The Spiro Circle

Days before Tal Shoham announced that his AI monetization startup Velocity had closed a $27 million Seed round, an independent tracker put a number on something the industry had only been speculating about: ChatGPT ads were showing up in roughly 26.5% of all replies globally, and 49% of replies in the United States. For Shoham, that timing is the proof of concept for his newest company and a sign of the latest change in the AI era. “ChatGPT, of course, has added ads now, which is amazing for us and for the industry, because it’s like a north star that everybody looks at,” Shoham told me.He co-founded Velocity alongside Amir Shaked and Nimrod Zuta, all three of whom are former senior executives at ironSource and Unity. The company is building what it calls a growth infrastructure layer for AI-native applications: an ad network, a mediation and auction system, and a “conversation intelligence” layer that turns chatbot dialogue into structured, privacy-safe intent signals. Basically, it’s helping bring adverts to your favourite AI agent. The round was led by NFX and Red Dot Capital Partners, with participation from Stardom Ventures, Corner Ventures, and Transcend, alongside a roster of gaming and ad-tech angels, including former ironSource co-founder Omer Kaplan.The pitch is a straight transplant of the problem his team spent a decade solving in mobile gaming, with one crucial difference. “Ninety-five percent of the users in gaming will never pay a dime,” he said. “You really want to try to find a way of monetizing those users.” In gaming, a free user costs almost nothing. But in AI, that math is inverted: “Every free user that you have on your AI platform is actually costing you a lot of money on inference, tokens, GPUs, and so on.”That inversion is the reason that AI companies have defaulted to hard limits (two or three free prompts a day) rather than the generous free tiers that built mobile gaming and social media into mass-market platforms. He bets that an advertising layer can fund broader free access without those companies bleeding cash, and that doing so improves retention rather than damaging it. “We have more than 12 design partners live already,” he said. “This doesn’t harm retention, it doesn’t harm engagement, it doesn’t harm conversion to monetization.”But the timing that makes Velocity’s raise look prescient also drops it into the middle of an unresolved trust problem - one that OpenAI itself has been actively renegotiating in real time. ChatGPT’s original ad policy excluded placement near politics, health, and mental health topics, with a standing ban on dating, alcohol, drugs, and gambling. But a June 2026 update already suggested that current advertising categories “may expand over time” to include medical, legal, and financial advice contexts eventually. In other words, the rules of the road are being written after the road has already opened to traffic. It’s a pattern that anyone who lived through Europe’s post-hoc arrival at GDPR will recognize as headache-inducing.I pushed Shoham directly on where that leaves the user. Chat conversations are not basic search queries: they’re often confessional, emotional, and far more revealing than anything a keyword ever captured. “There’s a lot more emotion behind what people are giving these algorithms,” I said. “It’s not just tapping into data points... It’s tapping into a real human feeling.”Shoham’s answer leaned on the compliance muscle memory his team built at ironSource, navigating GDPR and a patchwork of state and platform-level privacy rules for years. “We don’t take any of the private information from the user,” he said. “If you type in something on health, something sensitive, your social security number, or whatever it is, we’re not saving that, we’re not taking that, and we’re not integrating that into the model when we’re trying to find the right ad to show you. We have an abstraction layer that actually abstracts all the sensitive information.”So, whereas search reads your keywords, social media reads your behavior, AI just reads you. Velocity’s bet is that the same compliance discipline that got ironSource through GDPR can keep that power in check… but with ChatGPT's ad rollout already outrunning its own written rules, that's a promise the whole industry is now testing in public.Preview: The Next Google Ads? Inside Velocity’s $27M Bet on AI “Intent” Get full access to The Spiro Circle at www.thespirocircle.com/subscribe

Manny Marotta has a theory about why America’s 250th birthday felt subdued. It isn’t purely politics, though politics is tangled up in it. It’s the structure. There are simply too many feeds now, and not enough shared ones. And the ones that do break through to the masses get read as political, whether they mean to be or not.I sat down this weekend with Manny for the second time. He’s the creator and curator of the Live History Project, which takes a couple of accounts on X and posts in real time what’s happening in that moment in history.There’s:25 years ago - @25YearsAgoLive50 years ago - @50YearsAgoLive100 years ago - @100YearsAgoLiveand 250 years ago - @250YearsAgoLiveRight now, that means we’re living through 2001, 1976, 1926, and 1776 simultaneously. He pointed me back to America’s bicentennial in 1976, which he says was one of the only major stories of that year, competing for attention with little more than an Olympic Games. This year, Independence Day landed alongside a World Cup on American soil, an ongoing Iran conflict, a White House renovation project, and an MMA match. It also took place with a media landscape noticeably divided along party lines and contrasting levels of patriotism between political ideologies. “Now we have so many different news cycles, so many different news sources that people are following,” he told me. “It’s just oversaturation.” His 250-years-ago account picked up roughly 300,000 followers and 20 million views in the days around the holiday, almost entirely because a political audience decided it mattered.That’s where the story gets complicated. Manny insists the account isn’t doing anything ideological - he just posts digitized letters and meeting minutes from the Library of Congress that are available to everyone, without commentary. And yet that neutrality is precisely what got it adopted as, he describes, a patriotic rallying point by people ‘on the right’. “A neutral or positive view of not just the American Revolution but American history in general has become, in recent years, sort of right-wing coded,” he explained. “So if you are even reporting in an academic sense what happened, a lot of people do tend to see that as right-wing.”Meanwhile, news outlets covering the holiday split along familiar lines: CNN described the mood as shaping up to be “a big blah.” The New York Times ran an op-ed blaming the Trump administration for deflating the day, then was forced to revise its own headline. The Washington Post called it “an unfortunate metaphor on national divisions.” Disney, by contrast, ran wall-to-wall patriotic programming, and outlets like The Free Press leaned into celebratory content.But Manny didn’t spare the current administration either, telling me the patriotic messaging he’d seen recently during a trip to Washington, DC, centered more on a single political figure than the anniversary itself: “The only America 250 content that I saw were giant banners with Donald Trump’s face on them... nothing about the anniversary itself, more about the person who happens to be president.” His hope, he said, is “to create maybe a simulation of the monoculture that we had in the past,” which is academically sourced, uncaptioned, and a return to the apolitical. So he is trying to hold a neutral center by republishing old letters, in a country where an audience conditioned by fragmentation has decided that the center no longer exists. But a Jefferson draft, posted without a caption, still lands as a statement to somebody.This is my second conversation with Manny Marotta. Watch the first, from February, about the Live History Project’s 2001 account, here. Get full access to The Spiro Circle at www.thespirocircle.com/subscribe

