🚀 Portfolio in Action

Cnuic: We co-led Cnuic Technologies’ $3M pre-seed alongside Blankspace Ventures, with participation from Silicon Roundabout Ventures, Phasechange, Sands Capital, and Superlative Ventures. Based in Edinburgh, Cnuic has built a fundamentally new type of photolithography device, one that uses the wave properties of light to enable rapid, reconfigurable production of photonic chips with enhanced 3D control. Something that was previously impossible.

GRAI: We participated in GRAI’s $9M seed round, co-led by Khosla Ventures and Inovo.vc, alongside Tiny Supercomputer Investment Company, Flyer One Ventures, the a16z Scout Fund, and angels Andrew Zhai, Greg Tkachenko, Rob Reid, and Dima Shvets. GRAI isn’t building another generative music tool. They’re building the social and interactive layer that sits on top of music you already love: remixing tracks, changing styles, sharing derivatives with friends.

Pointee: We led Pointee’s $1.4M Seed round, with Seed Starter ČS and Seed Starter Slovenskej sporiteľne participating. Deploying one AI agent is easy. Managing hundreds, knowing what they’re doing, what they’re costing, and when they go wrong, is a fundamentally different problem. Pointee’s platform gives enterprises a unified layer to orchestrate, monitor, and govern AI agents at scale. Their core technology, Guardian Agents, acts as always-on AI operations support: detecting anomalies, tracing root causes, and automatically executing remediation playbooks.

Edmund: Edmund closed a €2.5M Seed round led by FORWARD.one, with participation from University2Ventures and Tensor Ventures. Edmund builds AI agents that understand how machines actually work, down to the level of PLC programs, schematics, and maintenance histories, so factory technicians can diagnose faults in minutes instead of hours.

AppFactor: ForgeCatalyst is now live, the agentic orchestration layer for the work nobody wants to do: corrective fixes, adaptive patches, dependency updates. Assign objectives; ForgeCatalyst runs multi-agent parallel workflows with built-in verification and human-in-the-loop controls. The maintenance treadmill, automated.

Antiverse: Co-founder Murat Tunaboylu sat down with The Entrepreneurs Network to unpack where AI-driven drug discovery is heading, from why GPCRs and ion channels remain the most compelling (and underserved) targets, to how the economics of medicine are shifting from blockbusters to personalised treatments, and why the UK’s regulatory stance might be more progressive than the headlines suggest.

PangeAI: The team launched on Product Hunt with Greenland as their playground, letting users query satellite, sensor, and geospatial data in plain English. Turns out you don’t need to buy Greenland to play with it.

👩🏽‍💻 Deep Tech Scene

DeepSeek drops V4: a year after rattling Silicon Valley. The Chinese lab unveiled V4 Flash and V4 Pro, with a Hybrid Attention Architecture for better long-conversation memory and a 1M-token context window that fits entire codebases in a single prompt.

Google’s TurboQuant could reshape inference economics. Unveiled at ICLR 2026, TurboQuant uses PolarQuant rotation and Johnson-Lindenstrauss compression to slash KV cache memory overhead by ~6x, the biggest bottleneck in serving large models with long contexts. As context windows grow, memory optimization may matter more than parameter count.

Cambridge builds a brain-like chip that cuts AI energy 70%. A nanoelectronic memristor made from modified hafnium oxide that processes and stores information simultaneously, mimicking how neurons work. Unlike conventional chips that waste energy shuttling data, this device operates at ultra-low power. If it scales, the implications for edge AI and data center sustainability are enormous.

MIT’s EnergAIzer predicts GPU power usage in seconds, not hours. Researchers built a framework that estimates GPU power consumption for AI workloads with only 8% error, matching approaches that take hours, but in seconds. When efficiency gains are multiplied across hyperscale compute, they translate into real megawatts saved.

GPT-5.5 enters early access. Early testing shows meaningful advances in coding, research, and creative tasks, the model excels when combined with tools, enabling complex outputs (academic papers, full software projects) with minimal input. Limitations in long-form creative writing remain, but the trajectory is clear: AI systems that do, not just draft.

📣 Events

Prague Hardware Meetup × PrusaLab. The very first Prague Hardware Meetup, co-organized with PrusaLab at Automat Matuška. Brian Walker on REVEL and Physical AI; Petr Vacek on architecture, design, and digital manufacturing, plus an open mic that made the night.

DeepTech Breakfast in Prague. Kyle Park of Harvest Management Partners joined Petr Ulvr and Jakub Kouril for an intimate gathering of founders, investors, and operators. The takeaway: CEE founders are building IP-driven companies that are not only technology breakthroughs but strategically relevant to major acquirers.

AI Tinkerers Prague × ValkaAI: Machine House Office. Engineers, researchers, and builders gathered for a deep dive into how you build a low-latency, real-time AI sports commentator. Jan Brukner, Jan Petrov, and Jan Vykruta walked through the full stack live: speech generation, LLM optimization, and real-time infrastructure. The kind of technical depth you don’t get from a slide deck.

SelectUSA Startup Pitching at US Embassy Prague. Martin Drdul joined the jury alongside Senta Cermakova, Pavlína Louženská, and R. Taylor Moore. Congrats to the winners: 🥇 Dronetag, 🥈 INFRAHEX, 🥉 Mewery. Dronetag earned a free registration to the SelectUSA Investment Summit (May 3–6, National Harbor, MD).

dstack × Crusoe × SGLang: AI Infra Meetup, San Francisco (Apr 30). 500+ signups. ~200 curated attendees from the GPU space. Talks from RadixArk, Makora, Tinfoil, Metamorphic, and others.

Nuqleus Pitch Storm at CEE Deep Tech. Ondrej Lipold is joining the expert panel, evaluating the next wave of deep tech startups alongside investors from AYMO Ventures, Silicon Gardens, and Deep Tech Momentum.

Podcast: Daniel Hastik on Budoucnost nepráce. Three signals from the conversation: (1) The CEE opportunity is real, and time-bound. (2) AI is infrastructure, not a feature, that’s the line separating deep tech from everything else. (3) How a team uses AI internally tells us more than their product roadmap. Listen here →

DYNANIC × Verification Academy Live, Brno (May 21). A full-day seminar on functional verification at FIT VUT Brno, co-organized by DYNANIC. Three sessions from the team: Marcela Zachariasova and Lukáš Kekely on DYNANIC productivity with Questa One, AI adoption in verification (based on an industry-wide survey), and a closing panel on the role of AI in verification.

Cheers,

The Tensor VC team

The semiconductor industry has run the same basic playbook for sixty years. Photolithography — the process of etching circuit patterns onto chips using light — has been refined to extraordinary precision, but its fundamental architecture hasn’t changed. What has changed is the physics: silicon-based chips are approaching their physical limits, and the performance gains that once came automatically from shrinking transistors are slowing to a crawl.

Photonic chips offer a compelling path forward. By transmitting data using photons rather than electrons, they deliver dramatically higher speeds at a fraction of the energy cost — a combination that matters enormously for AI inference, data center cooling, and the next generation of computing infrastructure. The bottleneck has never been the idea. It’s been manufacturing: photonic chips are technically complex and expensive to produce at scale, and the tooling to make them efficiently simply hasn’t existed.

Cnuic, an Edinburgh-based 🏴󠁧󠁢󠁳󠁣󠁴󠁿 deep tech company, has built that tooling.

The company has developed a working prototype of an entirely new type of photolithography device — one that exploits the wave properties of light to enable rapid, reconfigurable production of photonic chips with enhanced 3D control. This is not an incremental improvement on existing methods. It is a fundamentally different approach to patterning chips, making possible what was previously impossible: flexible, high-precision manufacturing of photonic structures that legacy equipment cannot produce.

Cnuic has raised $3M in a pre-seed round led by Tensor Ventures and Blank Space Ventures, with participation from Silicon Roundabout Ventures, Phasechange, SANDS, and Superlative. The round was oversubscribed — a signal of how seriously the deep tech investment community is taking both the technology and the team behind it.

Why This Matters Now

The AI infrastructure buildout has made the economics of photonic chips newly urgent. Data centers running large-scale model training consume staggering amounts of power — much of it lost to heat generated by electron-based communication between processors. Photonic interconnects eliminate that bottleneck. For the hyperscalers investing hundreds of billions in AI compute, the savings from switching to photonic infrastructure at scale are not marginal. They are structural.

But the promise of photonic chips has long outpaced the ability to manufacture them cost-effectively. Cnuic’s technology directly addresses that gap — and does so with a platform that is inherently reconfigurable, meaning it can serve not just photonic chips but a broader range of light-based applications: metalenses, 3D photonic crystals, AR and VR waveguides, and beyond.

As co-founder Omar Durrani puts it: “Every major leap in human capability has come from learning to use a new medium better. We learned to use electrons. Now we are learning to use light. Cnuic is building the tools that make that possible at scale.”

Our Perspective

Tensor Ventures has conviction in deep tech companies that address structural constraints in industries approaching inflection points. The semiconductor industry is at one of those inflection points now — and the manufacturing layer is where the leverage sits.

What drew us to Cnuic is the nature of the breakthrough itself. This is not a software layer on top of existing hardware, or a process optimization within the current paradigm. It is a new category of manufacturing tool, protected by novel intellectual property, developed by a team with deep roots in photonics research.

As Ondřej Lipold, partner at Tensor Ventures, notes: “Cnuic’s technology can democratize the production of photonic chips in much the same way that PCs democratized computing power.” Martin Drdúl, co-founder of Tensor Ventures, who oversaw the investment alongside Ondřej, adds: “From a deep tech perspective, this is a completely new technology and a major breakthrough that could mean a whole new role for Europe in the semiconductor industry.”

