Lately, we’ve been thinking about how quickly venture capital is evolving. Not long ago, investing in tech meant betting on a few brilliant ideas in a largely speculative space. Today, the landscape feels less like a single garden and more like a sprawling country, growing outward in every direction. AI, especially, has cracked the earth open. There’s white space everywhere.
What stood out to us in Martine’s conversation wasn’t just the obvious excitement around AI and infrastructure — it was the shift in how firms operate in response to these growing markets. Venture is starting to look more like engineering: deep specialization, layered platforms, and teams optimized for scale.
In the past, generalists built great portfolios with gut feel, pattern recognition, and good timing. Now? Now entire careers are being built around slices of infrastructure — databases, compilers, even compute economics. If you’re going to serve founders in a world that’s increasingly technical and fast-moving, you can’t just be an excellent judge of character. You have to speak the language of the domain.
Specialization isn't about siloing. It’s about depth, nuance, and scalable support. At Dellecod, we find that when investors or advisors truly understand something — say systems-level design or how a model architecture will play in production — the advice stops being generic and starts being generative. We’ve felt it ourselves when we get feedback from someone who’s built what we’re building, at scale. It changes things.
That same shift is happening with media — maybe unexpectedly. A decade ago, it might have seemed strange for venture capitalists to be so public-facing. But with traditional media getting increasingly skeptical of tech, firms are now becoming publishers in their own right. They’re building their own distribution channels, not just to broadcast their opinions but to support their companies when it matters most.
What we’ve learned is that timing matters. When a major AI release drops, or a security flaw makes headlines, founders aren’t waiting for a weekend blog post. They’re looking for people who are online, tuned in, contributing. It’s not about loud takes — it’s about showing up at the right moment with clarity, proof, and reach.
That’s increasingly tied to talent, too. Recruiting right now feels entirely different from even five years ago. Top engineers, researchers, and operators are choosing startups not only based on product but also on whether they believe the company — or the firm behind it — can tell a coherent story, attract great people, and withstand noise. In most markets, competition is about winning customers. In AI right now, it feels more like winning minds.
It’s a new kind of scarcity.
And with that comes pressure. There’s a temptation to over-index on stories, or to chase markets before they’re ready. Martine’s point about infrastructure is a reminder to step back. The apps will come and go — but the platforms, the developer tools, the compute backbone, the bottlenecks in scale and latency — this is where lasting value sits. Each platform shift creates new needs, new foundations. Mobile and cloud proved that already. AI just makes the case stronger.
It’s also leading to some uncomfortable tensions — like the growing pushback we’re seeing against open source. In 2023, we saw real efforts — even from within the tech world — to close models down, restrict access, and consolidate IP. The reasons are complex, and concerns around AI misuse are valid. But as builders, we believe in open ecosystems because that's where innovation tends to emerge. Monopoly logic may win in the short term — but long-term progress usually comes from community breakthroughs.
You can’t plan for every use of open tools. But you can create structures that encourage stewardship. The moment we treat open source as a liability rather than a foundation is the moment we lose some of what allowed the last wave of breakthroughs to happen — whether it's Linux, TensorFlow, or the modern LLM stack.
We also appreciated Martine’s views on boards and value. A lot of early-stage founders imagine board seats as strategic guidance hubs — but most of the real day-to-day value comes from access, infrastructure, and actual help. The support that matters comes between the meetings: when you’re hiring, when you’re stuck, when you need infrastructure tested or thought partnership around a sticky problem. The best partners show up quietly and consistently.
That model — of distributed teams offering grounded, real-time help — feels right to us. It’s how we try to work, too. Don’t solve problems with meetings when platforms can do it better. Don’t wait for quarterly updates if you can just share logs and start debugging together. The best investors and partners aren’t wise sages up the hill. They’re a Slack message away.
AI may well change what it means to write code or launch a company. But the fundamentals — building a strong team, thinking in decades rather than cycles, choosing quality of work over hype — those things still matter. Maybe even more so now that things are moving faster.
We’ve felt that at Dellecod every time we’ve chosen depth over noise. Every time we chose to sit with ambiguity a little longer, knowing that the best ideas usually reward perseverance, not immediacy. AI feels like that — thrilling, still raw, and filled with questions we haven’t figured out how to ask yet.
