From Project Manager to Vibe Coder: A Personal Experiment

As someone who has spent years delivering complex digital systems across government and enterprise contexts, I've learned to look at new tech trends not just for what they can build, but for what they mean for how societies organise and govern technology. I've been a project manager and a presales engineer for some time now, but building a full-fledged website or app entirely on my own was still new to me—until quite recently, when I started vibe coding. By now I've vibe-coded and released a few products from scratch, each one growing in architectural complexity with every new attempt. The simplest of them all is a purely front-end system that calls APIs directly from the browser, deployed from an open-source setup and bound to a domain name. As functionality grew, I subsequently added cloud database and storage services, along with serverless functions.

Besides creating systems from scratch, I also got to try my hand at maintaining and enhancing some of the applications I built. From time to time I do run into issues—limited context windows, coding agents losing memory of past instructions, or producing lacklustre results. It has been a good learning experience: I've learned to tune my prompts, experiment with different architectural patterns and deployment models, try out different workflows, and validate assumptions about how best to achieve a desirable result with AI.

If anything, I'd say my vibe-coding experience has given me deeper insight into how cloud infrastructure plays a pivotal role in supporting this wave of AI-driven app development, and into which skills remain valid and valuable even as AI codes far better and faster than any of us humans do.

Why the Cloud Will Matter Even More

Let's first talk about why cloud will become even more important.

1. Cloud services allow pay-as-you-use

Many vibe-coded products are personal tools with substantially lower traffic than enterprise software. Without a flexible payment scheme like the pay-as-you-use model offered by cloud services, upfront costs would hinder the deployment of these products.

2. Always available and ready to use

Cloud infrastructure is immediately available when you need it. Someone new to coding doesn't need to go into the details of setting up a server or sorting out network configurations—as long as the app will be served on the internet, cloud infrastructure is always at your fingertips to support deployment.

3. Readily available APIs

Cloud infrastructure is naturally accessible via code, which also makes it easily accessible to coding agents. Beyond basic APIs for using specific resources, most cloud providers also offer infrastructure-as-code so you can easily set up complex architectures. All these code-based functions now come in handy with coding agents.

4. A diverse range of services following standardized protocols

There are plenty of cloud service providers out there, ranging from aspiring start-ups to established tech giants. Depending on a project's budget and needs, one can always choose from any desired provider as long as certain technical standards are enforced. Standards such as REST API, SSH, web sockets, and RBAC models form the basis of an open and diverse tech community, and ensure that it remains a fair competition. With GenAI, MCP has increasingly become the standard for agents to communicate with external services; the Skills framework allows repeatable processes to be framed into plug-ins and used across agent platforms. It is with such standards that we welcome this wave of enthusiastic involvement from developers and firms.

Taken together, these features of cloud infrastructure do more than lower the barrier to software creation—they fundamentally shift where power lies in the digital ecosystem. When a policy officer can deploy an app without going through a lengthy procurement cycle, the gatekeeping function of traditional IT departments and vendors is weakened. This is as much a governance shift as a technological one.

From Procurement to Prompting: When a Policy Officer Can Code

With the support of cloud and AI, one need not have extensive knowledge about infrastructure or budget and existing set-ups to launch their own products. All you need is motivation; AI is capable of handling the rest.

Let's take this further by considering the following scenario: a government agency wants to procure app development services.

• How it's usually done: justify needs, put up a budget, call for tender, wait for the app to be developed (typically on the order of months or even years), and eventually deploy and go live.
• How it can be done now: using AI and cloud services, an adequately trained government official could translate policy intentions into a functional MVP and even a complete product. Recently, an official in Beijing reportedly did just that—he vibe-coded a hazard warning app using 1 billion tokens (source: https://news.bjd.com.cn/2026/07/13/11862189.shtml)

  

  .

Beyond Coding: The Four Skills That Matter More Than Ever

What is interesting about this example is how GenAI has reshaped our definition of important digital skills in the new era. We are not talking about getting rid of coding entirely, but rather a rethink of the relative importance of these digital skills. What are some of these skills that remain important—if not more important—despite GenAI's increasing capability in writing code?

1. Analytical thinking

The ability to identify gaps, formulate a problem statement, and create a vision for a product—all belong to the category of analytical thinking. With GenAI, it matters just the same. You need to express a clear intention about what needs to be done: the clearer the intention, the closer the outcome will be to your vision. Analytical thinking is what lies between ambiguous needs and clear task prompts. The good news is that with GenAI, one need not perform that analysis entirely alone. New techniques such as AI-assisted discussion can always be employed to help you frame it better. But at the end of the day, you will have to be the one making the final call, so knowing what you want—and the ability to decipher the good from the mediocre—still takes a great deal of analysis on your end.

2. Architecture and systems thinking

Knowledge about software architecture becomes ever more valuable. You can create all you want, but as applications grow in size and complexity, good architectural design will be key to performance and user experience. It requires you to take a step back and think about your system holistically, and sometimes to make necessary compromises in design.

3. Digital governance

Governance is a broad term, and what I'm specifically referring to here is the ability to control and manage digital applications and infrastructure within legal, geopolitical, and sovereign frameworks. This includes a wide range of things: how and where data is stored, cleaned, used, and shared; how access is controlled and maintained; how system security is monitored. With growing geopolitical complexity worldwide and the fast-paced advancement of AI, digital governance becomes ever more relevant.

4. Professional knowledge about business

I loosely group all knowledge about lines of business, processes, and control mechanisms into this category. A lot of the anxiety we see from the field comes at an individual level—about skill mismatch and job losses. Rarely have we thought about how this might impact B2B and enterprise software. Why are enterprise software giants such as SAP not worried? Ultimately, enterprise software offers not just technical functionality—how you click this button and create that table. Enterprise software offers a platform where business processes consolidate into organizational knowledge. And that is the true value that makes enterprise software difficult to replace. If we take this understanding a step further and look at it from an industry level: the digital industry creates real value when these new technologies find real use cases and bring enablers to the entire ecosystem of economic activities. It takes professional knowledge about business to ground AI, put it to real use, and tackle problems that might not have been solvable before.

A Window of Opportunity: What This Means for Global Digital Cooperation

It's a massive undertaking to discuss how our education and skills programmes could keep up with the pace of AI. I'd like to look at this as an opportunity for latecomers to the digital era to think about what AI could potentially offer. It could possibly be a chance to quickly outgrow the stage of building up a base of software engineers and jumpstart new growth opportunities. This cannot take place overnight: a new guide on how we understand this trend, build capabilities as an economy, and a new framework for collaboration and standards to support open-source data, models, and products will need to be in place.

This is precisely where multilateral institutions and established tech giants have a critical role to play. Updating global digital skills frameworks to include AI-assisted development literacy, creating model curricula that governments can adapt, and fostering open-source standards for cloud-based public applications—these are the tangible next steps that translate this vision into practice. The opportunity is clear; the institutional machinery to seize it needs to be set in motion.