

Bonatsos: The exact same thing was happening when I arrived as a grad student at Stanford in 2009. The iPhone was two years old, the App Store was one year old, and there were days when there were more VCs on campus than students. Today is one of those singular moments again. If you’re 22 years old in San Francisco and building something in AI, there may be a seed term sheet in your inbox — but if you’re 19, oh my God, this means you’re really good [laughs]; you might already have a Series A [offer]. And look, age is all relative at this point — I was talking to a founder here in Athens this week who’s 24, and when I said he wasn’t that young, I meant it: I met the Mercor kids when they were 19, and look where they are now.

Blume: If you try to generalize just from age, I think you miss what you’re actually looking for: an extremely high level of intensity, the ability to move ahead of the pace the market is moving, and the mental dexterity to adapt in a landscape that’s changing constantly. If you have those things, it’s more important than the age on the passport.

What do you make of shady behavior happening around metrics — particularly how companies are reporting ARR [annualized recurring revenue]?

Blume: People are being relatively liberal with how they define the A and the R and the R. New pricing models — token-based billing, free tokens being counted as revenue — create a lot of ways to express these numbers. Our job as investors is to cut through that and make decisions based on the actual truths. Is it fine from a marketing perspective? Probably. Is it fine for deciding which companies get capital? No. But sophisticated investors can generally cut through it.

Bonatsos: Sometimes I’ll get an email with a very high ARR number from a portfolio company I didn’t remember doing that well, so I’ll contact the founder. The answer? It was 365 times what they made the day before because a campaign hit. I told him, can you please use a quarterly basis at least? Whenever a lot of money is chasing specific themes, some people develop a grifting mentality for short-term gain.

In venture you can only lose your money once on a bad investment, but the right one can return 100x — so you write off the bad actors and move on.

For the aspiring founders in the audience, where do you actually see white space right now?

Bonatsos: Every VC firm used to have at least half its partners doing consumer internet investing. Today, maybe they have half a person — they’ve left the field altogether. But one of the best AI companies of the last few years, OpenAI, became massive because of ChatGPT. Consumer is coming back, which is almost a crazy statement. Those founders today have maybe five investors they can pitch for their first or second round. I think there’s also a new movement emerging that’s going to help restore the American dream through new consumer fintech ideas.

Blume: The opportunity of AI interacting with the physical world is orders of magnitude larger than what we’ve seen so far in workflow automation and digital process. The physical world still shapes a large part of the economy. The bet on robotics in all its forms — not just the humanoid doing a backflip — is still one of the biggest wide-open spaces over the next 10 years.

If you’re interested in learning more about what the three think — including about whether Stanford University has grown too cozy with the venture capital industry — you can check out the full conversation below:

Gemini Spark is Google’s new 24/7 agentic assistant, designed to help you help you “navigate your digital life,” which essentially means getting your online to-dos done, summarizing the things you don’t have time to read (like the entirety of your inbox), or organizing something that would have otherwise involved too much screen time-filled manual labor, like a personal expenses spreadsheet.

The service was first introduced at Google’s annual developer conference in May, where CEO Sundar Pichai joked that Spark, which runs on virtual machines in the cloud, means that “yes, you can close your laptop.” The in-joke here is that he’s comparing Spark to other agentic AI systems, like the ever-popular OpenClaw, which require keeping the machine awake to run its tasks.

Spark, he’s suggesting, is agentic AI for the rest of us — those who would rather get things done without nerding out about it by setting up an always-on AI machine.

In practice, Spark is still very much designed for work-adjacent tasks, given its integration with Google’s productivity apps like Gmail, Calendar, Docs, Sheets, and Slides. (After all, how many times are you preparing a deck for in your personal life? Unless you’re a Gen Z creator explaining the latest meme to your chronically offline friends, that is?)

Google also struggles a bit to come up with real-world examples that would convince someone that Spark is a “must-have” rather than a “nice-to-have” tool for personal use.