Scotland has quietly built one of Europe’s strongest photonics research ecosystems, centered on Edinburgh and Glasgow. Cnuic is a product of that ecosystem — and potentially its most consequential commercial expression yet.

Music is one of the most social things humans do — and one of the least social experiences on the internet.

You press play. You listen. Maybe you share a link. That’s roughly where it ends. In a decade defined by creator economies, interactive media, and participatory culture, music has remained stubbornly passive. No other major consumer category looks quite like this: you can remix a video, annotate a document, co-author a story — but the song you love stays locked behind a stream.

GRAI, a music AI company founded by Ilya Liasun, Andrei Avsievich, and Dzmitry Kamarouski, is building the infrastructure for what comes next. The company has just raised a $9M seed round co-led by Khosla Ventures and Inovo.vc, with participation from Tensor Ventures, Tiny Supercomputer Investment Company, Flyer One Ventures, the a16z Scout Fund, and angels Andrew Zhai, Greg Tkachenko, Rob Reid, and Dima Shvets.

Not Generation, Participation

The current AI music conversation has largely been about creation: give the model a prompt, get a song. Suno, Udio, and others have demonstrated that generating plausible music at scale is now technically tractable. But GRAI’s founding insight is that most people don’t actually want to generate music from scratch. They want to participate with music they already love — to remix a favorite track, change its style for a joke, play with a song the way you might edit a meme.

The distinction matters enormously. Creation is solitary; participation is social. And the social layer around music — the layer that drives discovery, cultural spread, fandom — is almost entirely underdeveloped compared to what AI now makes possible.

GRAI’s current products reflect this thesis directly. Music with Friends (iOS) lets users remix trending tracks collaboratively. Remix Playground (Android) offers an open-ended AI music sandbox. Both are experiments in consumer behavior as much as they are products — ways to understand how Gen Z and Gen Alpha users actually want to interact with music before culture has named the behavior.

Infrastructure Built for Artists, Not Around Them

What separates GRAI from a feature is the infrastructure layer underneath it. The company has built its own taste and participation graph, a real-time audio system, and a derivatives pipeline — technical architecture designed to transform tracks while preserving the identity of the original. Remixes are identifiably connected to their source, not detached from it.

This matters for the business model as much as the product. GRAI’s commercial thesis is that remixed derivatives can generate a new royalty stream for artists and labels — making participation economically aligned with the rights holders, not adversarial to them. Rather than deploying first and seeking forgiveness later, the company is approaching labels proactively, building opt-in/opt-out controls directly into the platform architecture.

The team brings relevant credibility here. Liasun, Avsievich, and Kamarouski previously built Vochi, a video creation app that was acquired by Pinterest — a company that understands participatory media at scale. They know what it takes to build creative tools that go viral, and they know how to negotiate the rights infrastructure that makes that legally sustainable.

Why Tensor Ventures Invested

At Tensor Ventures, we look for founding teams with rare combinations: deep technical capability, genuine product intuition, and an ability to operate at the intersection of consumer behavior and complex stakeholder ecosystems. GRAI has all three.

The music industry is one of the most technically interesting and commercially difficult markets in consumer technology. The IP landscape is Byzantine, the incumbents are powerful, and the history of startups building on licensed content without permission is not encouraging. GRAI’s approach — building the rights infrastructure first, centering artist control, and treating labels as partners rather than obstacles — is the only viable path to building something durable here.

We also believe the underlying bet is right. Music is a massive surface area for AI-driven social interaction that has barely been touched. If GRAI can establish what “participating in music” means — the verb that doesn’t yet exist — they won’t just have a product. They’ll have defined a category.

We’re proud to be part of this round alongside Khosla Ventures, Inovo.vc, and an exceptional angel syndicate. Congratulations to Ilya, Andrei, and Dzmitry — this is where the next era of music begins.

Deploying one AI agent is a product decision. Managing hundreds of them is an organizational one — and most companies aren’t equipped for it. That’s the gap Pointee was built to close, and it’s why Tensor Ventures is doubling down.

Pointee has closed a $1.4M Seed round led by Tensor Ventures, with Seed Starter České spořitelny and Seed Starter Slovenskej sporiteľne participating. The investment will go toward expanding the platform, growing the commercial team, and advancing Pointee’s core Guardian Agent technology — developed in collaboration with Microsoft in San Francisco.

The Problem with AI at Scale

The enterprise AI conversation has largely focused on adoption: getting agents deployed, automating invoices, sorting emails, generating reports. That part, it turns out, is the easy part. The harder problem emerges when one agent becomes ten, then fifty, then hundreds — each running autonomously, each consuming compute and budget, each capable of behaving unexpectedly in production.

At that scale, the question is no longer “can we automate this?” It’s “how do we know what our agents are actually doing?”

As Jan Strnad, co-founder of Pointee, puts it: moving from one AI agent to hundreds is like moving from one employee to an entire company — you need management infrastructure. Without it, costs spiral, quality degrades, and the very efficiency gains that justified the investment evaporate.

What Pointee Builds

Pointee is a unified platform for agentic automation — from initial deployment to enterprise-scale orchestration. It brings together workflow automation, process transparency, and AI governance in a single layer, letting teams manage humans and AI agents side by side with full visibility into what’s running, why it’s running, and what it costs.

The platform’s most distinctive capability is Guardian Agents: always-on AI that watches over your other AI. When an agent starts behaving anomalously — degrading in output quality, running up unexpected costs, or posing a security risk — Guardian Agents detect it in real time, trace the root cause, and can automatically execute remediation playbooks without waiting for human intervention. No alert fatigue. No manual firefighting. Self-healing automation infrastructure.

The results among existing customers speak to the platform’s impact: a 12x increase in process completion speed, 47% more cases processed, and 99.7% SLA adherence for critical workflows. Customers include KBC, Raiffeisenbank, UNIQA, Deloitte, Intel, AXA, and Atlas Copco.

Why Tensor Ventures Is Following On

We first backed Pointee when the company’s thesis was about bringing transparency to RPA and workflow automation. That thesis has aged well — and sharpened considerably. The question Pointee is now answering, AI governance at operational scale, is one of the most consequential infrastructure problems in enterprise software today.

The team, founded in Liberec by Ivo Knejp, Jan Strnad, and Martin Vitouš, has consistently built ahead of the market. When everyone was talking about deploying AI, they were already thinking about how to manage it. Their collaboration with Microsoft in San Francisco is a further signal of how seriously the technical community is taking the problem they’ve defined.

This follow-on reflects our conviction in both the problem and the people solving it.

Manufacturing has an invisible tax. It shows up not in the cost of raw materials or energy, but in the hours lost every time a machine stops and no one immediately knows why. By most estimates, up to 80% of unplanned downtime is spent diagnosing a fault — not fixing it. The wrench is ready. The technician is not.

Edmund, an industrial AI company founded in Ostrava in 2023 by Jakub Szlaur, Benjamin Przeczek, and Miroslav Marek, has built a platform to close that gap. The company has just closed a €2.5 million Seed round led by FORWARD.one, with participation from University2Ventures and Tensor Ventures.

The Problem Is Context, Not Data

Modern factories are rich in data and poor in understanding. PLC programs govern how machines behave, service records capture what went wrong before, and technical documentation describes how systems are architected — but none of these speak to each other. When something fails, a skilled engineer must mentally stitch these sources together under pressure, often while production is stopped. That synthesis is the bottleneck.

Edmund’s platform acts as a reasoning layer across all of it. It ingests PLC projects, electrical schematics, maintenance histories, and real-time machine data, and builds a contextualised model of each specific machine. When a fault occurs, technicians can identify its root cause in minutes rather than hours. In controlled deployments, the platform has reduced the diagnostic phase by up to 90% — with one customer, Amcor Flexibles, reporting faster repairs of over 25% and an annual saving of approximately 440 hours of skilled labor.

As Szlaur puts it: “The real challenge isn’t a lack of data — it’s a lack of context. We’re building AI agents that understand how machines actually work, down to the level of the control program.”

Why Now, Why CEE

The urgency is structural. European manufacturing faces a compounding talent problem: tens of thousands of maintenance positions sit unfilled, and roughly one in five current workers is expected to retire within the decade. The knowledge these engineers carry — accumulated over careers spent learning specific machines in specific plants — cannot be easily transferred. Edmund’s thesis is that this expertise can be captured, structured, and made accessible to the next generation of workers before it walks out the door.

Central and Eastern Europe, where Edmund has built its initial commercial base, is an ideal starting point for this kind of product. The region houses some of Europe’s most sophisticated manufacturing operations in automotive, food processing, and heavy industry — sectors where margins are tight, machine complexity is high, and the cost of downtime is acute. CEE is not a fallback geography; it is a proving ground for industrial depth.

Our Perspective

At Tensor Ventures, we back deep tech companies addressing structural problems in industries that have long operated below their technical potential. Industrial maintenance sits squarely in that category. The gap between what modern AI can do and what actually reaches the factory floor is still enormous — not because the technology is unavailable, but because translating machine intelligence into operational context is genuinely hard. Edmund has done that translation work.

What distinguishes the team is their grounding in the specifics of industrial control systems. Building an AI that reasons over PLC logic is not a generative AI wrapper problem — it requires understanding the engineering domain deeply enough to model how failures propagate through real systems. That domain depth, combined with early commercial traction and a clear expansion path into Western Europe and the US, is why we participated in this round.

We’re pleased to be backing Jakub, Benjamin, and Miroslav as they build toward what they describe as a standard technology layer between machine and human in manufacturing — not another chatbot, but infrastructure for how industrial work actually gets done.

🚀 Portfolio in Action

Antiverse closed a $9.3M Series A led by Soulmates Ventures, with participation from Innovation Investment Capital, DOMiNO Ventures, and existing investors DBW, Kadmos Capital, and i&i Biotech Fund — bringing total funding to over $20M. The round will scale Antiverse’s AI-driven antibody design platform and advance its first wholly owned candidates toward later-stage preclinical development by 2027.