But if you believe that value comes from infrastructure, from open systems, and from showing up with conviction when the map is still blurry — then maybe we’re on the right path.
What stood out to us in Martine’s conversation wasn’t just the obvious excitement around AI and infrastructure — it was the shift in how firms operate in response to these growing markets. Venture is starting to look more like engineering: deep specialization, layered platforms, and teams optimized for scale.
In the past, generalists built great portfolios with gut feel, pattern recognition, and good timing. Now? Now entire careers are being built around slices of infrastructure — databases, compilers, even compute economics. If you’re going to serve founders in a world that’s increasingly technical and fast-moving, you can’t just be an excellent judge of character. You have to speak the language of the domain.
Specialization isn't about siloing. It’s about depth, nuance, and scalable support. At Dellecod, we find that when investors or advisors truly understand something — say systems-level design or how a model architecture will play in production — the advice stops being generic and starts being generative. We’ve felt it ourselves when we get feedback from someone who’s built what we’re building, at scale. It changes things.
That same shift is happening with media — maybe unexpectedly. A decade ago, it might have seemed strange for venture capitalists to be so public-facing. But with traditional media getting increasingly skeptical of tech, firms are now becoming publishers in their own right. They’re building their own distribution channels, not just to broadcast their opinions but to support their companies when it matters most.
What we’ve learned is that timing matters. When a major AI release drops, or a security flaw makes headlines, founders aren’t waiting for a weekend blog post. They’re looking for people who are online, tuned in, contributing. It’s not about loud takes — it’s about showing up at the right moment with clarity, proof, and reach.
That’s increasingly tied to talent, too. Recruiting right now feels entirely different from even five years ago. Top engineers, researchers, and operators are choosing startups not only based on product but also on whether they believe the company — or the firm behind it — can tell a coherent story, attract great people, and withstand noise. In most markets, competition is about winning customers. In AI right now, it feels more like winning minds.
It’s a new kind of scarcity.
And with that comes pressure. There’s a temptation to over-index on stories, or to chase markets before they’re ready. Martine’s point about infrastructure is a reminder to step back. The apps will come and go — but the platforms, the developer tools, the compute backbone, the bottlenecks in scale and latency — this is where lasting value sits. Each platform shift creates new needs, new foundations. Mobile and cloud proved that already. AI just makes the case stronger.
It’s also leading to some uncomfortable tensions — like the growing pushback we’re seeing against open source. In 2023, we saw real efforts — even from within the tech world — to close models down, restrict access, and consolidate IP. The reasons are complex, and concerns around AI misuse are valid. But as builders, we believe in open ecosystems because that's where innovation tends to emerge. Monopoly logic may win in the short term — but long-term progress usually comes from community breakthroughs.
You can’t plan for every use of open tools. But you can create structures that encourage stewardship. The moment we treat open source as a liability rather than a foundation is the moment we lose some of what allowed the last wave of breakthroughs to happen — whether it's Linux, TensorFlow, or the modern LLM stack.
We also appreciated Martine’s views on boards and value. A lot of early-stage founders imagine board seats as strategic guidance hubs — but most of the real day-to-day value comes from access, infrastructure, and actual help. The support that matters comes between the meetings: when you’re hiring, when you’re stuck, when you need infrastructure tested or thought partnership around a sticky problem. The best partners show up quietly and consistently.
That model — of distributed teams offering grounded, real-time help — feels right to us. It’s how we try to work, too. Don’t solve problems with meetings when platforms can do it better. Don’t wait for quarterly updates if you can just share logs and start debugging together. The best investors and partners aren’t wise sages up the hill. They’re a Slack message away.
AI may well change what it means to write code or launch a company. But the fundamentals — building a strong team, thinking in decades rather than cycles, choosing quality of work over hype — those things still matter. Maybe even more so now that things are moving faster.
We’ve felt that at Dellecod every time we’ve chosen depth over noise. Every time we chose to sit with ambiguity a little longer, knowing that the best ideas usually reward perseverance, not immediacy. AI feels like that — thrilling, still raw, and filled with questions we haven’t figured out how to ask yet.
But if you believe that value comes from infrastructure, from open systems, and from showing up with conviction when the map is still blurry — then maybe we’re on the right path.