Among its suggestions for “personal productivity” is using Spark to scan your emails and calendar for the day and send you a recap with your top three must-do tasks,” which already assumes you are a person who jots down your to-dos in a calendar or email app, instead of a notepad (virtual or otherwise), or just keeps a running list in your brain. (E.g., Grab prescriptions and shampoo at Walgreens. Buy more dog food. Hang out with friends on Saturday.)

Google also suggests you could use Spark as a weekend planner, by drafting a Google Doc “suggesting three free activities based on my open calendar blocks for the upcoming weekend,” which, again, assumes you are some sort of scheduling nerd in your offline life.

Nevertheless, with early access to Gemini Spark, I decided to put it through its paces, with what are perhaps some real-world suggestions of my own. I came away surprised that it was a fairly useful implementation of consumer AI, but not one that deserves to have its own brand.

Finding Savings

For one initial task, I asked Spark for help with a shopping-related research. The idea was to help me with an everyday local drugstore trip for household items, so I asked Spark for product suggestions based on weekly deals and coupons I could clip.

At first, Spark seemed to do pretty well here, as it told me exactly what products were on sale that matched my needs, and suggested coupons to clip in the Walgreens app for extra savings. It even suggested how I could stack coupons for one item by combining online promo codes, if I were placing an online pick-up order and was planning to spend more on personal care items.

However, as is often the case with AI, the devil was in the details, as one of the promo codes was invalid when I tried it, despite meeting what the AI said were the requirements. Still, Spark pointed me to some other savings — like buy-one-get-one-free and rewards deals that made up for this gaffe.

Planning a packing list for a day trip

In another test, I asked Gemini for help with a packing list for a day trip out of town. I asked it to check the weather, gather the event details, and make suggestions of what to bring with us, like sunscreen or water, to see what it would come up with, after it learned more about the activity. I asked for the final list to be imported into Google Keep.

Guess what Spark can’t do? Use Google Keep.

That’s a huge oversight, given that Google’s notetaking app would be essential for anything in the realm of personal productivity. Instead, it offered to make me a doc or draft me an email because, sure, that’s the sort of thing I’d want to check for my list of to-brings. (??)

In terms of the list itself, however, Spark was spot-on, suggesting lawn chairs or blankets, water, sunscreen, sunglasses, a light layer for when the sun goes down, a reusable shopping bag, and an umbrella for possible light showers that day. It also reminded me that dogs were not allowed, despite the event being outdoors. (Sorry, Princess!)

Summer Camp / Activity Suggestions

My child has aged out of summer camps for kids (and should probably just get a job), but before we went that route, I wanted to scour the local area to find out if there were any summer activities available for teens that she could do in addition to her engineering camp in June. I asked Spark to do a thorough search and find any and all suggestions, keeping in mind that we would not want to drive more than around 30 minutes.

Spark generated a decent list of ideas for activities that matched my child’s interests, and plotted out how far they were from home. Unfortunately, I forgot to prompt Spark to get the costs or dates of the programs, and it didn’t bother to tell me, which meant I still had to do more manual research on my own.

Recurring Task: Summarize newsletters from email

Like many, I subscribe to too many newsletters, so I put Spark to work on preparing me a weekly summary, which would arrive every Friday, focused only on the top five posts or articles I shouldn’t miss reading, along with a link.

The AI got to work, digging into my inbox and, within moments, had presented a summary of several interesting articles to read that included context and a link. (The link ended up being a Google.com redirect that didn’t work — I had to click the link displayed on the redirect page, as it never automatically sent me to the site in question.) While I generally liked the suggestions, Spark only returned four articles to read when I had requested five. Spark had interpreted the request as “4-5” for some reason.

Recurring Event: Suggest Weekend Activities

For another request, I asked Spark to compile a list of weekend activities around town for me on Fridays, so I can get to planning my weekend fun. As someone who lives in a smaller city, there aren’t always big events or things to do, so making sure you don’t miss the anticipated street festival or hot show when it comes to town is key. But there’s no single source to find everything there is to do — you have to read multiple local newsletters, visit websites and Facebook Groups, read the newspaper online, and more.

