Pietro Bezza

💥 We are thrilled to unveil today our new visual identity and website 💥 tl;dr  Same love for product, more love for product founders and a ...toggle. connectventures.co

Europe now has category leaders across the AI stack A new report by Headline has mapped the top 100 AI companies across Europe by each layer of the AI stack. From the frontier labs to intelligent hardware, Europe is now building category leaders at each layer of the stack. What's particularly amazing to see is how quickly things have changed in the last year. There have been multiple $1bn rounds that have gone into the companies that are on this image, some of these companies didn't even exist a year ago. HUGE CONGRATS to those on the top 100 list. @JonUsero @HeadlineVC

@ttunguz go go go

Three years ago, we launched Theory Ventures with a simple premise : AI would reshape how software is built, sold, deployed, & operated. Within that world, we would build a concentrated, thesis-driven firm. The market moved faster than even the most bullish expectations after the ChatGPT moment. Frontier models leapt from delicate demos to production systems. Open source models have become substitutes for enterprise workloads. Inference emerged as the dominant market in AI. Underpinning all of this, AI compresses time. New models are released every 41 days. Companies reach $100m in revenue in record time. We all achieve more faster. In celebration of our anniversary, we wanted to trace that mechanism through the market shifts of the last three years. The first casualty of compressed time is the old language of venture capital. Seed, Series A, Series B categories still exist, but they describe the financial product companies seek rather than rather than company maturity. Venture firms have left the idea of offering a standard financial product to bespoke offerings : seeds range from $1m to $500m in size. Can we really call it all the same thing, anymore? Three years ago, a seed company was often a small team with a product concept & early signs of product-market fit. Today, some seed rounds are larger than IPOs, fueled by great ambition, a supportive VC ecosystem, & the promise of generational scale businesses to be built. Part of this is inflation in private markets. But more of it is time compression : the best companies mature much earlier than software companies did in prior generations. We’ve learned as an ecosystem how to build software companies & AI accelerates product development. Compressed time also redraws the map of where great opportunity lies. When we first launched Theory, most AI conversations centered on models. Remember the debate of whether model companies would be the airlines of the era? Today, inference is becoming the dominant market. The market is segmenting because the workloads & buyer preferences have evolved - very few companies can afford state-of-the-art AI for everyone - & each specialized constraint creates a new infrastructure category. Companies like @sailresearchco are building the systems that operationalize intelligence : serving it cheaply, routing it intelligently, & specializing it around use cases like video, batch, local, agentic, & real-time workloads. Databases followed this path a decade ago. They fragmented into OLTP, OLAP, vector databases, & streaming systems. Those markets have evolved with AI, a pattern we’ve backed through @motherduck & @lancedb , with @omni in the AI analytics layer above them. Inference infrastructure is now specializing the same way. The expense of inference reinvigorates a sedate market that has been controlled by behemoths for a decade : advertising. Every major interface shift, TV, web, mobile, streaming, found its answer to monetizing a massive audience in ads, & AI is no different. AI advertising is emerging as the subsidy for inference costs, letting applications grow usage & revenue together rather than against each other. We wrote about this dynamic when we led @koahlabs ' Series A : native ad formats inside AI conversations are producing click-through rates 4-5x the display baseline, & an agentic app builder can provide inference offset by ads. The same compression closed the gap between closed & open models, cloud models & local models. The conventional narrative holds that frontier closed-source models lead & open source follows. We’ve reached the iPhone 15 moment of AI. Many models are good enough for most work. Running a model locally reduces cost, improves latency, increases control, & minimizes data governance concerns. Enterprises are adopting local & open-source models for sensitive workloads, & frontier capabilities compress toward consumer hardware within a few years. What once required a hyperscaler cluster runs on a laptop just a few quarters later, a shift @ollama brings to millions of developers. The promise of AI is that software will ultimately be more secure : machines that read every line of code, patch faster than attackers move, & never tire. In the meantime, the attack surface is exploding. MCP servers, skills, plug-ins, & coding agents each introduce new entry points, & enterprises are deploying them faster than security teams can review them. Attackers are massively parallel & shrinking necessary response times from months to minutes. Defenses must respond. It’s why we backed @DropzoneAI , whose AI analysts investigate the alert flood no human SOC can keep up with, @Maze_Security , which applies agents to cloud vulnerability triage, & @artemis , securing the new agentic surface itself. The same agentic wave is rewriting operations. ERP & back-office systems have resisted change for decades because the work is unglamorous, the data is messy, & the switching costs are enormous. One CFO we interviewed, when asked about a startup said, “that company has only been around 15 years; they are too immature.” Agents invert that math. Systems that read documents, reconcile records, & execute workflows can attack operations from the inside rather than demanding a rip-&-replace. It’s the thesis behind Doss, rebuilding ERP for teams that move at modern speed, & Backops, applying agents to the back-office work no one wants to do by hand. AI has impacted crypto, another market fueled by data. Prediction markets, stablecoins, micropayments all have an AI infusion to them. Today, crypto companies need to generate revenue & use AI to provide better experiences, which led to our investment @AlliumLabs , the data layer underneath that institutional wave. Recognizing shifts early requires fingers on keyboards, wrestling AI agents into compliance rather than observing it. We built Theory as a technical organization, experimenting with AI across research, sourcing, diligence, portfolio support, & internal operations. Working inside these systems sharpens our understanding of where the stack is breaking & where new workflows are emerging, while deepening our empathy for founders deploying real AI systems inside enterprises. It’s harder than social media says. AI also changes the economics of an investment firm. Over the last decade, venture firms scaled by adding people. AI-native companies are demonstrating that much smaller teams can operate at 10x+ the leverage of prior software generations, & the same dynamic applies to us : since launch, we’ve analyzed 2x the investment opportunities with a team of just 3 investors working alongside a nine-person intelligence organization. None of this works without the team behind it. Theory started three years ago as a handful of people & a thesis. Today we are thirteen strong. We believe this is the structure of a modern venture capital firm : engineers & researchers who build the systems we use every day : agents that map markets, pipelines that surface companies months before they raise, & research infrastructure that lets a small team cover the ground of a firm several times our size. Everyone at @Theoryvc works with the technology we invest in, & that shared fluency shapes every decision we make. The firm we’ve built over three years is itself a product of the thesis : a small team, deeply technical, operating with the leverage AI makes possible. But the real story of these three years is the founders. They compressed decades of company-building into quarters & shipped products that rewrote what enterprises expect from software. The next three years will make these look slow. The most ambitious builders we meet are just getting started, & we can’t wait to see what they do.