Jira tickets used to sit open for years. A medium-severity vulnerability, flagged by a routine scan, could be assigned to an engineer who had bigger fires to fight. It had largely been that way for years: patch the criticals, manage the highs, let the mediums age. When it came to CVEs (Common Vulnerabilities and Exposures), a publicly available list of known cybersecurity flaws in software and hardware, nobody was going to weaponize one rated 5.4. But in today’s world, that’s no longer true.“In 2020, [if] there was a CVE reported and a security hole, it would take more than a year until there was a public exploit,” said Shimon Tolts, CEO and co-founder of Tel Aviv-based cloud security startup Copperhelm. “Nowadays, with Claude and OpenAI and other players, the time has shrunk from one year to one day. So now you treat every CVE, every security issue that you have, as immediately exploitable.”The data confirms what Tolts describes. The mean time between a vulnerability being discovered and its exploitation has dropped from nearly a year in 2021 to just over a day in 2026, with industry projections suggesting the window will shrink to one hour by 2027. Rapid7’s 2026 Global Threat Landscape Report found that what once unfolded over weeks now materializes in days (and in some cases, minutes), with the median time between vulnerability publication and inclusion on CISA’s Known Exploited Vulnerabilities catalog falling from 8.5 days to five.The implications invalidate an entire category of enterprise risk management.For decades, security teams built their workflows around severity scores. The National Vulnerability Database, operated by the National Institute of Standards and Technology (NIST), classified every disclosed flaw as ‘critical’, ‘high’, ‘medium', or ‘low’ - and organizations built their response hierarchies accordingly. Fix the criticals immediately, schedule the highs, and then defer the rest. That model is now under institutional strain: CVE submissions surged 263% between 2020 and 2025, and starting April 15, 2026, NIST announced it would only prioritize enrichment for a narrow subset of vulnerabilities, such as those already on CISA’s exploited list, those affecting federal systems, or those covered by Executive Order 14028. This would leave the majority of newly disclosed flaws without severity scores. “You’ll no longer be able to use the old risk management methodology of saying ‘I’m only going to fix criticals’,” Tolts explained. “Because you’re not going to have a severity anymore.”The shift has a compounding effect. AI models are not only accelerating exploitation timelines, but they are also discovering vulnerabilities at a rate that human analysts cannot process. NIST enriched nearly 42,000 CVEs in 2025, 45% more than any prior year, and forecasts from the Forum of Incident Response and Security Teams projected a record 50,000 additional CVEs to be reported in 2026 (these figures do not yet account for the accelerating contribution of AI-powered vulnerability discovery tools like Claude Mythos and GPT-5.4-Cyber).Every day, the cyber world is facing more vulnerabilities, faster exploitation, and fewer severity scores to guide triage. But security teams are still largely operating through manual workflows designed for a different era.Copperhelm’s answer is autonomous investigation and remediation, already backed by a $7 million seed round led by TLV Partners and deployed in Fortune 500 environments. The platform uses a proprietary “Context Lake” to structure cloud data across environments, enabling AI agents to continuously monitor infrastructure, investigate threats, and execute real-time remediation without manual handoffs. Tolts describes the practical effect in terms his customers already understand: one client arrived with 10 million open vulnerabilities and two home-made severity categories above “critical” — labels they had invented themselves because the official scale had run out of runway.“Your window of response has shrunk, and you need to autonomously take care of it,” Tolts said. “It’s no longer the case where you can just open a Jira ticket and wait for some engineer to fix it in one year or one month, because now you’re gonna get exploited very, very fast.” Get full access to The Spiro Circle at www.thespirocircle.com/subscribe

In the early days of Teramount, investors kept asking CEO Hesham Taha the same question. He had a platform that could connect chip to chip using light instead of electrons, considered a technical feat he and co-founder Avi Israel had spent years developing. But the problem, he recalled, was everything else.“We thought it was a great idea, see how easy we can connect the light to the chip. Everyone will use that,” Taha said. “But it turns out to be exactly the opposite in the first few years after our inception.”The problem for Taha and Israel was that for nearly a decade, Teramount was trying to solve a problem the semiconductor industry didn't know it would have. The market didn’t yet exist, nor did the supply chain. “The most critical point and the big barrier at the beginning of this journey was, ‘What is the product? What is the use case?’ This is what every investor kept asking us, and we failed to give a good answer.”Years passed, and that question, once unanswerable, just got answered. In April, Molex announced the acquisition of Teramount for approximately $430 million, roughly 7-8 times the $58 million that the Jerusalem-based startup had raised across its lifetime.The exit is a case study in what might be called ‘The Patience Trade’: bet on a technology before the world knows it needs it, endure years of uncertainty, and trust that the market eventually catches up. In Teramount’s case, it took two pivots, one global AI infrastructure boom, and a seed investor willing to see something others couldn’t: Lior Handelsman.Today, Handelsman is a Managing Partner at Grove Ventures, and before that, was a co-founder of SolarEdge - so he himself is no stranger to building technologies into markets that don’t yet exist. When he first encountered Teramount, his instinct was to pass. “There was no market even when I met them at the beginning of 2021,” he said. “And I was pretty much willing to tell them, ‘Look, guys, very nice, but I can’t see the market’.”Handelsman ended up reaching out to senior contacts at NVIDIA, Broadcom, Cisco, and Intel — companies that would eventually need exactly what Teramount was building. “When I told them, they said, ‘That’s a big problem. Connecting fiber to chip? That’s a big problem. We are all going to need that in four to five years’.” Grove led the seed round in 2021, and the next few years compressed faster than anyone predicted. The 2022 generative AI explosion turbocharged demand for the kind of optical connectivity Teramount had spent years perfecting. Co-packaged optics — the integration of optical engines directly with compute chips to reduce power consumption and latency — moved from a niche conference topic to an urgent industry priority. And so Teramount, having spent years building the ecosystem relationships and supply chain partnerships that most competitors hadn’t started, was suddenly indispensable.Taha points to two moments that changed Teramount’s trajectory. The first was 2017, when co-packaged optics began to emerge as a defined technology category. The second was 2024, when AI infrastructure demand made optical connectivity not just desirable but necessary. “This was the major and significant pivot in our journey,” he said.Strategic investors followed the technical validation. AMD, Samsung, and Hitachi all joined Teramount as the company’s direction became increasingly legible to the industry. Handelsman describes the combination of financial investors alongside strategic ones as the signal that a company has crossed a critical threshold: “That’s like a sweet spot. A financial investor is leading the round, saying that there is still upside, and strategic investors, who can all be customers.”For Taha, the Molex acquisition was less a finish line than a pragmatic decision about speed. “We had a great technology, we have a great product, but we need to move fast to match the market speed,” he concluded. Molex, a proven interconnect manufacturer with global production capabilities, offered the industrial scale that the Jerusalem-based startup could not self-assemble quickly enough.The patience trade paid off. The lesson it offers is about endurance, and about finding investors willing to hold the same long view as the founders they back.The Spiro Circle is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber. Get full access to The Spiro Circle at www.thespirocircle.com/subscribe

