Vibe coding, an AI-powered model where software developers—or anyone—use natural language prompts instead of writing code, is no longer a Silicon Valley meme. By late 2025, it had become one of the fastest ways to build software, slashing costs and accelerating development.
Coined by ex-OpenAI researcher Andrej Karpathy in February 2025, the term describes a high-level approach where AI generates entire applications from simple prompts. It slashed early startup burn rates by as much as 85–95%, enabled weekend-built Micro-SaaS tools to reach paying customers within days, and made the idea of the $1 million solo founder feel attainable rather than mythical.
Across Africa, where access to capital and engineering talent has historically constrained innovation, the appeal is obvious: ideas can now move from concept to product at unprecedented speed.
But speed has a cost. And that cost is increasingly showing up in the form of fragile systems, leaked data, and broken trust.
This is the challenge that Cencori, a Nigerian AI startup positioning itself as the “Cloudflare for AI,” says it wants to address.
From AI optimism to infrastructure reality
Cencori began in June 2025 not as a business idea, but as a concern. Co-founder and CTO, Bola Roy Banjo, traces its origins to his early work as Design Manager at FohnAI, a digital security platform, working at the intersection of artificial intelligence (AI) and cybersecurity in 2024.
“The evolution of AI is fast, and we can’t just trust its decisions blindly,” Banjo says. “I was trying to infuse cybersecurity directly into AI systems, to secure the entire landscape. That led to Cencori.”
What emerged was a thesis: as AI products proliferate, they will need an infrastructure layer like the internet needed Amazon Web Services and Cloudflare.
In Banjo’s view, future AI applications should ship with ethics, security, and reliability built in by default, not bolted on after something breaks.
“AI products are going to come with other problems, ethics, security, and data leaks,” he explains. “With Cencori, those things are in-built. You don’t have to worry about your AI leaking sensitive data or exposing user information.”
The hidden risks of vibe coding
Vibe coding, Banjo argues, is powerful but dangerous. Unlike no-code tools, which rely on visual editors and constrained workflows, vibe coding allows AI to generate entire codebases from prompts. Developers often accept outputs wholesale, pushing them directly into production.
That speed has consequences.
“There are many AI products out there right now, built this way,” Banjo says. “And we’ve already seen cases where vibe-coded apps exposed locations, phone numbers, and addresses to the public internet.”
These aren’t theoretical risks. Data leaks involving large language models happen constantly, as models inadvertently surface private or sensitive information embedded in prompts, logs, or training data. A notable example is the DeepSeek Database Exposure (January 2025), in which over one million lines of log data were leaked and the EchoLeak 0-Click Attack (December 2025), which allowed an attacker to send a specially crafted email with a “hidden prompt”. Even OpenAI warns users not to share personal information, acknowledging that Large Language Models (LLMs) can leak data.
Cencori’s approach is pragmatic rather than philosophical. Instead of trying to “solve AI ethics” in the abstract, the platform focuses on known failure modes. Developers configure what data must be protected, including emails, phone numbers, internal records, and Cencori enforces those constraints at the infrastructure level.
“We already know the kinds of data LLMs leak,” Banjo says. “So we built tools to stop that, based on real use cases.”
The Cloudflare for AI production
If Banjo provides the technical vision, co-founder and Chief Operating Officer, Oreofe Ojurereoluwa Daniel supplies the framing. He describes Cencori as “the Cloudflare for AI production.”
Just as Cloudflare sits between websites and the Internet, handling security, reliability, and traffic routing, Cencori positions itself between AI applications and the models they depend on.
“Most AI builders are focused on getting something working,” Daniel says. “But what happens when users start paying? How do you guarantee uptime, security, and reliability?”
Cencori answers that by acting as a middleware layer. One of its core features is automatic failover across major AI providers. If OpenAI experiences downtime, requests are rerouted to Anthropic or Gemini without developers’ intervention.
“It’s like a power switch,” Daniel explains. “If electricity from the grid goes off, you switch to the generator. Our system does that automatically for AI.”
This redundancy enables what Cencori claims is 99.9% uptime for applications built on its platform, an essential requirement as AI products move from experiments to businesses.
Early traction, real stakes
Though still largely operating in stealth, Cencori is already gaining traction.
Its platform is embedded in the codebases of three Y Combinator–backed startups: Sonarly, an AI-powered bug detection platform; 1uI, a generative UI platform for building AI-native interfaces; and Laurence, an AI “digital brain” for advertising that uses quantitative models to automate Amazon ad spend.
Collectively, Daniel claims these deployments process more than 20,000 requests each week. In addition, around ten solo developers use Cencori for side projects, some of which the founders hope will grow into full-scale companies.
The stakes are high. As Daniel points out, poorly secured AI platforms can cause irreversible harm. He cites cases where private images and personal data escaped into mainstream media, situations that could have been prevented with proper infrastructure controls.
“With vibe coding, these incidents will only increase,” he says. “We built Cencori to plug those holes before they become disasters.”
Building for Africa, competing globally
“We understand the African landscape,” Daniel says. “Cost matters. Complexity matters. We built Cencori with that in mind.”
By consolidating security, observability, reliability, and cost management into a single platform, Cencori aims to reduce both financial and cognitive overhead for builders.
Integration takes under 20 minutes, according to the co-founders, an important detail in a world where speed defines adoption.
The company is currently bootstrapped, funded by personal savings and support from friends and family, but is in conversations with investors to expand its infrastructure roadmap.
Cencori is betting that the next phase of AI innovation won’t be won by those who move fastest alone, but by those who make speed safe. In a world awash with AI-generated code, the companies that endure may be the ones built not just on vibes but on solid foundations beneath them.