Products people love create enduring companies.

@Dessn_ai well done team

And this is the product that had us raise $6M in a week from incredible investors 👇 What we're the most excited about is this new feature, showed at 0:40 called Surfaces, which allows you to pull an existing page and iterate on it. This feature only exists in Dessn, and it's a huge unlock for our users. Right now it's in private beta, available only to members of our Design Partners Program. DM if you're interested in getting early access! @eminimnim @GabriellaHach @k_grajeda

A year ago everyone said LLMs would get commoditized but the opposite happened (Save this). And the reason is something most investors are still underestimating, the model itself was never the moat but the moat is what surrounds it. Sarah Friar laid out the thesis that the agentic layer creates context and memory that compounds with every interaction. Her own Codex instance knows her role, her communication style, her priorities, her family situation because the system has been learning her across hundreds of sessions. That distinction matters enormously for the investment case. A generic model can be replicated by any lab with enough capital, context and memory accumulated over time cannot. The longer an agent operates inside a person's life or a company's workflows, the more it knows that is irreplaceable specific preferences, past decisions, undocumented institutional knowledge, the way a particular CFO thinks about risk versus the way her predecessor did. This is what the enterprise AI race is actually about, not which model scores best on benchmarks which system owns the context graph. Salesforce built an entire agentic memory architecture specifically to solve this. Their work describes the problem precisely, stateless agents that reset at the start of every session fail at enterprise scale because they cannot learn, cannot improve, and cannot be trusted with autonomous long-horizon work. The companies building persistent memory infrastructure episodic memory for past events, semantic memory for accumulated knowledge, procedural memory for learned workflows are building something that becomes more valuable the longer it runs, not less. That compounding dynamic is fundamentally different from every prior generation of enterprise software. A seat license for a CRM tool has the same value on day one as it does on day 1,000 and the software does not know anything new. An agent with deep memory and context integration is worth multiples more on day 1,000 because it has absorbed a thousand days of decisions, corrections, preferences, and outcomes. This is why enterprise CEOs are moving because they can see the switching cost accumulating in real time on the other side of these deployments. The pricing model is changing to match the economics. A16Z and the leading enterprise SaaS analysts have been tracking a shift away from per-seat pricing toward outcome-based models, pay per resolution, per contract closed, per ticket resolved, per line of code shipped. Gartner projects that by 2030, at least 40% of enterprise SaaS spend will shift toward usage, agent, or outcome-based pricing. That is a complete restructuring of the software revenue model.

The best AI-native products give non-engineering teams the power to build, automate, and ship. If you are building a vertical or horizhontal product that pushes this shift, I'd love to talk (DM me)

With AI, EVERYONE is becoming an engineer. Security teams → engineering teams @AikidoSecurity Design teams → engineering teams @Dessn_ai Customer support teams → engineering teams @plainsupport Business intelligence teams → engineering teams @steepapp

Is it the most competitive industry out there? Yes. Do we have a radically different take on it? Also yes. Come see what designing in prod feels like: create.dessn.ai/create

@GabriellaHach @Dessn_ai @betaworks @eminimnim 😍😍😍 More story and thesis here: connectventures.co/opinions/why-w…

So excited to share that @Dessn_ai has raised $6m, led @pietrobezza , with participation from @betaworks , N49P, and a few other amazing partners and angels. @eminimnim and I started the company 2 years ago with one conviction: the future of product development wouldn’t happen in disconnected mockups or recreated environments. It would happen directly in production. Today, Dessn is the only product that enables an entire team to design and prototype directly in prod — visually, collaboratively, and in one click.