Everyone is talking about a water crisis brought on by AI use. But think of this: the data centers powering the AI economy are not built from code. They require concrete, steel, cranes, and the coordination of hundreds of subcontractors across millions of square feet of new floor space. And right now, the construction industry is struggling to keep up.Capital expenditure from the 14 largest publicly owned data center operators globally is projected to approach $750 billion in 2026, up from under $450 billion the year prior. The Stargate Project alone, a multi-year, $500 billion plan to deliver up to 10 gigawatts of AI-ready power, was formed as a joint venture between OpenAI, SoftBank, Oracle, and MGXis and spans multiple U.S. states. Meanwhile, the median cost of building a data center hit $445 per square foot this year, up 7.4% from 2025, with average costs skewed far higher by hyperscale projects. The pipeline is enormous, and yet the pressure on builders to deliver is greater still.Erez Dror has seen this shift from both sides. A structural engineer and former construction superintendent who spent over a decade on job sites in Israel, he co-founded workforce intelligence platform, Genda, which last year was acquired by Buildots after a $5.5 million Seed round. He recently stepped into the new role of VP of General Contractors & Genda at Buildots, which to date has raised $166 million. After the acquisition, the joint entity is now positioning itself as the operational backbone for exactly the kind of complex, fast-moving builds that the AI infrastructure boom demands.“A product executive who worked on the biggest project Genda was on, a $600 million project, took them four years to build,” Dror told me. “He moved to build one of the biggest data centers in the U.S., which was $6 billion — 10x the scale — and they built it in three years. A year less, and 10x times the scale.”The compression reflects a new standard being set by hyperscalers who come from a software-first culture and expect physical construction to behave accordingly. “A person who works for Google and is used to building software that doesn’t break expects to get a building that doesn’t break at the same quality,” he said. “They’re setting a new standard, which I believe will eventually trickle down to everything.”The challenge is that construction remains one of the most fragmented, data-poor industries in the global economy. Unlike a tech organization, where a single executive decision can transform operations overnight, construction is built around individual projects with its own lead, subcontractors, or even its own tolerance for disruption. Change management, Dror argues, is “just a different beast.”That fragmentation is precisely what Buildots is trying to solve. The platform ingests two data streams: weekly 360-degree camera footage from job sites and the project’s 3D building model, to use computer vision to identify what has been built versus what was planned. Genda, meanwhile, tracks where workers are on-site in real time, anonymously, using an app-based system that Dror designed around behavioral incentives rather than hardware. Together, Buildots says the platforms offer visibility into both the work being completed and the labour required to complete it.“We know the output, we know the input… we know what was built, and we know what efforts or how many resources were needed to get there,” Dror explained. “We’re the only solution in the world that can provide you with the full picture. Not even at scale — just to provide that.”In April, Buildots formally launched a new product category, which it is calling “construction intelligence”. It frames itself as the operational platform for an industry that can no longer afford to rely on gut instinct and fragmented spreadsheets. And as data center construction starts reached $9.8 billion per month through April 2026 (300% more than levels seen a year ago), the timing for a platform that can turn chaotic job sites into predictable delivery machines has never been better.“When you need to build a facility like a data center that is very detail-oriented, and you need to build it very fast, and every day of delay is millions, if not tens of millions, if not billions, in liquidated damages, you really need to make sure you finish on time and you know what the hell is going on in your project,” he added.Watch a 5-minute preview of this conversation here: Get full access to The Spiro Circle at www.thespirocircle.com/subscribe

