I was lucky to have a fireside chat with Guy Podjarny yesterday evening at our GenAI London Meetup. It made me wonder if I could take inspiration from our discussion of Vibe-Coding and apply it to drafting a blog post—perhaps ‘Vibe-Scribing’ might be the best term for this.
The blog text is vibe-scribed by Gemini 2.0 Advanced (with some light human editing) from a transcript of our talk.
The pace of change driven by AI right now is simply breathtaking. It feels like the ground is constantly shifting beneath our feet, especially in software development. Yesterday’s best practices feel increasingly inadequate, making way for what Guy Podjarny calls AI-native software development.
I’ve known Guy for nearly a decade, we invested in the Series A of Snyk and more recently the Seed of his new company, Tessl. We dove deep into how AI isn’t just tweaking the edges but demanding a fundamental rethink of how software gets built.
For fellow investors and the developer community trying to make sense of it all, here are 5 key takeaways from our conversation that really stuck with me:
1. Beyond Automation: AI Demands a Workflow Revolution
Much of the current AI adoption is merely “sustaining” — applying new tech to old workflows. Think automating code snippets or sharpening video footage. That’s useful, but the true potential, the “transformative” leap, comes from rethinking the entire workflow from first principles, assuming AI’s capabilities exist. Like cloud-native wasn’t just elastic VMs but new practices, AI-native requires us to challenge how we build. Guy used Synthesia as an example — text-to-video isn’t just better video editing; it’s a whole new way to create. It forces the question: Are we just automating tasks, or are we ready to rebuild the factory?
2. Navigating the Chaos: Mapping the AI Dev Tool Gold Rush
Anyone exploring AI dev tools knows it’s chaos out there. Guy aptly described it as “messy” — hundreds of tools, often sounding the same, with ill-defined scopes and low differentiation. Developers are understandably jumping between tools, lacking loyalty because few deliver fully on their promises. To help navigate this, Guy and the Tessl team launched the AI Native Dev Tool Landscape. It’s an impressive, open-source effort on GitHub mapping ~170+ tools (and crucially, taking community submissions). It’s a much needed effort and can serve as an up-to-date resource. I encourage everyone to check it out and contribute.
3. The Future is Spec-Centric (So Perhaps Code Really Could Be Dead?)
The most provocative idea Guy shared was the shift from “code-centric” to “spec-centric” development. His argument is compelling: code loses the intent behind decisions over time, creating fragility and tech debt. We discussed the hilarious “More Realistic” flight simulator sketch — a perfect analogy for blindly trusting AI “magic” without defining guardrails. The AI-native approach uses LLMs’ strengths by defining what needs to be built (the spec) at a higher abstraction level, clearly delineating human-owned decisions (the must-haves) from areas where LLMs can adapt and manage complexity. It suggests a future where code becomes an implementation detail, not the source of truth.
4. Developers Evolve, Not Disappear (Calling all Architects & Product Minds!)
So, is AI coming for developers’ jobs? Guy’s view, which I share, is no — but the role is fundamentally changing. As the act of writing code becomes less central, developers move towards:
- Architecture: Making critical trade-offs, applying taste, designing complex, novel systems.
- Product Focus: Deeply understanding user needs and defining what the system should achieve (the spec!). The value shifts from line-by-line implementation to strategic design and user empathy. The challenge remains to keep the joy and creativity in building, even as the tools change. Your craft is becoming more strategic.
5. The $1B Solo Startup? A Compelling Myth.
Finally, we’ve all seen the incredible efficiency metrics – startups hitting huge ARR with tiny teams. Our own portfolio company Stackblitz scaled its Bolt.new product to a $30M run rate in less than three months and did it with extraordinary efficiency — approaching $2M revenue run rate per employee. It inevitably leads to talk of the $1B one-person company. Guy offered a strong counterargument. Yes, one person might soon create $1B in value thanks to AI leverage. But capturing that as $1B in revenue is unlikely. Why? Because if one person can do it, others can too. Supply rises to meet demand, and what was once magical quickly becomes the expected baseline. We joked that to get in front of a deluge of competition the startup might have to get to the Billion dollar scale in less than a day in the future. Extreme efficiency might exist in fleeting moments, but market dynamics prevail. To close Guy said “Startups should anchor toward the future–try to imagine what will happen in five years’ time.” It’s a reminder to focus on durable value creation, not just seductive efficiency narratives.
This AI transition is profound. It’s pushing us to question assumptions we’ve held for decades about how software is built and valued. Embracing the AI-native mindset feels essential for anyone building or investing in the next generation of technology.
You can follow my thoughts and GV’s, and definitely keep an eye on Guy Podjarny and Tessl.
And seriously, check out the AI Native Dev Tool Landscape.