I wrote the full investment thesis here: Design in production. Why we invested in Dessn lnkd.in/gjZ7VbWZ

The best product founders are overly technical and bold. And they build products that let users express themselves in the ways they actually want to work. 🚀 Thrilled to announce we led a $6m seed round in @Dessn_ai, backing the awesome @GabriellaHachem @CheemaNim23684

No other design tool can "get a view of your app rendered to use as a starting point" because no other design tool runs your production repo like we do This is pretty spot on from Karri😍

Great insights @soleio and fully aligned with the @Dessn_ai thesis

That line hit me: “Transitioning from Sketch to Figma was a no brainer because all of a sudden we went from working in local files to web based collaboration” People frame the current moment as “designers will code now”. I think the bigger story is simpler. We are quietly going back to local again. We already lived through local pain once. In the Photoshop era, a design file was a thing that lived on your machine. Big files, messy versions, “who has the latest” and collaboration felt like passing a large file from person to person. Even later, in a large company, we used Sketch with a semi cloud setup. Basically: shared storage, a versioning workflow, and rules everyone had to learn. We used Abstract for branching and merging. It worked, but it came with onboarding cost. New designers did not just learn the product, they learned the system. UI kits made it heavier. Consistency depended on process. Sync depended on discipline. Prototyping was also split across extra tools. If you wanted “real”, you learned a separate craft: After Effects, Principle, Origami, ProtoPie, or even React with early @Framer. It was doable, but it was not flowing. It was tool switching. Then Figma happened and it was obvious. Not because it was prettier, because it moved the work into shared space. Collaboration became the default, not an add on. AI coding tools are bringing back the same old friction Now designers are building “coded prototypes” with Claude Code, Cursor, and similar tools. They are powerful, but the workflow pulls you into local reality again: repos, env vars, local DBs, running servers, PRs, deployment, and “it works on my machine” That is what the report calls “we’re back in local space” And I agree. The problem is not capability. The problem is location. ⎯⎯⎯⎯⎯ Why I keep reaching for Figma Make? My current workflow at @diffusionhq is simple: we design in @Figma, and if needed, I prototype in Figma Make. Not because it's magically better than Cursor or v0. Because the setup cost is almost zero, and the output is easy to share. Click, prompt, iterate, send a link. That matters more than people admit. I mostly use it for one thing: previewing the experience at true scale, in the browser, at 100% zoom, with real interaction. Since we are building a browser tool, that feedback loop is gold. It helps me catch issues early, make decisions faster, and reduce back and forth before handoff. Big prototypes still take time, sure. But the difference is the collaboration stays online, which keeps the team moving. ⎯⎯⎯⎯⎯ The real “next switch”? Photoshop to Sketch was a productivity jump. Sketch to Figma was a collaboration jump. This next jump will be the same type of collaboration leap, but for coded prototypes. This is not “designers can code now”. It is about keeping design work shareable and close to production. The teams that win will not be the ones with the fanciest local setups. They will be the ones who keep making, testing, and reviewing work in the same shared space. That idea is a big part of how we think at Diffusion. A browser based video editor where work stays shared, friction stays low, and iteration stays fast

We’re excited to announce our investment in Mosaico, an open-source, next-generation data management platform built specifically for robotics. connectventures.co/opinions/lets-…

Boom 💥💥 A huge day for @AikidoSecurity as they announce a $60m Series B and for Connect Ventures as we add another 🦄 to our portfolio. I wrote the full story here: connectventures.co/opinions/built…

AI = trajectories. "Fifth, trajectories are the basis for optimizing AI models through reinforcement learning. Smaller specialized models trained on high-value paths replace massive generalists. tomtunguz.com/ai-trajectorie…

@spolu this is awesome

"9 months later but 9 times better" We're releasing a deep-dive agent that has access to your entire company data (structured and unstructured) and all the MCP tools connected to a Dust workspace to perform high level tasks on longer time horizons. It has taken over a number of weekly multi-hours tasks that used to be done by humans including our team.weekly meeting preparation as well as aggregation of user testimonials and feedback from call transcripts. I used it last week during a call with an investor to generate a specific retention graph they were asking for, sharing the results before the call was ended instead of going to our data team. Yes it's 9 months after deep research. But this is definitely 9 times better if your goal is to get things done at work. It has access to all our unstructured data presented as a file-system (notion, slack, drive, ...), all our data-warehouse with discoverability capabilities, and all of the MCP servers connected to our Dust workspace. The team has been using it increasingly shifting from custom agents accomplishing tasks to deep-dive based-agents accomplishing outcomes. 7 different technical initiatives went into enabling deep-dive in Dust. We're diving into the detail of each of them in the blog post below. A great read if you're looking to use or build long-time-horizon agents. blog.dust.tt/building-deep-…

The secret behind Gemini 3? Simple: Improving pre-training & post-training 🤯 Pre-training: Contra the popular belief that scaling is over—which we discussed in our NeurIPS '25 talk with @ilyasut and @quocleix—the team delivered a drastic jump. The delta between 2.5 and 3.0 is as big as we've ever seen. No walls in sight! Post-training: Still a total greenfield. There's lots of room for algorithmic progress and improvement, and 3.0 hasn't been an exception, thanks to our stellar team. Congratulations to the whole team 💙💙💙