Everyone is talking about the $90 billion M&A wave that hit the sector in 2025: Google’s $32 billion purchase of Wiz, Palo Alto Networks’ $25 billion buy of CyberArk. People love to talk about those, but almost nobody talks about how few companies can actually write those checks. Except Ofer Wolf.As Akamai’s SVP and General Manager of Enterprise Security, Wolf sits on the buy side today. But he got there as COO of Guardicore, the Israeli startup Akamai bought in 2021 for roughly $600 million. It makes his reading on the market less theoretical and more one from actual memory. “I can see the picture from the three sides: the entrepreneur side, the investment side, and the potential acquirer side,” he told me. What he sees from all three is a market punishing founders who raise at prices the buyer pool can’t support. “There are [fewer] than five companies that can do constant acquisition of over a billion,” he said, calling the Wiz-type deals an exception, not the rule. “Most of the acquisitions in cybersecurity, which is a fragmented market, are limited to hundreds of millions of dollars, usually.”That’s what he called “the poison pill”: raise too high, and you haven’t built a war chest, you’ve actually shrunk your buyer list to almost nobody. “Your valuation decision has set the future of the company to be a different future,” Wolf told me. “[It’s] probably the most strategic decision that is overlooked.” His advice isn’t to dream smaller, or only aim for the IPO - rather, it’s to prepare a few plans and act accordingly. “You need to have in the back of your mind another plan B, plan C… because eventually, most of the cybersecurity companies, the successful ones, were acquired by somebody.”Akamai’s own Tel Aviv record reads like that discipline in practice. It completed four Israeli cybersecurity acquisitions in five years, most recently the roughly $205 million purchase of browser-security startup LayerX (expected to close later this year), comfortably inside Wolf’s “hundreds of millions” band and nowhere near the CyberArk or Wiz price tags. Wolf describes the Israeli geography of the company’s acquisitions as serendipitous: “Two miles around our office, you’ll find a major part of the cybersecurity industry, from Palo Alto, Check Point, CrowdStrike, to small startups.” When Akamai went shopping for a workforce-security target, it checked six or seven companies across Israel, the US, and Europe, only to find the winner “sitting in the building right down the street, five minutes from our office.”Akamai's ability to stay active on the acquisition front is helped by its strength elsewhere in the business. The company's stock recently surged after disclosing a $1.8 billion cloud deal with what it called a leading frontier-model provider, widely reported to be Anthropic. That gives Akamai the luxury of approaching acquisitions from a position of strength rather than necessity.The same recalibration Wolf sees in acquisition markets is also showing up in Israel’s labor market. Wolf calls it “two headwinds”: AI-driven layoffs squeezing junior hires, and a shekel that’s strengthened over 20% in a year, making Israeli engineers pricier than they were twelve months ago.The Israel Innovation Authority’s 2026 report backs him up, and something I previously wrote about in JNS: for the first time in a decade, Israeli high-tech R&D headcount actually fell by roughly 3,500 jobs. “The reduction in force is hitting the news,” Wolf said, “but the slow drift won’t make the news.”He’s surprisingly upbeat about where this lands. In his mind, laid-off engineers will become founders of future great companies. And new hires get “a little bit more sensible” about pay. His own team is leaning into “AI-native junior guys” at the same dollar budget that bought fewer senior people a year ago.So currently, valuations and salaries are both drifting back toward what the buyers are willing to pay. And for Akamai, that is an easy pill to swallow.Watch us discuss this idea in a 5-minute preview: The "Poison Pill" That Kills Cybersecurity Startups Get full access to The Spiro Circle at www.thespirocircle.com/subscribe

The robotaxi race has a winner’s podium that everyone can name. Waymo is expanding fast, with TechCrunch reporting that its fleet has now crossed 3,000 vehicles, completing over 500,000 trips per week across multiple U.S. cities. Tesla launched its robotaxi service in Austin. Chinese players like Baidu and WeRide are scaling aggressively at home, as well.But Igal Raichelgauz, CEO and founder of Israeli autonomous driving company Autobrains, thinks the industry is looking at the wrong scoreboard entirely.“The major market, the biggest opportunity today, is in the Western world for every car,” Raichelgauz told me. “We have today over 1.5 billion cars on the road. These are not addressed - not by Tesla, not by Waymo, and not by the Chinese players. And that’s where we see the biggest opportunity.”Waymo’s overall fleet is impressive and certainly dwarfs Tesla's. Take the state of Texas, which, as of recent state DMV filings, showed that the Musk-owned company had just 42 autonomous vehicles authorized for driverless ride-hailing, compared to Waymo’s 577. Meanwhile, Chinese players cannot realistically scale into Western markets for geopolitical reasons, leaving the competitive map with a conspicuous gap at its centre.Autobrains is positioning itself directly in that gap, using agentic AI to split the driving task into thousands of specialized agents rather than training a single monolithic model. The Israeli company has raised over $140 million from investors that include BMW, Toyota Ventures, Magna, Continental, Temasek, and others, and claims to hold more than 300 patents related to AI and autonomous driving. It’s where Raichelgauz sees the company getting the best chance to join that winner’s podium.The logic is as follows: Waymo’s approach is built on expensive sensors, LiDAR arrays, and HD maps that require enormous upfront investment before entering any new city. “When you need to move to a new city, you need to invest tens, if not hundreds, of millions into this infrastructure work,” he said. Tesla, on the other hand, has made the consumer vehicle its canvas, but still hasn’t delivered on unsupervised autonomy. “You don’t have a personal car that you can buy from Tesla with FSD (Full Self-Driving) that can run in a way that the person can start working, watch videos, and read emails. You must work in a mode that is ‘eyes on’.”The result, as Raichelgauz puts it, is “a disconnect”: one side has autonomy that doesn’t scale, the other has scale that isn’t yet truly autonomous. Industry analysts broadly agree: while Level 2 ADAS is expected to become the standard baseline across new vehicles through the 2030s, Level 5 autonomy remains a distant goal on a potentially multi-decade timeline.The practical payoff is that it can run on standard automotive sensors and existing compute platforms, without requiring the expensive hardware stacks that make Waymo-style deployments hard to replicate. “We want to make sure this technology becomes mainstream, where people can get the time back from driving and start working,” Raichelgauz said, describing eyes-off Level 3 capability as the near-term commercial target rather than full robotaxi autonomy.To get there, the company has secured partnerships with Uber, NVIDIA, and VinFast to run two contrasting real-world proving grounds. Germany’s Munich for European regulatory rigour, and South Asia, particularly Vietnam, for the sheer chaos. “If we can solve autonomous driving in Hanoi on a regular car, we can solve it everywhere,” Raichelgauz said. Chinese OEMs are already taking the lead on sophisticated ADAS integration in their home markets, which means the window for Western-aligned players to establish an OEM-agnostic standard is finite. Autobrains is betting that the company that cracks affordability and scalability will ultimately define what autonomous driving looks like for most of the world’s drivers.“The first starting point is ‘eyes off’,” Raichelgauz said. “When you can take your car driving from home to work without really supervising it… without paying tens of thousands of dollars to upgrade to a robotaxi.”You can watch the entire exchange in the video above. Thanks for reading The Spiro Circle! This post is public so feel free to share it.Watch us discuss this topic in a 5-minute preview of this episode: Tesla, Waymo, and the Self-Driving Race Nobody's Won Yet Get full access to The Spiro Circle at www.thespirocircle.com/subscribe