ValkaAI: We’re proud to back ValkaAI, which has secured a €12M pre-seed round led by Rockaway Ventures, with participation from J&T Ventures, Tensor Ventures, BD Partners and Fond Naše Česko. ValkaAI is building real-time, fully controlled AI video — turning video from a passive format into an interactive interface. Think AI commentators reacting live to games, personalized broadcasts, and digital characters that evolve with user input.

AppFactor officially joined the Tensor family. We led a $4M seed round alongside Begin Capital, Adara Ventures, and Narwhal Investments in this London-based agentic orchestration platform that autonomously maintains, modernizes, and deploys enterprise software — without disrupting end users. Thanks to Knarik Avanesyan for scouting this gem!

Bioo: Something interesting is growing at the intersection of biotech and space. Our portfolio company Bioo is rethinking how biology can power the future — turning natural processes into usable energy and sustainable infrastructure.

“Time to update the spaceship we all live in.”

👩🏽‍💻 Deep Tech Scene

Karpathy drops microGPT — a full GPT in 243 lines of pure Python. No PyTorch. No TensorFlow. No dependencies. Just math, an autograd engine, a tokenizer, the Adam optimizer, and a training loop. Karpathy called it "an art project" and "the irreducible essence" of what makes LLMs work — the culmination of a decade-long obsession to simplify transformers to their bare minimum. microGPT blog post.

AMD and Meta sign a 6-gigawatt GPU deal. The largest single hardware procurement in AI history. AMD and Meta announced a multi-year, multi-generation partnership to deploy up to 6GW of AMD Instinct GPUs — with the first gigawatt (custom MI450 silicon + 6th Gen EPYC CPUs on AMD Helios architecture) shipping in H2 2026. AMD issued Meta performance-based warrants for up to 160 million shares (~10% of the company), vesting as shipments scale. AMD announcement.

Majorana qubits get their first single-shot readout. A QuTech team published results in Nature demonstrating single-shot parity readout of a minimal Kitaev chain — confirming the protected nature of Majorana qubits with millisecond-scale coherence. Nature paper.

NousResearch just dropped a fully open-source AI agent that lives on your server forever. It’s called Hermes Agent and it’s the personal AI assistant that actually grows the longer it runs. Hermes Agent link.

📣 Events

Ondra Lipold joined Munich Security Conference that was a blast!

dstack is hosting an informal meetup at NVIDIA GTC in San Jose on March 18 together with Makora — bringing together GPU experts, AI researchers, and AI infra builders. Request a spot if you are around!

DYNANIC: The team was in Barcelona for MWC26, meeting key players in intelligent infrastructure.

Sofie Bukovska from Tensor Ventures joined the EUROTRIP Afterhours in San Francisco. Happy to support our friends at European Startup Embassy!

Cheers,

The Tensor VC team

We’re excited to share that our portfolio company Antiverse has raised a $9.3 million Series A, bringing its total funding to over $20 million since inception.

The round was led by Soulmates Ventures with participation from Innovation Investment Capital and existing investors DBW, Kadmos Capital, and i&i Biotech Fund.

Alongside the financing, Antiverse also announced a research agreement with the Cystic Fibrosis Foundation to design novel antibodies targeting the extracellular region of the CFTR protein, one of the most technically challenging targets in cystic fibrosis research.

This is a major step forward, not just for the company, but for AI-driven therapeutic design more broadly.

Rethinking Antibody Discovery with AI

Antibody therapeutics have transformed modern medicine. But some of the most biologically important disease targets, including G-protein coupled receptors (GPCRs) and ion channels, have historically been extremely difficult to drug using conventional discovery methods.

Antiverse is building a different approach.

At its core is a generative AI platform trained over seven years on difficult antibody targets, capable of designing epitope-specific antibodies while optimizing for:

• Binding specificity

• Physicochemical stability

• Developability

• “Humanness” for therapeutic compatibility

But what truly differentiates Antiverse is its lab-in-the-loop model.

Rather than stopping at in silico design, the company integrates:

• Proprietary programmable cell-line engineering

• In-house wet lab validation

• Iterative build–test–learn workflows

This tight integration between computational modelling and experimental validation dramatically accelerates the path from target identification to functional therapeutic-strength antibodies.

A Research Collaboration Targeting CFTR

As part of this announcement, Antiverse entered into a research agreement with the Cystic Fibrosis Foundation to design antibodies against the extracellular region of CFTR (Cystic Fibrosis Transmembrane Conductance Regulator).

Cystic fibrosis is a progressive genetic disease caused by mutations in the CFTR gene, leading to disrupted chloride transport and the buildup of thick mucus that damages lungs and other organs.

The extracellular portion of CFTR has remained notoriously difficult to target using traditional antibody discovery platforms.

Under the agreement:

1. Antiverse will use its AI modelling and optimisation platform to design novel antibodies.

2. Candidates will be screened using its hyper-expressing engineered cell-line systems.

3. Optimized antibodies will be transferred to the Foundation’s lab for testing in natively expressing cell models.

This workflow is designed to accelerate early discovery research and enable faster exploration of emerging therapeutic modalities.

Scaling the Platform, Advancing the Pipeline

The Series A financing will allow Antiverse to:

• Expand its proprietary AI antibody design platform

• Accelerate development of its internal therapeutic pipeline

• Advance lead programs toward in vivo efficacy studies

• Deepen pharmaceutical and foundation collaborations

Antiverse has already secured partnerships with multiple top-20 global pharmaceutical companies. The company is now targeting progression of its first wholly owned candidates into later-stage preclinical development by 2027.

As CEO and Co-Founder Murat Tunaboylu shared:

“Many biologically important targets have remained difficult to treat using conventional antibody discovery methods. This Series A financing enables us to scale our generative antibody design platform, accelerate our internal pipeline, and expand strategic collaborations.”

Why This Matters

Drug discovery for complex membrane proteins, GPCRs, and ion channels has long represented one of the hardest frontiers in biotech.

What makes Antiverse compelling is not simply the use of AI — it is the integration of:

• Large-scale generative modelling

• Target-specific antibody optimisation

• Proprietary cell engineering

• Real experimental validation

This full-stack, AI-native approach creates a new model for therapeutic antibody development, one capable of addressing historically “undruggable” biology.

We’re proud to see Antiverse reaching this milestone and continuing to push the boundaries of computational antibody design.

The next phase, advancing internal programs toward in vivo validation while expanding global partnerships, marks an important transition from platform development to therapeutic impact.

Congratulations to the entire Antiverse team on this achievement!

We are thrilled to share that Tensor Ventures has participated in a €12 million pre-seed investment in ValkaAI, a Prague-based startup pushing the boundaries of real-time interactive video using AI-driven digital avatars. The round was led by Rockaway Ventures with participation from J&T Ventures, Tensor Ventures, BD Partners, Fond Naše Česko, and other strategic investors — making it one of the largest pre-seed financings in Central and Eastern Europe.

ValkaAI is defining a new category of “AI-native entertainment”: synthetic video environments that react in real-time to events, context, and individual viewers, enabling fully controllable digital beings with natural spontaneity and agency.

ValkaAI: Real-Time, Responsive AI Video for the Next Generation of Interactive Content

Today’s video experiences are overwhelmingly linear — users press play and watch. ValkaAI believes we’re on the cusp of a fundamental shift toward interactive, adaptive, and personalized video experiences that feel alive and responsive.

The company’s platform is designed to generate high-fidelity synthetic video content on demand, driven by fully controllable digital avatars that:

• React in real time to live events (e.g., sports or esports action)

• Tailor responses to individual users’ context and behavior

• Support AI commentators and dynamic characters within live or on-demand content

• Scale interactive experiences across audiences and use cases

ValkaAI’s technology tackles one of the hardest problems in AI video: achieving real-time responsiveness with visual quality, interactive depth, and personalized behavior — a combination not yet attainable at scale.

A New Interface for Media and Entertainment

ValkaAI’s vision reframes video from static media into a living interface that audiences can engage with actively — shaping narrative, tone, and personalization as they interact. This new paradigm has clear applicability in:

• Sports and Live Broadcasts — AI commentators and personalities that react to gameplay and viewer input

• Esports and Gaming Content — immersive, adaptive experiences that rival linear streaming

• Interactive Media & Events — digital personas that host, guide, or engage users in real time

• Personalized Entertainment — bespoke interactions tailored to each viewer’s preferences and context

What makes ValkaAI compelling isn’t just what it creates, but how it does so: the platform aims to produce responsive, expressive video content that feels genuinely alive — bridging voice, cognition, motion, and visual dynamics in a unified real-time system.

The Founders: Experience Meets Vision

ValkaAI was founded in 2026 by Vlastimil Venclík and Miloš Lokajíček, leaders with deep experience in technology, AI, and entertainment innovation. Venclík previously built and scaled technology ventures in esports and data infrastructure, bringing global perspective and product leadership to ValkaAI’s mission. Lokajíček drives execution and global scaling across research and product development.

The team spans Prague, San Francisco, and Dubai, and includes researchers with backgrounds at Meta, DeepMind, and other top technical institutions, all focused on advancing real-time interactive video and AI persona technology.

Why We’re Invested

At Tensor Ventures, we focus on foundational deep tech that reshapes how people interact with machines and media. ValkaAI exemplifies this: it’s not merely a tool for content creation, but an entirely new computational interface for media, engagement, and human-computer interaction.

The challenges ValkaAI is tackling — real-time generation, interactive agency, personalized audiovisual response — are core to the next chapter of AI experiences. We believe ValkaAI’s platform can unlock new forms of entertainment, engagement, and narrative that don’t just respond to users’ input, but evolve with them in real time.

We’re excited to support ValkaAI as it builds this future.

