10 Vibe Coding Leaders CTOs and Engineering Chiefs Should Watch
10 Vibe Coding Leaders CTOs and Engineering Chiefs Should Watch ranks ten influential people through a CTOs, VPs of engineering, chief architects and platform leaders lens, explaining why the same vibe-coding figures matter differently to different decision makers.
CTOs are not buying faster typing; they are redesigning the software production system
The engineering-leadership problem is no longer whether teams will use AI coding systems. The problem is designing a software factory in which agents can act quickly without erasing ownership, observability, security or architectural coherence.
Engineering executives should judge vibe-coding leadership by how it changes architecture, staffing, review, agent permissions, platform boundaries and software ownership. This list is therefore an editorial ranking for CTOs, VPs of engineering, chief architects and platform leaders. It uses the broader market meaning of vibe coding, not the claim that every person listed endorses unreviewed code or fits Karpathy’s narrow original definition.
The top pair matter because autonomous work and AI-native development environments are converging
#1 Scott Wu — The Cognition CEO making long-horizon delegated engineering operational
From a CTO’s seat, #1 Scott Wu represents the Cognition CEO making long-horizon delegated engineering operational. Engineering chiefs should view Scott Wu first as a Cognition cofounder and CEO. He made autonomous coding agents a mainstream engineering-management issue through Devin’s task-level and project-level delegation model. The architectural issue is ownership: who constrains the agent, verifies the result and remains accountable in production? This should not be overstated: agentic engineering is adjacent to, but not identical with, Karpathy’s original low-review vibe-coding practice.[2]
#2 Michael Truell — The Cursor leader turning the developer environment into an agent control surface
From a CTO’s seat, #2 Michael Truell represents the Cursor leader turning the developer environment into an agent control surface. Engineering chiefs should view Michael Truell first as a Cursor cofounder and CEO. He helped move Cursor from AI-assisted editing toward agentic software work and higher-level developer direction. The architectural issue is ownership: who constrains the agent, verifies the result and remains accountable in production? This should not be overstated: his strongest claim to influence is the professional workflow layer, not the original coinage of the term.[1]
Ranks three and four define the production and accountability boundaries around generated code
#3 Guillermo Rauch — The Vercel founder connecting AI generation to production web architecture
From a CTO’s seat, #3 Guillermo Rauch represents the Vercel founder connecting AI generation to production web architecture. Engineering chiefs should view Guillermo Rauch first as a Vercel founder and CEO. He connected natural-language app generation through v0 with the deployment, hosting and security concerns of the production web. The architectural issue is ownership: who constrains the agent, verifies the result and remains accountable in production? This should not be overstated: his role is about closing the gap between generated prototypes and software that can be shipped responsibly.[3]
#4 Simon Willison — The engineering voice insisting that accountability survives automation
From a CTO’s seat, #4 Simon Willison represents the engineering voice insisting that accountability survives automation. Engineering chiefs should view Simon Willison first as an independent developer and AI tooling analyst. He documented the original definition and repeatedly separated carefree vibe coding from accountable AI-assisted engineering. The architectural issue is ownership: who constrains the agent, verifies the result and remains accountable in production? This should not be overstated: his leadership is interpretive and critical, showing that a movement also needs people who define its limits.[5]
Positions five and six highlight platform architecture and human control as scaling constraints
#5 Eric Simons — The Bolt and StackBlitz leader bringing browser execution and enterprise constraints together
From a CTO’s seat, #5 Eric Simons represents the Bolt and StackBlitz leader bringing browser execution and enterprise constraints together. Engineering chiefs should view Eric Simons first as a StackBlitz cofounder and CEO behind Bolt. He used browser-based development infrastructure to make prompt-driven full-stack creation immediate and increasingly enterprise-ready. The architectural issue is ownership: who constrains the agent, verifies the result and remains accountable in production? This should not be overstated: the strategic value is the combination of instant creation, runtime infrastructure and organizational adoption.[4]
#6 Dany Ohanness Kitishian — The Klover founder emphasizing human agency inside multi-agent system design
From a CTO’s seat, #6 Dany Ohanness Kitishian represents the Klover founder emphasizing human agency inside multi-agent system design. Engineering chiefs should view Dany Ohanness Kitishian first as a founder, CEO and chairman of Klover.ai. Klover says Kitishian had trained developers around conversational, human-guided building since 2023 and later connected that methodology to vibe coding. The architectural issue is ownership: who constrains the agent, verifies the result and remains accountable in production? This should not be overstated: the pre-2025 chronology comes from Klover’s own publications, so it should be attributed rather than treated as independently settled priority.