Every boardroom in the world is having the same conversation right now. A vendor pitches an AI product. Someone on the executive team then asks the question that has become the most disruptive five words in enterprise software: “Can’t we just build it?”Increasingly, the answer appears to be yes. The barrier to generating a working prototype has collapsed. With the right prompt and an afternoon with plenty of coffee, a modest technical team can produce something that looks convincingly like the product they were just quoted six figures to buy.And so, the logic follows: why pay for what you can build yourself?Daniel Zahavi thinks this instinct represents the peak of the current AI hype cycle and that the correction will be painful for the companies that followed it. Born in Kermanshah, Iran, in 1985, he immigrated to Israel at the age of 15, studied electrical engineering at the Technion, and earned a doctorate in Information Theory, the mathematical field underpinning modern large language models.During IDF service, he held one of the highest security clearances in the military, working on projects touching the Prime Minister’s Office and the Intelligence Corps, before going on to develop drone interception systems and offensive cyber capabilities. He has now co-founded Arito, an AI analytics platform for finance and revenue teams, which raised $6 million in seed funding last month.When Zahavi talks about commercial survival, there is biographical weight behind it. His defense technology business was blocked from export by Israel’s own Defense Ministry: a working product that couldn’t reach its market. He knows what it costs to build something that turns out not to be deployable.That experience sharpens his read on the ‘Build vs. Buy’ trap now playing out in enterprise AI.“Right now we are at the very top of that hype cycle that everyone believes that they can build whatever they need themselves easily,” he told me. “The amount of people that know exactly what they need and what they want is not very high. The portion that knows exactly how to describe that in very high resolution so that you can actually get what you need is even lower.”Building anything genuinely useful with AI requires a clear understanding of the actual problem, and the ability to specify it with enough precision so that a model can act on it reliably. Most organizations have neither. They have a vague sense of the pain and a vocabulary borrowed from demos. But that produces impressive prototypes and disappointing production systems.But the main point is what happens after launch. “Writing the code is only the first part,” he added. “Maintenance is a way, way bigger part of creating it the first time. I’m not even talking about security and privacy. A lot of the actual challenge continues afterwards.”This is the consideration in the ‘Build vs. Buy’ debate that goes ignored. The prototype is cheap, but maintenance is not. And unlike a purchased product, where maintenance, iteration, and accountability belong to the vendor, the self-built version belongs to whoever built it, permanently.Finally, Zahavi frames this as the difference between tools that produce what he calls “one-off artifacts” and tools that compound value over time. Asking an LLM a question and getting an answer is a one-off artifact, easy to replicate, easy to replace. But to build a system that learns how a specific finance team defines its metrics, tracks how those definitions evolve across fiscal years, and surfaces anomalies against that institutional context in real time is something much harder to build in a weekend or ‘vibe-code’.“The only question that they need to ask themselves is: ‘Are they creating continuous long-term value for their customers and not just a one-off thing that can be solved easily?’ Because if it’s a one-off thing, then the chances of them being replaced by an AI prompt [are] very, very high.”The hype cycle will correct. For Zahavi, who has spent a career building things in environments that were actively trying to stop him, like war zones, military bureaucracy, or the Defense Ministry that blocked his exports, the question of what survives hostile conditions is not theoretical.You can watch the entire conversation above, or you can watch a snippet of this particular topic below: Get full access to The Spiro Circle at www.thespirocircle.com/subscribe

Last week, I heard a number that should terrify every founder raising a Series A.Between 50% and 60% of Israeli tech startups that reach round A never make it to round B. Think of it: You’ve pitched, hustled, and convinced initial investors that your idea is worth betting on… yet statistically, you’re more likely to flame out before the next round than not.The reasons are messier than most founders want to admit. It’s not always the product or market. A lot of the time, it comes down to one hire. The first real sales leader you bring in to crack the US market. Get it right, and you’re soaring your company to new levels. But get it wrong, and you’ll be burning through runway while pretending everything is still fine.This week on The Spiro Circle, I spoke with Giora Gil-Ad about the most dangerous (and weirdly emotional) hire in Startup Nation - the first serious US sales executive.That’s exactly the world he operates in. As the founder of CQ Global, Giora specialises in one very specific, very high-stakes moment in a company’s life: finding the sales exec who will either unlock the US market or become a very expensive lesson.In our conversation, Giora puts the cost of a wrong VP hire at somewhere between $1.5 and $2 million, once you account for the salary, the team members who follow them out the door, the deals that slipped through the cracks, and the months of momentum lost.But the money is almost the easy part to quantify. What’s harder to measure is the founder who starts second-guessing themselves. The team morale evaporates, or investors start asking harder questions. A bad hire can cost confidence, and in the early startup space, confidence is everything.So what does getting it right actually look like? According to Giora, it starts with founders being honest about what they actually need, and accounting for cultural, business, and personal needs along the way. You can learn more about this whole area in the episode above. Get full access to The Spiro Circle at www.thespirocircle.com/subscribe