🔗 https://valka.ai

We are excited to announce that Tensor Ventures has led a $4M seed investment in AppFactor, an agentic orchestration platform designed to automate the modernization and ongoing maintenance of enterprise software.

Founded in London by Keith Neilson, Gabriel Ciuloaica, and Attila Vajda, AppFactor is redefining how large organizations manage complex application environments — enabling software to be discovered, regenerated, and redeployed with minimal human intervention.

The round includes participation from Begin Capital, Adara Ventures, and Narwhal Investments. The funding will support platform expansion, accelerate go-to-market efforts, and help more enterprises address the growing burden of technical debt.

At its core, AppFactor is driven by a clear objective: free engineering teams from maintenance-heavy workloads so they can focus on innovation.

AppFactor: Autonomous Software Maintenance for the Enterprise Era

Enterprise software environments are becoming increasingly difficult to manage.

Large organizations often operate hundreds, sometimes thousands, of applications across fragmented infrastructure. Over time, these systems accumulate vulnerabilities, outdated dependencies, inefficient architectures, and operational risk.

Maintaining them consumes a significant portion of engineering capacity, slowing modernization efforts and delaying business-critical initiatives.

AppFactor addresses this challenge with an agentic orchestration platform that functions as a coordinated system of specialized AI agents capable of operating across the full technology stack.

The platform can:

• Discover where software is running

• Identify code and infrastructure issues

• Refactor and regenerate applications

• Run system-level tests

• Validate performance

• Deploy updates securely

All while preserving enterprise guardrails through verification workflows and human-in-the-loop oversight.

Rather than acting as a coding assistant inside a developer’s workflow, AppFactor operates at the system level, helping organizations maintain and evolve software environments without disrupting end users.

Why AppFactor?

The industry has largely focused on tools that help developers write code faster. Yet one of the biggest constraints on enterprise progress is not code creation, it is code maintenance.

Bug fixes, dependency updates, vulnerability remediation, cloud migrations, and architectural refactoring collectively consume enormous engineering time.

AppFactor approaches this problem differently.

Its platform maintains broad contextual awareness, including application topology, repositories, dependencies, infrastructure, and runtime behavior. This enables autonomous actions to be executed with the confidence and traceability enterprises require.

Every agentic task generates human-readable success criteria tied to defined objectives, ensuring teams retain visibility into what the system is executing.

One example is vulnerability remediation: when a library exceeds a CVSS threshold or a known CVE is detected, the platform can locate the issue, propose a fix, generate tests, verify outcomes, and progressively roll changes into production.

A New Path from Legacy Systems to Modern Architectures

A particularly compelling capability is autonomous code regeneration, including migrations into modern languages such as Rust, increasingly valued for its memory safety and performance.

Historically, such migrations required extensive specialist resources and long timelines.

AppFactor reduces this barrier by automating large portions of the process while integrating documentation, testing, and deployment workflows.

For enterprises running legacy software or compute-intensive workloads, this creates a viable path toward more secure, performant, and cost-efficient systems.

Importantly, the platform fits directly into established engineering processes. Autonomous changes move through pull requests, approval gates, and code review pipelines, maintaining auditability and operational control.

What’s Their Story?

The AppFactor team brings decades of experience in advanced software engineering and enterprise systems.

Through their work, the founders repeatedly observed a pattern: highly skilled engineers spending the majority of their time maintaining existing applications rather than building new capabilities.

They saw an opportunity to rethink maintenance entirely, not as a manual process to be optimized, but as a function that could become autonomous.

This vision led to the development of a platform designed to deliver continuous modernization, allowing software systems to evolve alongside business needs.

As CEO Keith Neilson explains:

“Enterprises have brilliant engineers spending the majority of their time maintaining the past instead of inventing what’s next. AppFactor changes that by turning software upkeep into an autonomous, closed-loop process.”

How Are They Positioned?

AppFactor enters the market at a moment when enterprises face mounting pressure to modernize.

Accumulated technical debt, often described as a trillion-dollar global challenge, continues to slow AI adoption, cloud transformation, and infrastructure efficiency.

By understanding the full topology of an application environment, AppFactor enables autonomous action while honoring the safeguards mature engineering organizations depend on.

Early enterprise customers are already using the platform to compress modernization timelines and transition toward more scalable operating models.

Why We’re Invested

At Tensor Ventures, we support deep-tech companies building foundational infrastructure for the next generation of computing.

AppFactor aligns strongly with that mission.

Software complexity is increasing, distributed systems are expanding, and organizations require new approaches to remain adaptable. Autonomous maintenance represents a meaningful shift in how enterprise software will be operated in the years ahead.

We believe AppFactor has the potential to help organizations unlock engineering capacity, accelerate modernization, and build more resilient software environments.

And we are proud to support the team as they advance this vision.

🚀 Portfolio in Action

• QurieGen launched a fresh website, hosted a webinar showcasing its AI-powered drug development approach, and kicked off its expansion into New York City and Boston, key hubs for biotech and life sciences partnerships.

• dstack’s published a practical Slurm migration guide to help AI teams scale infrastructure without rewriting existing workflows, a helpful step for teams moving from legacy batch systems to more flexible experimentation and deployment.

• Passengera’s connectivity and data platform is now supporting autonomous shuttle tests on public roads in Hamburg, underscoring that autonomous mobility infrastructure is advancing beyond prototypes into real-world operation.

• Aireen’s AI retinal screening algorithm achieved full MDR Class IIb integration with Optomed’s certified medical hardware, bringing AI-powered diabetic retinopathy screening closer to primary care and mobile programs across Europe.

• Repsense’s information analytics platform reached a significant deployment milestone: it is now operational in a NATO context, validating its robustness and security in high-stakes environments.

• Superface: At Generation AI Paris, Superface introduced ArcPay, the first payments solution built for autonomous AI agents, enabling machine-to-machine transactions, positioning agentic commerce as an emerging infrastructure layer.

• Antiverse: Hiring a Bioinformatician to our Cardiff (UK) HQ. Join their interdisciplinary team on a mission to design antibody-based therapeutics against the targets that the big pharma cannot crack using AI methods and top-tier wet lab techniques.

  More on our portfolio updates:

👩🏽‍💻 Deep tech scene

• Google just ran a fault-tolerant quantum computer that makes fewer mistakes than its own qubits. That sentence used to live in theory papers and conference fantasies. Now it’s in Nature. Quantum error correction officially left the “someday” folder. Chemistry, cryptography, optimization - buckle up.

  → Nature paper
  → Blog by Google DeepMind & Google Quantum AI

• Fusion energy quietly crossed engineering breakeven.
  Not “the plasma was hot” breakeven - full system, lasers, cooling, everything.

  Translation: fusion is no longer a physics flex. It’s now an engineering + scaling headache. Which is exactly how real industries are born.

  → Official announcement from National Ignition Facility
  

• 18 countries just agreed to track who’s training giant AI models and how much compute they burn.
  Signed in Davos, of course. No slowdown. No bans. Just: “Hey, maybe frontier compute is critical infrastructure now.” The era of vibes-based AI governance is ending. Spreadsheets have entered the chat.

  → World Economic Forum announcement
  

💡 What we read

• State of AI Report 2026 - “The Compute Era”:
  The latest State of AI Report 2026 lands with a clear thesis: AI progress is now governed less by model architecture and more by compute access, energy, and geopolitics. The report details how frontier AI development has consolidated around a handful of compute-rich actors, while national strategies increasingly treat GPUs, fabs, and power grids as strategic assets, placing AI squarely in the realm of industrial policy.

• The Real Cost of Frontier AI (Tomasz Tunguz):
  In a widely shared analysis, Tomasz Tunguz breaks down the full-stack economics of frontier AI systems, from chip fabrication and networking to cooling, power, and long-term depreciation. His conclusion: the limiting factor for next-generation AI isn’t talent or ideas, but capital intensity at unprecedented scale, reshaping who can realistically compete at the frontier.

• “The Return of Physics” (Nicolas Colin):
  In a sharp essay, Nicolas Colin argues that technology is exiting a software-dominant phase and re-entering an era where physics, materials, energy, and hardware constraints define progress. From AI to climate tech to biotech, innovation is once again bounded by atoms, not just bits, forcing investors, policymakers, and technologists to rethink timelines, risk, and value creation.

📣 Events we loved

Our first event at Machine House office! 🚀

It was a real pleasure to kick off our first deep tech community event of the year with such a thoughtful, curious, and engaged group. We loved the conversations, the energy in the room, and the genuine connections that started over coffee and snacks.

A special thank you to the team who made it all happen: Sofie Bukovska, Petr Ulvr, and Ondrej Lipold!

Join our next events, subscribe for the notifications: https://luma.com/isaji1ue

Cheers,

The Tensor VC team

1. Passengera: Autonomous Mobility Is Already Here

Autonomous shuttles are no longer a future promise in Europe, they are already on the streets.

In Hamburg, the HOLON autonomous minibus has begun testing, powered in part by technology from Passengera. Their connectivity and data platform enables real-time fleet operations, the kind of infrastructure most passengers never see, but without which autonomous mobility simply cannot function.

This milestone is a reminder that the future of transport isn’t just about vehicles. It’s about software, data, interoperability, and collaboration across European teams that know how to ship complex systems together.

🔗 Press release

2. QurieGen: Expanding AI-Driven Drug Development to the US

QurieGen has been on the move.

The team launched a new website and hosted a webinar showcasing their approach to AI-powered drug development, while also kicking off their US expansion through the ScaleNL program.

New York City and Boston are next 🇺🇸✈️ two of the most important hubs for biotech and life sciences.

This step marks a new chapter in bringing European AI-first drug discovery platforms closer to global pharma, research institutions, and clinical partners.