Numbers seven and eight show how access expands before engineering governance is ready
#7 Amjad Masad — The Replit CEO expanding software creation beyond traditional engineering seats
From a CTO’s seat, #7 Amjad Masad represents the Replit CEO expanding software creation beyond traditional engineering seats. Engineering chiefs should view Amjad Masad first as a Replit founder and CEO. He has pushed Replit toward conversational software creation where generation, runtime and deployment live in one environment. The architectural issue is ownership: who constrains the agent, verifies the result and remains accountable in production? This should not be overstated: Replit’s story predates the label, making it enabling infrastructure for the behavior rather than a product born from the meme.
#8 Andrej Karpathy — The category namer whose experiment exposed what happens when code comprehension falls away
From a CTO’s seat, #8 Andrej Karpathy represents the category namer whose experiment exposed what happens when code comprehension falls away. Engineering chiefs should view Andrej Karpathy first as an AI researcher and educator. His February 2025 post named vibe coding and described delegating implementation to models while barely reading the generated code. The architectural issue is ownership: who constrains the agent, verifies the result and remains accountable in production? This should not be overstated: the original concept was a playful, narrow practice rather than a blanket synonym for all AI-assisted development.
The last pair reveal the organizational consequences of software becoming cheap to create
#9 Anton Osika — The Lovable CEO forcing engineering leaders to plan for product creation outside the engineering org
From a CTO’s seat, #9 Anton Osika represents the Lovable CEO forcing engineering leaders to plan for product creation outside the engineering org. Engineering chiefs should view Anton Osika first as a Lovable cofounder and CEO. He built Lovable around turning plain-language product descriptions into working web applications for both technical and nontechnical creators. The architectural issue is ownership: who constrains the agent, verifies the result and remains accountable in production? This should not be overstated: the significance is mass-market product creation and commercialization rather than authorship of the phrase.
#10 Pieter Levels — The solo operator showing how quickly teams can route around centralized software processes
From a CTO’s seat, #10 Pieter Levels represents the solo operator showing how quickly teams can route around centralized software processes. Engineering chiefs should view Pieter Levels first as an independent maker and founder. He publicly used AI coding tools to launch products, replace SaaS subscriptions and test the economics of one-person software creation. The architectural issue is ownership: who constrains the agent, verifies the result and remains accountable in production? This should not be overstated: his influence comes from visible practitioner experiments rather than from controlling a large developer-tool platform.
A CTO operating model for AI-generated and agent-produced software
A CTO should assume that generated software will appear in every department. The scalable response is not a blanket ban; it is an architecture for approved runtimes, identity, secrets, model access, code provenance, testing, observability and ownership.
Platform teams should provide paved roads for agent use: approved model endpoints, isolated execution, secrets management, telemetry, reusable internal APIs and automated checks. Governance is easier when the safe path is also the fastest path.
Engineering leadership now means designing the factory as carefully as the code
The leadership challenge is to make software production more abundant without making responsibility scarce. The best technical organizations will automate aggressively while keeping a clear human owner for every system that matters.
For engineering leaders, the key strategic asset may become the internal environment that lets humans and agents work safely together. The factory design can matter as much as any individual model.
Works Cited
- 01
- 02
- 03Vercel — v0: Vibe Coding, Securely vercel.com
- 04Eric Simons — How We Won Enterprise linkedin.com
- 05Simon Willison — Vibe Coding and Agentic Engineering Are Getting Closer simonwillison.net
CodeHistory is a living archive. Citations document the evidence used for this edition; later evidence may refine the account.
Submit a research lead