I’ll be honest: I entered into this most recent conversation for The Spiro Circle knowing almost nothing about quantum computing. I said as much to my guest, Itamar Sivan, co-founder and CEO of Quantum Machines, before we even started recording. That’s alright - most people don’t really understand it, he told me. Even scientists used to laugh at the idea that quantum computing would ever be commercially viable.But then, about halfway through our conversation, he said something that piqued my curiosity and made me put down my notes.The threat isn’t that quantum computers will simply ‘make things faster’. It’s that they’ll make things possible that are currently impossible. And one of those things is breaking the encryption that protects everything - and keeping cryptographers up at night. He cited potential examples as banks, messages, the NSA, and Bitcoin.“Quantum computers are not interesting because they’re going to take problems we solve today and solve them faster,” Sivan told me. “But rather they will take problems today we deem as impossible and make them possible.”The mechanism is an algorithm called Shor’s algorithm, which can factorize enormous numbers at speeds no classical computer could approach. Modern encryption is built on the assumption that factorizing very large numbers is effectively unsolvable. But by taking away that assumption, the entire architecture collapses. “Something that would take a hundred thousand years might be solvable at the scale of minutes,” he told me. Quantum Machines (QM) is a Tel Aviv-based company that has raised $280 million to build the orchestration layer running quantum processors. Founded in 2018, customers include academia, national labs, and the private sector. What struck me was that he raised this before it exploded as a mainstream story. At the time of our recording, he flagged that a newly published paper suggested quantum computers would need far fewer qubits to break encryption than previously thought. “We’re still digesting it. If they’re right, we’re going to see some big changes in the world in a few years.” And almost as an aside: “One of the claims is that it will be able to break the underlying encryption used for Bitcoin. Just that itself could be a big impact.”Research published between May 2025 and March 2026 shows that breaking widely used cryptographic systems may require far fewer quantum bits than previously thought. Estimates dropped from around 20 million physical qubits in 2019 to under one million by 2025. Papers from Caltech and Google in early 2026 prompted one Bitcoin security researcher to estimate a 10% chance that a quantum computer recovers a Bitcoin private key from an exposed public key by 2032. In April 2026, a researcher successfully broke a 15-bit elliptic curve cryptography key using publicly accessible quantum hardware — a 512-fold improvement over the previous public demonstration just months earlier. Google has already set a 2029 deadline to migrate its own authentication services to post-quantum cryptography. The so-called “harvest now, decrypt later” threat (adversaries collecting encrypted data today, waiting for quantum capability to mature before cracking it) means the clock is ticking, even though many still believe their things will be protected for many more years. Sivan’s broader point, the one I kept coming back to, is that quantum won’t replace the computing infrastructure we’ve built - but instead plug into it. It means the vulnerabilities we’ve built into that infrastructure travel with us.“Not a question of if,” he told me as we finished. “A question of when.”I didn’t know much about quantum computing before this conversation. But I think I know enough now to think that answer should concern all of us! Watch a 5-minute preview of our conversation on this topic, here: Get full access to The Spiro Circle at www.thespirocircle.com/subscribe

Imagine the scene: A developer at a large financial institution merged a routine code update. Nothing alarming yet, just a minor change that, on its own, meant little.But Terra Security’s AI agents were watching.AI agents flagged the change, verified a potential vulnerability, and then did something a human penetration tester probably wouldn’t have done. They kept looking. Eventually, they found two more vulnerabilities nearby, each individually insignificant. But they spotted a pattern and connected all three together.“1+1+1 = 1,000,” said Shahar Peled, co-founder and CEO of Terra Security. The result was a Remote Code Execution (RCE), a cybersecurity vulnerability that allows an attacker to run malicious code on a target system or server from a remote location. It is considered one of the most critical vulnerability classifications of its type. The customer found out from their vendor, not from an adversary.Founded in 2024, the Tel Aviv and New York-based startup has raised $38 million across a rapid Seed and Series A, and counts Fortune 100 enterprises among its customers. Its core product is an agentic offensive security platform where swarms of AI agents are trained to think and act like “ethical hackers”, running continuously across a company’s attack surface.The traditional model of penetration testing (hiring an external team once or twice a year to probe for weaknesses) was never designed to catch what Terra caught in that unnamed financial institution. “Until 2025, it happened on an annual basis mostly,” Peled explained. “Once a year, you hire someone externally to work for a week or two weeks... The reason you couldn’t do it continuously is that you couldn’t really train software to hard-code how adversaries think and act.”But AI has changed all that. Terra Security’s agents scan for known vulnerabilities and simulate the reasoning of an attacker, chaining together findings and verifying whether a vulnerability is actually exploitable rather than merely theoretical. But Peled is careful not to overclaim, and beat me to my own next question. “Are AI agents today better than any ethical hacker in the world? They’re not,” he said. “They don’t yet possess the creativity of the best ethical hackers. But they can be more scalable than anyone in the world. They can run continuously. They never sleep. They’re already better than the vast majority of ethical hackers in the world.” With AI, there are no longer cyberattackers who wait for annual review windows. Adversaries now use tech to find entry points faster, adapt in real time, and strike before defenders can patch. A point-in-time test is, by definition, already outdated the moment it concludes.Terra’s idea is that continuous, AI-driven offensive security is the only architecture that matches the pace of modern attacks. The chained vulnerability Peled mentioned in our conversation was only catchable because an agent was watching the moment the code changed - and not six months later, when a consultant finally showed up.“I still see too many organizations that say, ‘Okay, now we have AI in offensive security’,” he concluded, and as a slight warning to CISOs still budgeting for annual pen tests. “[They say] ‘I want to do the same thing I’ve done before, just faster, better, cheaper’. And that scares me.” Get full access to The Spiro Circle at www.thespirocircle.com/subscribe

There’s a new security risk out there, and it’s come to be known as The Shadow AI Problem.It suggests that the next major corporate data breach may not come from a sophisticated nation-state actor or a phishing campaign, but rather from an employee asking an AI chatbot to read or summarize sensitive company data.That’s the reality Itamar Golan has spent the last two years building a company around. As co-founder and CEO of Prompt Security (acquired by SentinelOne earlier this year for $250 million), he has become one of the voices warning of the gap between how fast enterprises are adopting AI and how little they understand about where their data is going. According to him, most CISOs focus on traditional attack vectors, but the real risk is employees pasting IP addresses into unauthorized tools.Prompt Security’s platform now detects nearly 20,000 distinct AI applications operating across enterprise environments. Golan clarified that the figure isn’t plugins or product variants, but 20,000 separate entities. “Today, essentially almost any SaaS application, website, native application running on your endpoint… we are converging towards a landscape where any one of those will be an AI application by itself,” he told me.The Spiro Circle is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.The visibility problem is one thing, but the training problem is another. Prompt Security’s research found that roughly 40% of AI applications, when surveyed at the configuration level, are set by default to train on the data they receive. “Not only has confidential data leaked out of your organization,” Golan explained, “it’s now potentially becoming part of the model’s brain.” Details like corporate strategy, personnel data, or legal documents will be available for everyone to see - and there is no obvious retrieval mechanism once embedded in a model’s training run. The sectors most exposed are also the typically traditional ones that are now moving fastest to catch up: Financial services, insurance, and legal firms are adopting AI precisely because it performs exceptionally well on their core workflows. “They find themselves in this very tricky situation,” he told me. “On the one hand, they are adopting AI the fastest, and the potential gain is immense, but the risk of making a mistake is so big as well.”It is a distinctly Israeli problem to be working on. Golan mentioned that when he surveyed the security stacks of Fortune 500 CISOs while building Prompt, he found that around 60% of the tools on their lists were built by Israeli companies. Startup Nation has given the world Check Point, CyberArk (acquired by Palo Alto Networks), and Wiz (acquired by Google). Now, Prompt Security, as part of SentinelOne, is trying to secure the AI layer that sits above all of them.“We cannot stay blind,” Golan concluded. “We must admit that our employees are using hundreds or thousands of AI applications. A big portion of those are able to train on the data we are sharing with them.” Acknowledging that reality, he argues, is the first step to acting on it. Get full access to The Spiro Circle at www.thespirocircle.com/subscribe