🔗 Website & webinar
🔗 ScaleNL announcement

3. dstack: Practical Guides for Scaling AI Infrastructure

Many AI teams still rely on Slurm for training and batch jobs, but scaling experimentation and deployment can get painful.

dstack released a concise Slurm migration guide showing how teams can move existing workloads to dstack step by step, without rewriting their entire setup. It maps familiar Slurm concepts to dstack workflows and focuses on practical, day-to-day usage.

🔗 Migration guide

4. Aireen × Optomed: AI Retinal Screening Meets EU Medical Standards

A major healthcare milestone from Aireen.

Optomed US Lumo® is now officially approved for integration with Aireen’s AI algorithm for diabetic retinopathy screening in the EU.

This achievement reflects far more than a partnership announcement. Behind it:

• Deep technical integration

• Full MDR Class IIb certification

• Execution under some of the highest regulatory and clinical standards in Europe

Together, Optomed and Aireen are bringing AI-powered retinal screening into primary care, mobile programs, and healthcare systems across the continent.

Portfolio milestones like this are built slowly, and matter immensely.

🔗 More details

5. Repsense: Deployed Where It Matters Most

Repsense reached a notable deployment milestone: their information analytics platform is now deployed in a NATO context.

This is a strong validation of both technical robustness and operational trust, particularly in environments where reliability, security, and decision speed are non-negotiable.

🔗 Announcement

6. ArcPay by Superface: Payments for Autonomous Agents

At Generation AI Paris, ArcPay was introduced as what it truly is:
the first agentic wallet built for autonomous agent payments.

As protocols from Google, OpenAI, and Coinbase enable agents to discover, negotiate, and transact autonomously, a critical gap has emerged: traditional payment systems are not designed for agents.

ArcPay is tackling this head-on, think Apple Pay, but for AI agents.

The real shift isn’t agents serving humans. It’s agents transacting with other agents:

• Autonomous supply chains

• Agent-to-agent B2B marketplaces

• High-frequency, machine-speed commerce

ArcPay’s beta is now live, with early builders already experimenting at the infrastructure layer of agentic commerce.

🔗 Learn more & sign up
🔗 Presentation

7. Quantagonia: 30 Under 30 Recognition

A special highlight from Quantagonia.

Founder Philipp Hannemann was named to the 30 Under 30 list, a recognition that reflects an exceptional journey. From convincing major sports leagues to adopt optimization software before turning 20, to building Quantagonia (now acquired by Strangeworks), Philipp continues to push decision intelligence forward.

Quantagonia’s work applies AI and quantum computing to planning problems with millions of constraints, making mathematical optimization accessible at unprecedented scale.

🔗 More about the list

8. Solvemed: AI Supercomputing for Medical Research

Solvemed has been granted access to Isambard-AI, one of the world’s most powerful AI supercomputers (ranked 11th globally), powered by over 5,400 NVIDIA Grace Hopper Superchips.

Through the $1.5M EyeWarn project, funded by UK Research and Innovation, this computational power enables pupillary biomarker research that until recently was purely theoretical.

Solvemed’s work has already been featured in BBC News and NewsBytes, and this access unlocks an entirely new research frontier.

🔗 Full story

Looking Ahead

Across our portfolio, a common thread stands out:
deep technical execution in domains where complexity is unavoidable, and impact is real.

From autonomous mobility and medical AI to agentic commerce and quantum-inspired optimization, these teams are building foundational layers that will quietly shape the next decade.

We’re proud to support them, and excited for what’s next 🚀

We’re excited to invite you to an in-person gathering for founders, researchers, and builders shaping the future of deep tech in Prague!

📅 DeepTech Breakfast @ Machine House
Friday, 23 January
9:00–11:00 AM
Machine House, Pernerova 635/57, Prague 8-Karlín

Start your Friday morning with great coffee, open conversations, and hands-on insights — hosted in collaboration with the Tensor Ventures community. We’ll kick things off with a short intro, then open the floor to share best practices around scaling, selling, and fundraising in deep tech — all in a relaxed, informal setting.

Whether you’re a founder, operator, researcher, or investor in AI, quantum, biotech, or adjacent deep tech fields, this meetup is a great space to connect, collaborate, and learn from peers.

👉🏼 Request to join & register here DeepTech Breakfast @ Machine House or just email us back here.

We’d love to see you there — let’s kick off 2026 with meaningful conversation and community momentum in Prague!

Warmly,

Tensor Ventures Team

At Tensor Ventures, 2025 was defined by execution: disciplined investing, tangible outcomes, and growing momentum across the deep tech ecosystem.

Investment Activity

Throughout the year, we deployed capital across our funds with a clear focus on foundational technologies and critical infrastructure. Our investments targeted companies solving hard, systemic problems across AI, quantum computing, and next-generation infrastructure, areas where long-term value creation matters more than short-term hype.

Each new investment reflected our core thesis: backing exceptional technical teams with global ambition, and supporting them through the earliest and most challenging phases of company building.

OpenVibe

Edmund

Wultra

DYNANIC

QurieGen

PangeAI

Repsense

Exit Activity

2025 also marked an important milestone for Tensor Ventures with a successful portfolio exit.

In August, Quantagonia, a German quantum computing software startup, was acquired by Strangeworks, a U.S.-based leader in quantum and high-performance computing software. The acquisition, completed less than two years after our investment in Quantagonia’s €3.3M round in 2023, represents a strong outcome for our fund and validates our early conviction in the company’s technology and team.

This exit is particularly meaningful as it underscores the global relevance of deep-tech innovation emerging from Central Europe. It also reinforces our long-standing focus on quantum technologies, an area where commercial adoption is now accelerating.

Beyond financial return, the acquisition enables Quantagonia’s technology to scale to a broader enterprise audience worldwide, amplifying its real-world impact. While Quantagonia was the headline exit of 2025, we see several other portfolio companies progressing toward similarly transformative outcomes.

Building for the Long Term

As we look ahead, Tensor Ventures enters 2026 with a stronger team, deeper networks, and an increasingly global footprint. Over the past year, we expanded our capabilities to better support founders post-investment, from strategic guidance to international partnerships and follow-on capital.

The achievements of 2025: new investments, portfolio growth, a successful exit, and team expansion; have laid a solid foundation for the years ahead.

We extend our sincere gratitude to our portfolio founders, limited partners, co-investors, partners and the broader deep tech community. Your ambition, resilience, and commitment are turning fundamental research into technologies that address some of the world’s most complex challenges.

As we head into the holiday season and prepare for 2026, we are energized by what lies ahead. Europe’s deep tech ecosystem has never been stronger, and we are proud to support the scientists, engineers, and entrepreneurs building its future.

Wishing you a restful holiday season and a successful year ahead. 🎁

Cheers,

Tensor Team

We are excited to announce that Tensor Ventures invested in DYNANIC, a deep-tech startup building solutions that unlock the full potential of programmable chips for high-performance applications like AI, virtual reality, and next-generation networking.

Founded by Pavol Korček, Lukáš Kekely, and Jan Kořenek as a spin-out from Brno University of Technology, DYNANIC simplifies how programmable chips are used and integrated — making them far more accessible, efficient, and performant across computing environments where speed and flexibility are essential.

The company has secured €550,000 in pre-seed funding from Tensor Ventures alongside a group of experienced angel investors, including industry veterans from Dell, AMD, Intel, and Hewlett Packard Enterprise. This investment marks the first deployment from Tensor’s second deep-tech fund and will accelerate product development, market expansion, and technology maturation.

At its core, DYNANIC is driven by a clear purpose: to democratize access to programmable computing — empowering developers, companies, and platforms to harness flexible hardware without specialized expertise.

DYNANIC: Making Programmable Chips Practical for the AI Era

High-performance computing is evolving fast. Today’s workloads — from AI inference engines to real-time analytics and virtual reality — demand not just raw compute but adaptable compute that can be reconfigured in the field to meet changing requirements.

Programmable chips such as FPGAs (Field-Programmable Gate Arrays) offer this flexibility, but they have historically been hard to program and integrate. This has kept them on the sidelines of many modern applications — despite their performance and efficiency advantages.

DYNANIC is changing that.

By building tools and frameworks that simplify how programmable chips are used, the company makes advanced chip technologies accessible to teams without deep hardware programming expertise, and unlocks performance benefits previously available only to a narrow class of specialists.

Why DYNANIC?

Programmable chips have long been a performance frontier — capable of delivering acceleration, power efficiency, and adaptability. Yet unlocking that value normally requires teams of experts writing low-level code and managing complex toolchains.

This complexity has limited their adoption in emerging compute areas that urgently need them, especially:

• AI workloads where latency and throughput are critical

• High-speed networking such as 100 Gbps and 400 Gbps links

• Edge and data center acceleration where flexibility matters

DYNANIC addresses this challenge by offering abstractions and tooling that make programmable chips as accessible as software — dramatically shortening time to value and reducing reliance on scarce hardware programmers.

What’s Their Story?

DYNANIC was born out of research at Brno University of Technology, where its founders spent decades developing deep expertise in programmable logic and high-speed computing.

Formerly known as BrnoLogic, the team combined academic rigor with a practical mission: make programmable chips usable in real world applications without the prohibitive complexity they’ve traditionally required.

Co-founder Pavol Korček and his team saw programmable chips as a generational opportunity — a way to give systems unparalleled flexibility and performance — if only the barriers to use could be removed.

With this vision, they built a technology stack that automatically optimizes and adapts chip configurations, enabling customers to customize performance without deep FPGA expertise.

A New Approach to Programmable Chips

What makes DYNANIC different is its focus on usability without compromise.

Programmable chips are powerful precisely because they can be reconfigured after production. But that flexibility also makes them hard to work with.