Spark instead set up a web search, combined (at my request) with a search of my Gmail for any relevant local newsletters, digests, or lists with keywords indicating a local activity suggestion. It then compiled a list of upcoming weekend events and noted that if I wanted to add any to my calendar, I could just reply.

If it wasn’t for Spark, I would have never known there is an Annual Beaver Queen Pageant nearby, which apparently features people in beaver costumes raising money for wetland conservation? OK, I might need to check that out. (You still have to tell Spark to add it, then click a button to confirm, but this is easier than the manual labor of reading through so many sources for ideas.)

Recurring Event: Check for Price Drops

For my last request, I set Gemini Spark to work on tracking price drops for an expensive eye cream. As a penny-pincher, I’d never buy it unless there was a crazy sale. I wanted Spark to keep track of the price changes for me and alert me if the eye cream ever became more affordable. However, Spark’s interpretation of this request was to simply recheck the price every two weeks to see if it dropped below my target. I’m not sure that would be frequent enough to spot a deal. (I’ll update if the results are successful, but I believe I’ve set too low a bar as my target — even after raising my bar by another $10! — so this is probably just wishful shopping at this point. But I’m always hopeful some online retailer will make a pricing mistake one day!)

More Ideas to Come

I can already see how I’ll be able to integrate Spark into my everyday life in other ways, too — I already have ideas for more email monitoring and cleanup tasks, for instance. The next time I change the home’s air filter, I’m going to ask Spark to remind me in three months to swap it out. If I ever get around to taking a vacation, I’ll probably have some tasks for it then, as well.

Room to improve

While Spark already performed fairly well on my tasks with only small quibbles, the biggest criticism I had was that there’s no need for this to be a standalone product with a different branding. I think that adds to consumer confusion in this day and age, where there are so many things happening in the AI space, and where every new model has its own name and number, and some of these are quite wild. (Nano Banana, anyone?)

Why not just pitch Spark as something Gemini can do out of the box, instead of making it its own product? Why does the toggle have to say “switch to Spark,” instead of just “switch to Tasks?” (If it even needs to have its own space in the user interface!) I personally don’t want to carry the mental load of trying to determine whether something is a question or a task; I just want to type in a question or request and be done with it.

I also think the lack of Keep integration is a major miss in terms of being helpful with your personal productivity. Google Docs is overkill for a packing list. And, unfortunately, for iPhone users, tapping into Gemini Spark directly from your device through a push of a hardware button or gesture won’t be possible — unless Apple announces this at next month’s WWDC? Instead, you’ll need to launch the Gemini app and use it from there. (Another issue with having Spark as its own toggle within Gemini — you can’t program the iPhone’s Activity Button to go directly to Spark, which is separate from Gemini’s chatbot interface. How great it would be if everything Gemini does were all in a single destination! Ugh!)

And while Spark will later be able to do more with MCP integrations, not being able to set it to perform certain tasks, like booking your favorite date night restaurant regularly through Resy or looking for flight deals on a preferred booking engine, for instance, makes Spark feel somewhat lacking for the time being, given that not everything you do online takes place in Google’s universe of services.

(Also, I’d really like to text Spark. I wish that were an option, too.)

Meta is developing an AI-powered pendant that it plans to start testing in the next year, according to a memo viewed by The Information.

This device would presumably build on the work of Limitless, an AI device startup that Meta acquired at the end of 2025. The startup made an AI pendant that users could attach to their shirt or wear as a necklace to record their conversations.  At the time, Meta said the acquisition would allow it to “accelerate our work to build AI-enabled wearables.”

Earlier AI wearables have failed to catch on with consumers — perhaps due to privacy concerns and tone-deaf marketing, or perhaps because they just weren’t that useful. But companies like OpenAI aren’t giving up.

The memo also reportedly states that the company is planning to expand its lineup of AI glasses and launch a business subscription called Wearables for Work. With all these planned devices, Meta is apparently hoping to reverse the fortunes of its hardware-focused Reality Labs division, which lost $4 billion in the first quarter of this year.

TechCrunch has reached out to Meta for comment.