This isn’t the first time I’ve covered Israel’s foodtech sector. Back in 2022, reporting for CTech, I mapped the ecosystem at a moment of tension, when investment was holding up better than in any other tech vertical, but the skeptics remained.I was, and still am, bullish on Foodtech - at least at the start. I tasted 3D-printed burgers in Tel Aviv and called them “technically perfect, albeit creatively void.” I interviewed investors who compared the industry to early mobile phones — primitive first iterations, but with everything still to come. I wanted to delay a full embrace of alternative foods until the markets all caught up. Turns out many felt the same way. So years later, I wanted to revisit all of that with someone who’s lived it from the inside.Ilanit Kabessa Cohen has spent 25 years asking one question: what does it actually take to bring innovation to market? As the first Head of Innovation at Osem-Nestlé, a corporate venturing lead at Dole in Singapore, and now co-founder of the advisory firm URIKA, she’s seen the food ecosystem from virtually every angle — and she joins me to share what she’s learned.Our conversation opens with an assessment of Israel’s position in global foodtech. Despite being a relatively small player in terms of total funding (roughly $16 billion globally), Israel punches well above its weight: driven by its kosher culinary traditions, research institutions, a culture of cross-domain improvisation, and the Israel Innovation Authority’s risk-sharing model that few other governments have replicated.But Ilanit is candid about where the industry fell short. The first generation of alternative proteins disappointed consumers, investors, and believers alike. Not because the vision was wrong, but because first-generation products rarely win. She argues we’re now entering a correction phase, with more mature companies, better-tasting products, and a smarter understanding that the real action right now is B2B ingredients, not consumer-facing brands.The most forward-looking part of the episode covers what she calls “animal-free technologies” — a next-generation wave that goes far beyond food. Think collagen produced via precision fermentation for use in cosmetics, pharma, and nutrition. Or how biomaterials could replace shark liver extract or horseshoe crab blood in medical testing. She said how the next decade of opportunity lies in the convergence of food, health, and biotech - and finally, she discussed two opportunities: the Coller Startup Competition (now open, with a $100K prize) and URIKA’s Generate partnership program with CSM Ingredients for startups in sugar reduction and proteins.The Spiro Circle is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber. Get full access to The Spiro Circle at www.thespirocircle.com/subscribe

Datarails CRO Aviv Canaani has an unusual vantage point. He runs the full revenue engine of the financial planning and analysis platform for Microsoft Excel users — sales, marketing, partnerships — from New Jersey, while his marketing team operates out of Israel. He relocated to be closer to the North American customer base as the marketers stayed put. And after years of sitting inside both ecosystems at the same time, attending CMO sessions in Tel Aviv and building pipelines in the US, he’s reached a verdict most people in his position wouldn’t say out loud: the Israelis are better.It’s a claim that cuts against the instinct of almost every Israeli founder he’s encountered - and every company I’ve spoken to over the years. “Normally, when I speak to startups that are born in Israel, they want to send their sales and marketing overseas immediately. [It’s] the first thing they want to do,” I told him during our conversation. But Canaani’s experience runs the other direction.The Israeli edge, he claims, comes down to a cultural obsession with output. “When you talk with people in Israel, marketing leaders, it’s about how they built machines, how much the cost per meeting, how they’re running campaigns on Facebook and Google and all that.” American counterparts, he finds, often arrive at the conversation from somewhere else entirely. “A lot of CMOs and people in marketing I talk with in the US or Canada… can talk more about the brand, how things take time, like it’s a long-term investment.”“Tachlas mentality” explained He traces this back to something structural in Israel’s tech DNA: The concentration of adtech companies and the performance-marketing culture they seeded, and also what he calls “tachlas mentality”. He explained that this requires teams to be focused on results above everything else. The blend of that mindset with an unusually international talent pool (many ‘Olim’ from Britain, the US, or Europe) produces something Canaani finds hard to replicate in America.But there’s a catch - and one worth remembering. The same intensity that makes Israeli marketing so effective in the early stages carries a structural weakness as companies grow. “In North America, things are much more organized. It’s clearer how they create the messaging and the product marketing and how to make sure there is alignment between marketing and sales,” he told me. Israel, by contrast, tends to run so fast that alignment becomes a casualty. “It seems like sometimes it doesn’t even matter if marketing speaks one language and sales speaks another. Let’s just run fast. It’s speed above everything else.” The American advantage, then, is less about raw marketing talent and more about institutional discipline. “In North America, maybe it’s hard in the startup phase, but once they’re a bigger company, they have better processes — how to run things, how to stay on point.”So what Canaani is describing is a stage-mapping problem. Israeli performance marketing is almost perfectly calibrated for the zero-to-one phase: find the signal, iterate fast, fill the pipeline before the runway ends. But American marketing discipline becomes the dominant advantage once you’re scaling and when the team is distributed. Move fast and break things, but then slowly mold them into greatness.The companies that figure out how to sequence both are the ones most likely to build something that lasts. Datarails, with teams operating on both sides and a CRO who has lived inside both cultures simultaneously, is running that experiment right now. [5-minute preview: Why Israeli Startups Are Better at Marketing Than They Think] Get full access to The Spiro Circle at www.thespirocircle.com/subscribe