DYNANIC’s approach:

• Abstracts the complexity of FPGA programming into higher-level tooling

• Automates optimization, selecting the best configurations for a given workload

• Boosts performance and throughput, especially for networking and AI-related tasks

• Integrates with existing infrastructure so companies can adopt without rewriting systems.

This combination gives organizations the performance of programmable logic with the accessibility of general-purpose software — a rare and valuable blend.

How Have They Grown?

Since launching, DYNANIC has already demonstrated notable progress.

Its technology has been licensed and adopted by major programmable chip manufacturers — including leading names such as Intel and AMD — and used by companies like Kakao Corp and Kaloom Networks to improve server efficiency and throughput.

With the support of Tensor Ventures and experienced angels, the company is now scaling its product, targeting broader adoption in high-performance computing segments and among global hyperscaler customers.

The investment will enable further refinement of the platform, acceleration of advanced networking technologies (including next-generation 800 Gbps solutions), and expansion into new markets where reconfigurable computing offers clear advantages.

Why We’re Invested

At Tensor Ventures, we invest in deep-tech companies that bring computational innovation into the physical and digital infrastructure of tomorrow.

DYNANIC reflects exactly that mission.

Programmable chips are a cornerstone of future computing — especially as AI workloads push traditional architectures to their limits. But their promise has been constrained by accessibility and tooling gaps.

DYNANIC bridges that gap, making advanced reconfigurable hardware practical for real-world use cases without requiring rare engineering skills.

We believe this technology has the potential to:

• reshape how edge and data-center acceleration is achieved

• unlock performance efficiencies for AI and networking

• broaden the addressable market for programmable hardware

This investment is a step toward a world where programmable computing is no longer a niche but a foundational layer of next-generation systems.

And we’re proud to support DYNANIC on that journey.

Repsense is a European dual-use deep-tech company building advanced data analytics for information tracking, threat detection, and narrative analysis, helping governments, institutions, and businesses understand how information spreads, mutates, and influences real-world outcomes.

The round was joined by Seed Starter ČS, alongside existing investors BSV Ventures and Coinvest Capital, and was completed in tranches throughout 2025.

At its core, Repsense is driven by a clear mission: to give decision-makers visibility and foresight in an information environment that has become a new frontline: across security, policy, and markets.

Repsense: Navigating the Information Frontline

Information has become one of the most powerful, and volatile, assets in the modern world.

Narratives spread faster than institutions can react. Disinformation campaigns shape public opinion. Algorithms amplify content at scale. And decisions in government, defence, finance, and business are increasingly influenced by forces that are hard to observe in real time.

While vast amounts of data exist across social and traditional media, turning that data into actionable intelligence remains difficult.

Repsense is changing that by combining large-scale data ingestion, advanced analytics, and predictive modeling, Repsense helps organizations see what’s emerging, understand its impact, and respond faster.

Why Repsense?

Repsense was founded in 2022 by Mykolas Katkus, a communications and strategic narrative expert, and Alfredas Chmieliauskas, a former engineer at Palantir.

The founding insight was simple but urgent: the information space had become a strategic domain, yet decision-makers lacked tools to observe and understand it at scale.

The team began by building systems capable of mapping information flows across platforms, sources, and formats, treating narratives not as isolated messages, but as evolving structures.

Growth accelerated in 2025 when Liudvikas Andriulis, an experienced technology sales leader, joined the company to scale commercial adoption.

Today, the team operates from Prague and Vilnius, with 25 people working across engineering, data science, and applied research.

A New Approach to Information Intelligence

Repsense’s technology goes far beyond traditional media monitoring.

Its flagship Havel platform enables clients to track, detect, and predict how information spreads, evolves, and influences public perception across both social and traditional media, with a level of granularity that is rarely possible today.

Core capabilities include:

• Predictive information intelligence
  Anticipating how narratives are likely to evolve, escalate, or dissipate, allowing institutions to act before impact materializes, not after.

• Granular opinion tracking
  Measuring how narratives affect specific audiences, public figures, or decision-makers over time, from broad public sentiment down to individual actors such as policymakers or elected officials.

• Narrative & information tracking
  Monitoring how themes, messages, and storylines emerge, spread, and mutate across platforms, regions, and media types.

• Advanced content analysis
  Including short-form algorithmic content and object-based recognition for emerging media formats where influence increasingly concentrates.

• Narrative DNA forensics
  Analyzing how information propagates, recombines, and gains traction, helping identify coordinated activity and structural patterns behind influence campaigns.

Together, these capabilities transform raw media data into decision-grade intelligence, enabling faster, more informed responses in complex information environments.

Wide Applicability Across Sectors

While Repsense has strong roots in security and defence, its applications extend far beyond a single domain.

Key areas include:

Government & Defence
Supporting strategic communications, resilience against disinformation, and situational awareness in the information space. The platform is currently deployed by NATO Strategic Communications Centre of Excellence and multiple European government institutions.

Finance & Banking
Helping institutions understand narrative-driven risk, market sentiment, and emerging reputational threats.

Marketing & Brand Protection
Enabling advanced influencer analysis, reputation safeguarding, and campaign intelligence in an algorithm-driven media landscape.

Across sectors, Repsense allows organizations to move from reactive monitoring to proactive information strategy.

What Makes Repsense Different?

Repsense is not a media monitoring tool.

It is an information intelligence platform.

Instead of focusing on isolated posts or metrics, Repsense analyzes information as a dynamic system: tracking structure, spread, coordination, and impact.

Key differentiators include:

• deep analytical depth rather than surface-level metrics

• predictive capabilities, not just historical reporting

• applicability across security, policy, and commercial use cases

This systems-level approach is what makes Repsense both defensible and scalable.

How Have They Grown?

Repsense has already achieved meaningful traction across both public and commercial sectors.

Its Havel platform is actively used by:

• NATO Strategic Communications Centre of Excellence

• multiple European government institutions

• commercial clients operating globally

At the same time, the company continues to expand its technical capabilities, particularly in short-form content analysis and object-based recognition for emerging media formats.

The successful completion of the €2 million seed round marks an important step in scaling both product development and international adoption.

Why We’re Invested

At Tensor Ventures, we invest in deep-tech companies operating at the intersection of technology, systems, and real-world impact, especially where new forms of infrastructure are emerging.

Repsense sits squarely at this intersection.

Information is now a strategic asset and a strategic risk. Repsense gives organizations the tools to understand and defend that space, while also unlocking new possibilities for commercial insight and market intelligence.

As Ondřej Lipold, Partner at Tensor Ventures, puts it:

“The information environment has become a frontline — for governments, businesses, and institutions alike. Repsense is building capabilities that will matter far beyond today’s information challenges.”

We believe Repsense is building foundational technology for the next era of information intelligence.

And we’re proud to support them as they scale.

Founded by Johanna von der Leyen and Marek Miltner, PangeAI is building AI agents that make complex geospatial analysis as simple as typing a prompt, turning spatial intelligence from a specialist-only capability into an instant, decision-ready tool for businesses and public institutions.

The company has raised over $1 million in pre-seed funding from a strong transatlantic syndicate, including Miton, Plug and Play, RTP Global, Stanford-affiliated investors, and Tensor Ventures, alongside a group of strategic angel investors across Europe and the United States.

At its core, PangeAI is driven by a powerful ambition: to make the physical world as searchable and understandable as the digital one, unlocking faster decisions, safer infrastructure, and more sustainable investments.

PangeAI: Making the Physical World Searchable

Much of today’s most important decision-making is grounded in the physical world — land, infrastructure, climate, and risk. Yet understanding this world remains surprisingly difficult.

The data exists, but it is locked behind complex geospatial tools, fragmented datasets, and specialist-driven workflows that take weeks to produce answers.

PangeAI is changing that.

By combining frontier AI with geospatial intelligence, PangeAI makes complex spatial reasoning accessible to anyone, not just GIS experts.

Why PangeAI?

Geospatial intelligence underpins decisions across energy, insurance, infrastructure, finance, and climate. Yet despite its importance, the way organizations work with spatial data has barely evolved.

Today, answering questions like where to build, where risk is increasing, or where nature can be restored typically requires hiring GIS experts, assembling datasets manually, running custom analyses, and translating technical outputs into management reports. It is expensive, slow, and poorly aligned with real decision timelines.

PangeAI removes this bottleneck.

Its AI agents automatically identify relevant datasets, select the right analytical methods, run spatial reasoning, and return clear, decision-ready outputs, all from a natural-language query. No GIS expertise required.

This represents a shift from consultant-driven geospatial projects to instant, AI-native spatial intelligence.

What’s Their Story?

PangeAI was founded by Johanna von der Leyen and Marek Miltner, who share an academic background at Cambridge and later reunited at Stanford University.

Both founders encountered the same structural problem from different perspectives.

Johanna worked on environmental and climate challenges from a business and policy perspective, including several years at Stanford and McKinsey. Across infrastructure, energy, and climate-risk projects, she repeatedly saw how critical decisions were slowed by inaccessible and expert-dependent geospatial analysis.

Marek brings deep technical expertise, with more than eight years of experience building AI-first products and multiple startup exits. His research at Stanford focused on the intersection of artificial intelligence, sustainable energy, and geospatial data, using spatial intelligence to optimize real-world systems.

Together, they set out to rebuild geospatial intelligence from first principles, not as better GIS software, but as an AI system designed around how decisions are actually made.

A New Approach to Geospatial Intelligence

Core product pillars include:

• Plain-Language Spatial Queries
  Users don’t need technical GIS skills. Just ask a question in everyday language and get actionable GIS insights.

• Autonomous AI Agents
  Behind the scenes, intelligent agents decide which datasets and analytical methods to apply, delivering results in minutes rather than months.

• Automated Data Integration
  The platform automatically ingests internal and external data sources, removing the traditional bottleneck of data preparation.

The outcome is spatial analysis that feels as simple as a search engine query, but grounded in real, complex earth data.