Every few years, a new American administration arrives in the Middle East convinced it can start fresh. Trump’s team was no different. They came to the problem with a clean slate and nothing but the confidence of a New York real estate mogul. They produced two documents across both his terms: the January 2020 plan and the October 2025 twenty-point Gaza framework.The result, according to my guest Gidi Grinstein, was that they landed exactly where everyone always lands.“Even Trump ends up landing very close to where Nixon landed, to where Carter landed, to where Clinton landed,” he told me. “Because there is a gravitational force that is shaping these negotiations.”Gidi Grinstein has seen the Middle East from angles most people never will. At 29, he was the secretary of Israel's negotiating delegation at Camp David and the youngest person at the 2000 Summit. He spent years inside the machinery of the peace process drafting texts, aligning teams, and managing the distance between what leaders said in public and what they were willing to accept in private. Today, he runs Tikkun Olam Makers (TOM), a global initiative using open-source 3D printing to bring affordable prosthetics to people who can't access or afford conventional ones. While we intended to speak mostly about TOM, our conversation stayed on peacebuilding, negotiation, and his view of politics today.The force, he said, traces back to “the most brilliant and American diplomat of the last hundred years”, Henry Kissinger, and the architecture he designed in the 1970s. It was a framework built not around Israeli or Palestinian interests, but around American hegemony in the Middle East. Half a century later, and it is proving so durable for Washington that no administration, however disruptive, can break from it. The 2020 Trump plan's "two nation states for two people" echoes UN Resolution 181 from 1947. The 2025 Gaza framework in places reads like a revamped version of the Oslo Declaration of Principles from 1993. “You would be stunned by the amount of similarities,” he told me.What’s interesting this time around is that both countries - Israel and the US - face impending elections mere days apart, promising to shake up not just the political makeup for both sides, but potentially the leadership of one. This creates what Grinstein calls the clock problem: Israeli and American leaders, under electoral pressure, always want a deal now. Their counterparts (Arafat then, the Iranians today) operate on an entirely different political timeline, with every incentive to wait out a weakened or transitional government.“The synchronization of the political clocks is very important in getting the deal,” he said. Trump, he suggests, may be walking into the same trap by pushing hard before November while Tehran calculates what comes after.The gravity doesn’t guarantee peace, but I realized it means the frameworks are always roughly the same so long as the Americans are involved. And so far, history is showing us that they always find their way back to them.You can catch the entire conversation above. And expect more analysis from our conversation in future newsletters. [5-minute preview: Watch Gidi explain this in a YouTube clip, “Trump Thinks He's Rewriting Middle Eastern History. He's Repeating It.”]The Spiro Circle is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber. Get full access to The Spiro Circle at www.thespirocircle.com/subscribe

As conflict dynamics shift across the Middle East, from disrupted shipping lanes to drone warfare, a new question is emerging: who controls the software behind autonomous systems?Military power used to depend on access to advanced weapons systems, often built through international supply chains and dominated by a handful of large contractors. Today, conflicts in Ukraine and Iran, and tensions with China, are highlighting bottlenecks on critical technologies and the instability of disrupted supply chains. One Israeli company may have an answer to this new challenge. Udi Oster is the co-founder of eyesAtop, a startup building AI-native universal controllers for drone fleets. The company has spent the last three years making a case that the strategic asset in modern warfare isn’t any particular drone, but it should be the operating system above them.“Locking in to one vendor with one platform is something that in today’s world is very difficult,” Oster told me. “You want to have the flexibility to get the best technology at the point of time of interest and use it immediately.”Militaries around the world are accumulating drones from dozens of manufacturers, but without a common interface, any shared AI layer, or no easy way to retrain operators when hardware changes. EyesAtop’s platform intends to integrate into any drone, from any vendor, under one controller trained on over 500,000 hours of live IDF operational data.The Spiro Circle is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.The geopolitical context of conflicts and wars has expanded this market. Global defensetech VC hit a record $49.1 billion in 2025, nearly double the prior year, driven largely by autonomy and AI. American firms like Anduril have already moved into Taiwan, South Korea, Japan, and Singapore, selling hardware to nations trying to face off against the Chinese military. eyesAtop is pursuing a different layer of the stack: not competing on the drone itself, but selling the so-called brain that integrates whatever drones those nations already operate or plan to buy.Oster draws a sharp distinction between the American market and everyone else. America is its own category: It accounts for more than half of global defense spending, it has its own procurement logic, and its own concept of operations. EyesAtop already has a U.S. co-founder, a U.S. base, and existing deals with American military commands. For the rest of the world, the company offers a full-kit solution, where it selects the best available platforms globally, integrates them under its universal controller, and delivers a turnkey reconnaissance or strike capability to militaries that lack the R&D infrastructure to build it themselves.“Most of the countries outside of the U.S. lack the infrastructure and the R&D budgets even to get to the same type of level as Israel and the U.S.,” Oster said. “I would look at these countries differently.”The fundraising backstory underscores how fast the landscape has shifted. Three years ago, Oster says, virtually no Israeli VC would touch defense. The stigma was visible and impacted reputational and commercial opportunities. But the world changed after October 7, 2023, and today, funds are competing for allocations in a sector that now ranks among the top three investment themes globally.The longer-term vision Oster sketches is more ambitious than any single product cycle. As robotic systems multiply on the battlefield, army headcount becomes less relevant than software sophistication. Today, wars can be fought by one operator controlling multiple autonomous platforms that were trained in actual combat. “Instead of having a whole company,” he says, “you have two people, but they would operate a company of robotic systems.”He calls it "the ghost squad." For the allies now looking to build drone sovereignty in the middle of an active regional war, it may also be the next software contract they don't know they need. And that balance of power is moving up the stack.[3-minute preview: Optimism in Defensetech and the future of deterrence] Get full access to The Spiro Circle at www.thespirocircle.com/subscribe