Wide Applicability Across Industries

PangeAI’s capabilities aren’t limited to one use case. According to their site, the product delivers the most value wherever decisions rely on land, risk, and environmental context.

Some of the most compelling areas include:

• Energy & Infrastructure
  Automated siting, risk screening, feasibility assessments, and monitoring workflows, reducing uncertainty on high-stakes projects.

• Natural Capital & Climate
  Tracking land-use changes, identifying restoration opportunities, and supporting climate-aligned investment decisions.

• Insurance & Financial Services
  Rapid hazard scoring, portfolio exposure mapping, and accumulation analysis that traditionally required whole data teams.

Across industries, PangeAI accelerates insights while lowering the cost of analysis, often replacing entire consultant engagements with an AI-driven workflow.

What Makes PangeAI Different?

PangeAI is not a mapping tool and not a traditional GIS platform.

It functions as a digital geospatial expert.

Its system of AI agents can:

• automatically ingest and combine geospatial and internal datasets

• select the appropriate analytical techniques based on the question

• run simulations and spatial reasoning

• generate maps, summaries, and reports

All in response to a plain-language prompt written by a business analyst, investor, or executive.

Rather than optimizing for layers, tools, or interfaces, PangeAI optimizes for clarity, speed, and action.

This agentic approach fundamentally changes how organizations interact with spatial data, shifting the focus from tooling and expertise to insight and decision-making.

How Have They Grown?

Although still early, PangeAI has already demonstrated strong market traction.

Within months of launching, the company has signed multiple paying enterprise customers, validating both the urgency of the problem and the demand for AI-native geospatial intelligence. Early adoption spans organizations for whom spatial data is mission-critical, and where speed, scale, and decision-readiness matter most.

The company initially developed with support from academic grants and quickly transitioned into commercial deployments, a pace that reflects how acutely outdated existing geospatial workflows have become.

Why We’re Invested

At Tensor Ventures, we invest in deep-tech companies that connect advanced computation with the physical world, especially where legacy industries are constrained by outdated tools and slow workflows.

PangeAI sits at exactly this intersection.

They are tackling one of the world’s largest and most under-digitized data categories, and doing so with a product that replaces manual expertise with autonomous reasoning at scale.

From our first conversation, we were convinced by the founders’ depth, clarity of vision, and ambition to redefine an entire category.

PangeAI is not just building a new product.
They are creating a new way of working with the planet.

And we’re proud to support them.

🚀 Portfolio in Action

• BEIT Breakthrough: BEIT published a major advancement in quantum-powered molecular simulation: a new algorithm that models full molecular motion (rovibrations) with up to 100,000× lower quantum-volume requirements than previous methods. A foundational leap toward practical quantum chemistry.

• Wultra’s Expansion: Wultra opened its first Asian office in Singapore to meet rising demand for post-quantum security. The expansion lined up with the Singapore FinTech Festival, where the team met new regional partners.
  (Their quantum-resistant authentication tech now protects 70+ banks across 25 countries.)

• DYNANIC Milestone: DYNANIC was named one of Deloitte’s Companies to Watch in the Technology Fast 50 Central Europe, spotlighting their momentum in SmartNIC, FPGA, and network-performance innovation. The team hints that something big is coming next.

• dstack’s Updates: Co-hosted a full house AI & Open-Source Models meetup in San Francisco during Open Source AI Week, uniting GPU specialists, AI researchers, and ML engineers for lightning talks and networking. On the product side, dstack rolled out a beta integration for Kubernetes, enabling its open-source orchestrator to natively manage GPU workloads on Kubernetes clusters – a big step toward easier multi-cloud AI deployments.

• Antiverse Expansion: Building momentum in South Korea. Our portfolio company Antiverse spent a packed week meeting leading biotech and pharma teams in South Korea, where demand for AI-powered antibody discovery and GPCR-focused cell line technologies is accelerating rapidly. With strong interest from one of Asia’s most dynamic innovation hubs, Antiverse is exploring new collaborations across the region.

  

• Veracity Protocol Technology: Joined forces with global materials leader Avery Dennison to scale up its microstructure fingerprinting technology. Selected for Avery Dennison’s ADStretch accelerator, Veracity is working to make Physical AI and digital product passports a new standard of trust for physical objects. (This partnership validates Veracity’s vision of using one-image Vision AI fingerprints to authenticate products without special tags or chips.)

• Edmund AI Hiring: Following its €500K pre-seed investment earlier this year, Edmund’s AI co-pilot for manufacturing is now deployed in 20+ pilot projects with industry heavyweights like Festo and Vitesco. The platform’s AI assistants help factory teams troubleshoot breakdowns in minutes instead of hours. Edmund has been expanding its team (hiring a Customer Success Engineer in September) and gearing up for a broader European rollout to tackle the skilled technician shortage in factories.

(Other Tensor portfolio notes: Superface continues to refine its connectivity platform for AI agents and strengthening the partnerships in the US, earlier this year it secured a $600K seed extension to build “AI-to-API” integrations. QurieGen is heads-down developing its single-cell “Google Maps for cells” techbio platform after raising €2.2M in pre-seed funding to accelerate oncology drug discovery with their new Chief Business Officer to strengthen sales and partnerships.)

👩🏽‍💻 Deep tech scene

• Google’s Quantum Breakthrough: Google unveiled a new algorithm called Quantum Echoes that runs on its latest quantum chip and outperformed classical supercomputers by 13,000×. The algorithm, detailed in Nature, is a major step toward practical quantum computing, with potential to aid drug discovery and materials science by generating unique data beyond classical reach.

• $12B Bet on RNA Therapeutics: Pharma giant Novartis agreed to acquire U.S. biotech Avidity Biosciences for about $12 billion in cash. Avidity’s platform delivers RNA therapies to muscle tissue, and Novartis’s hefty purchase (at a 46% premium) expands its pipeline of treatments for rare neuromuscular disorders. This marks one of 2025’s largest deep tech-driven biotech acquisitions, underscoring big pharma’s appetite for breakthrough RNA medicines.

• Exascale Supercomputer in Europe: Europe officially entered the exascale era, JUPITER, a new supercomputer in Jülich, Germany, surpassed 1 ExaFLOP/s (one quintillion operations per second) in November. JUPITER now ranks as the world’s 4th fastest computer and the most energy-efficient exascale system. With ~24,000 NVIDIA Grace Hopper chips, it opens unprecedented capacity for AI model training and climate or biomedical simulations at ultra-high resolution.

(Other notable news: The U.S. approved its first CRISPR-based therapy for cancer, a “warm” fusion breakthrough set a new record for net energy, and the UK unveiled a £100M fund to back quantum startups, deep tech is charging ahead globally.)

💡 What we read

• State of AI Report 2025 “AI Went Industrial”: The latest State of AI report dropped, and its key message is striking: Today’s biggest AI hurdles aren’t algorithms, they’re social and material. From data and compute bottlenecks to adoption challenges, this annual 300+ page analysis maps how AI’s frontier is shifting from pure research breakthroughs to scaling infrastructure and aligning with human context.

• OpenAI’s $1 Trillion Cloud Spend: VC analyst Tomasz Tunguz revealed jaw-dropping numbers on OpenAI’s ambitions. By his estimates, OpenAI has committed $1.09 trillion in cloud and hardware spending through 2035 across partners like Broadcom, Nvidia, Microsoft, AWS, and others.

• “Deep Tech Rising” (N. Colin): In a thought-provoking essay, Nicolas Colin argues that the shift to deep tech isn’t just another trend, it’s a return to venture’s original purpose: funding fundamental breakthroughs that create lasting competitive advantage. Deep tech flips the script by moving risk from market uncertainty to technological achievement.

📣 Events we loved

Petr Ulvr at “Czech AI: From Strategy to Execution”
Petr joined a panel on how the Czech Republic can turn its AI ambitions into real-world impact. He highlighted why deep tech is essential for economic growth and technological resilience, and how Tensor Ventures supports ambitious founders from the earliest stages.

Ondra and Jan at Slush 2025 (Helsinki)
In November, our team headed north to Helsinki for Slush, one of the world’s largest gatherings for founders, investors, and operators. With 70,000+ people on the ground and deep tech at the forefront, the energy was unmistakable.

Daniel Hastík at UP Business Camp 2025
Daniel spoke at UP Business Camp about what truly matters in investor meetings beyond technology and business models, and how we identify exceptional founders and support their growth at Tensor Ventures.

We also visited Truesdays, How to Web, Engaged Conf, and many more.
See you soon at Demo Day, Central Asia Venture Conference 2025, Deep Tech Momentum, and beyond! 🚀

Cheers,

The Tensor VC team

🚀 Portfolio in Action

HeartBeat.bio AG

We’re proud to see our portfolio company HeartBeat.bio AG join forces with Boehringer Ingelheim to advance gene therapy treatments for inherited heart muscle disorders.

With its Cardioid Drug Discovery Platform, HeartBeat.bio AG is pioneering how human cardiac organoids can model genetic cardiomyopathies, accelerating the path to transformative therapies for patients facing limited options.

This collaboration marks a significant step toward addressing one of the most pressing unmet needs in cardiovascular health. Congratulations to the HeartBeat.bio team!

Edmund is hiring. Refer a Customer Success Engineer and help Edmund to build the future of industrial AI by building an AI-powered co-pilot for manufacturing teams.

AI Infrastructure & Open-Source Models. In a few weeks, we co-hosting the AI infrastructure & open-source models meetup in San Francisco on 21st Oct with our portfolio startup, dstack.ai. Bringing together AI researchers, GPU experts, and open-source enthusiasts for lightning talks.

👩🏽‍💻 Deep tech scene

Oxford Quantum Circuits Expands to the U.S. with Hybrid Quantum-AI Data Center:
OQC is deploying its Genesis processor in New York alongside NVIDIA GPU clusters, blending quantum and AI infrastructure for financial and scientific workloads.

Sanofi Injects $625M into Biotech and Neuroimmune Deep Tech Ventures:
Sanofi Ventures received a fresh $625M allocation to fund startups in digital health, neurology, and immunology, signaling renewed big-pharma commitment to deep science.

“AI Is Outpacing Infrastructure” Survey Reveals Growing Gaps:
A study from Keysight shows that while 90% of organizations plan to scale AI infra, 59% cite cost and capacity constraints. The rapid evolution of models is straining compute, memory, networking, and storage systems.

💡 What we read

Agents vs Clouds. This post contrasts autonomous “agent” architectures with centralized “cloud” models, arguing that agents shift control toward edge or client-side systems while clouds centralize orchestration.

SaaS Valuation Multiples: 2015-2025. Aventis Advisors presents a data-driven analysis showing how SaaS valuation multiples expanded pre-2021 and have since contracted, with current public comps trading near ~6× EV/Revenue.

Why the Business Context Layer Is the Key to Making AI Work in Enterprise. The article proposes that to reliably deploy AI in enterprises, we need a “Business Context Layer” that codifies process rules, context, and exceptions so AI agents can act meaningfully within organizations.

📣 Events we loved

AI Tinkerers. Jan Faflík joined Jakub Krikava on stage for an inspiring talk at AI Tinkerers Prague. We also supported VC Drinks by AI Tinkerers, connecting early-stage founders, entrepreneurs, and VCs at Opero. 🎙️

Forbes Next Big Thing. Roman Smola, Martin Drdul, and Petr Ulvr joined discussions confirming that deep tech is moving from the margins to the mainstream of Czech investing.

Founder’s Cookbook Academy. Together with Startup Kitchen and Pražský inovační institut, Ondřej Lipold kicked off this year’s Founder’s Cookbook Academy with the first workshop, sharing tactical insights on crafting investor-ready pitches.

Catch us at SF Open-Source Week, Tech Happy Hours, Slush, and, of course, somewhere on planet Earth at other events 🌎

Cheers,

The Tensor VC team

🚀 Portfolio in Action

1. Quantagonia successfully exited to a U.S. company, marking another milestone for Tensor’s quantum investments.

   

2. Aireen secured a new investment round, fueling the next stage of its growth in digital health.

3. Sifted featured BEIT’s CEO & Co-founder Paulina Mazurek among 8 female founders to watch in quantum, highlighting another strong voice in our portfolio.

4. Edmund continues building momentum in AI-driven infrastructure, setting new standards for efficiency and scale.

5. QuireGen is pushing forward with its mission to revolutionize drug discovery.

6. Meanwhile, dstack’s hosting a hands-on developer event in SF, showcasing open-source tools for AI infra. Join here.

   👉 For more updates, follow us on LinkedIn.

👩🏽‍💻 Deep tech scene

• IBM & AMD Collaborate on Quantum-Centric Supercomputing

  IBM and AMD announced a joint initiative to build next-generation computing architectures that merge quantum computers with high-performance classical computing and AI accelerators. The goal is to enable hybrid systems that can tackle problems beyond either paradigm alone.

• QuamCore Raises $26M to Pursue 1-Million-Qubit Quantum Ambition

  QuamCore secured a Series A raise of $26 million to work on a 1-million-qubit quantum computer in a single cryostat. It’s an audacious scaling goal and signals strong investor confidence in quantum hardware advances.

• Top Biotech Deals in August 2025: Small Molecules, Antibodies & Royalty Models.

  The biotech space had a busy month. Big pharma made several key acquisitions and partnerships:

  • AbbVie acquired Gilgamesh Pharmaceuticals for ~$1.2B to take control of a novel psychedelic therapy.

  • Xoma Royalty and others struck royalty interest deals, and Interius BioTherapeutics was acquired by Kite to bolster its CAR-T pipeline.
    The trends show that small molecules, RNA, and creative royalty financing are very much in play

• ChronicleBio’s Push on Neuroimmune Tech
  ChronicleBio (a tech-bio startup) announced its mission to accelerate insight into neuroimmune disorders, combining deep tech and biomedical science expertise. It’s emblematic of the growing intersection between tech, AI, and biology.

• Fujitsu Eyes +10,000-Qubit Superconducting Quantum Computer
  Fujitsu announced R&D work toward a superconducting quantum computer exceeding 10,000 qubits, targeting fiscal 2030 to have an operational system. This includes a “STAR architecture” for early fault tolerance and is part of Japan’s push toward industrialized quantum.

📣 Events / On Our Radar

Petr Ulvr joined Jan Kavalírek and Robert K. on the panel “Czech AI: From Strategy to Execution” to discuss how our country can transform AI ambitions into real-world impact.

Meanwhile, Jan was at the AI side chat by Tech Europe in Berlin 🙌🏼

👋🏼 Catch us at Prague.bio conference 2025 this week - and at How to web and Bits & Pretzels next month.

Cheers,

The Tensor VC team

🚀 Portfolio in Action

1. OpenVibe took the stage in New York City at the Threads API Summit, showcasing how they’re redefining social and content aggregation.

2. QurieGen was selected among the Top 100 most promising AI startups of 2025 and is competing in the RAISE AI Startup Competition - a strong recognition of their groundbreaking work in AI-driven drug discovery.

3. Pointee launched a fresh new webinar series diving into product and data insights. Join on 28th August at 4pm CET.

4. dstack was featured in Nebius’ Q2 report, highlighting its role in shaping the future of AI infrastructure.

5. More on our LinkedIn follow us there!

👩🏽‍💻 Deep tech scene

• Anthropic Releases Claude for Financial Services

  On July 15, Anthropic launched Claude for Financial Services, a specialized AI solution tailored for financial analysts, portfolio managers, and underwriters. It offers advanced tools for market research, due diligence, and data-driven decision-making.

• AI Revolution: OpenAI Launches ChatGPT Agent
  On July 17, OpenAI unveiled ChatGPT Agent - transforming ChatGPT from a passive chatbot into a proactive assistant that can autonomously browse the web, manage files, and use applications. This marks a key leap toward agentic AI tools.

• Biotech Boom: MapLight Therapeutics Raises $372M
  July’s biggest private biotech round came from MapLight Therapeutics, which secured $372.5 million in Series D funding. The investment supports its lead program for schizophrenia and Alzheimer’s disease psychosis, positioning neurological therapeutics as a compelling VC focus.

• Robotics Milestone: Intel Spins Out RealSense with $50M
  Around mid-July, Intel spun out RealSense - its AI and robotics division - into a standalone company with a $50 million Series A raise. RealSense will continue to supply its depth-sensing tech to humanoids and autonomous mobile robots (AMRs), highlighting a renewed push in intelligent automation.

• Microsoft Edge Unveils “Copilot Mode”

  On July 29, Microsoft rolled out an experimental “Copilot Mode” in its Edge browser. This AI-driven feature aims to streamline browsing by predicting user intent, summarizing content, reducing tab clutter, and guiding interactions.

💡 What we read

• A VC’s Framework for Innovation
  Stanford professor Ilya Strebulaev shares a visual framework on how venture capital and innovation cycles evolve. The graphic makes clear where capital flows, where returns concentrate, and why timing matters more than we think. View the graphic ➝

• Yes, Ads Are Coming to AI
  Strange Loop Canon argues that ads in AI products are inevitable - and perhaps even desirable. The piece dissects economic incentives, user expectations, and how monetization may reshape the AI landscape. Read more ➝

• Levered Beta Is All You Need
  Ethan Ding revisits the role of levered beta in modern finance, showing how this simple concept underpins risk-adjusted returns. A sharp and technical dive that bridges academic theory with practical portfolio strategy. Dive in ➝

• Is Defense Tech Backable?
  Nicola Sinclair explores whether defense tech startups can attract venture funding. Beyond ethics, the piece lays out the structural barriers and why some investors are leaning back in. Read the essay ➝

• Contrary’s Research Rundown
  A sweeping roundup covering markets, startups, and technology from Contrary Research. The July digest highlights emerging founders, sector shifts, and under-the-radar insights worth tracking. Browse the rundown ➝

• AI Market Intelligence
  Artificial Analysis digs into the data behind AI funding, adoption, and product launches - mapping where the money flows in 2025 and which categories are rising fastest. See the analysis ➝

📣 Events On Our Radar

• TechBBQ – Aug 27–28 | Copenhagen, Denmark 🇩🇰
  Scandinavia’s flagship startup summit, where founders, VCs, and global innovators connect.

• TechVillage – Sept 5–7 | Butuceni, Moldova 🇲🇩
  A nature-immersed retreat with 50 curated startups and investors—dealmaking done differently.

• Bits & Pretzels – Sept 29–Oct 1 | Munich, Germany 🇩🇪
  Founders and investors gather during Oktoberfest for Europe’s most vibrant founder festival.

• How to Web – Oct 1–2 | Bucharest, Romania 🇷🇴
  Eastern Europe’s premier tech & VC gathering, drawing 3,000+ innovators and investors.

• Istanbul Slush’D – Oct 3–4 | Istanbul, Türkiye 🇹🇷
  Slush goes global—live pitches and cross-continental VC networking at the Bosphorus.

• Tech Week – Oct 6–12 | San Francisco, USA 🇺🇸
  A city-wide spotlight on AI & robotics with 1,000+ decentralized tech events.

• Web Summit – Nov 10–13 | Lisbon, Portugal 🇵🇹
  The biggest global stage for startups, investors, and thought leaders—70,000+ attendees.

• Slush – Nov 19–20 | Helsinki, Finland 🇫🇮
  The ultimate founder-first gathering: 5,000+ startups, global investors, and dealmaking at scale.

Cheers,

The Tensor VC team