AI / Human Authorship Note

This text was human generated (ie, not written by AI). It was captured on the 13th of December 2025 using speech to text but should be mostly free of basic errors.

Other pages in this repository may be the product of AI, ideation, or outputs based upon the questions posed here.

Statement Of Purpose

Agentic code generation is becoming an increasingly viable and important part of modern technology development. In fact, it is increasingly regarded as an indispensable tool. While some industries may be forced to be more reticent than others about their adoption, those that completely refuse to use AI code development face a powerful disincentive in the form of decreased code velocity relative to competition.

As AI assisted development emerges more prominently into the mainstream, a new type of technology professional is fast emerging: The expert operator of agentic tools for software development.

The purpose of this repository is to gather together some notes and thinking about how this new type of technologist can best position themselves for long term success in the job market.

The repository is open source in case it might spur thinking among others.

The motivation in my case is my own career planning: I write this from a personal perspective. I've been using AI for code generation intensively for the past year (and more). I regard mastering these tools as a long term project rather than as a temporary productivity boost. Hence, I'm doing some thinking into how to do that!

This repository will be a collaborative process between a human (me) and AI.

What Do You Call A Career "Vibe Coder"?

As mentioned:

The purpose of this repository is to generate some ideas and thinking into how a professional who finds themselves enjoying and excelling in the challenge of leveraging AI systems for technology implementations can best define themselves and grow in a job market that, until very recently, did not consider agentic AI to be a real capability.

There are a multitude of important facets to explore in this regard.

At the more foundational level is the question of how professionals intent on leveraging this skill set as part of their career might self define. Self definition is important because it provides a bedrock for professional identity and skill set generation. It's also important because from these definitions spring job titles and keywords which are ultimately mechanisms to match talent to demand. Hence it is suggested that there is more to self definition than simply coming up with a good description for a resume.

It is suggested as well that the process of self definition necessitates collective thinking.

Although there isn't a shortage of tech conferences that don't achieve much other than keeping conference venues in steady use, I remain a big believer in the value of networking and knowledge exchange.

I believe as well that in order for these processes to work efficiently, "we" need a better name than vibe coding. A thoughtful and mature process of standardizing nomenclature and skills will, I predict, be enormously important and valuable. As vibe coding has stuck as a point of reference, however, it is used here.

To the extent that "vibe coding" may be dismissed as a hobbyist pursuit, I bookmark the argument that that can be said about any field. We all cook. Few of us become professional chefs. The difference usually lies in how committed and passionate and deeply one wishes to master this particular skill set.

This process of standardizing or thinking about the required skill set, especially in a field as fast moving and rapidly evolving as this, is key to strategic positioning in the industry.

The extremely fast pace of change in AI poses a significant challenge. Question of what will happen to those whose skill sets are eroded by AI is frequently discussed, including in the technology context.

A separate challenge is that of finding ways to leverage AI that aren't self deprecating.

To retain a useful skill set in this regard, one has to learn technology that isn't rapidly deprecated. This requires identifying a curriculum that can be periodically updated, but which isn't so short term that by the time one learns the course, it's already irrelevant.

This new type of professional poses a challenge to the conventional prisms through which we conceive of technology workers:

• This person may not be a "coder" in the sense that the code generation they do is significantly offloaded to AI agents. These professionals may find themselves pushed out of discourses by entrenched skepticism from "real" developers
• This person may not be "just" a "product person" or creative: the workflows they touch upon may be deeply integrated into development workflows. They may ideate features and contribute to their development even if later QA is done by a different team member. Redirecting talent in these directions risks missing skills by applying outmoded frameworks onto emerging ones.

Furthermore: "AI assisted code generation" (only one potential name) is emerging as a very real and specific set of competencies drawing upon broad but specific knowledge of AI tools:

• Understanding the advantages and disadvantages of specific models for workflows. The availability of code specific models is rapidly accelerating. Today's playing field, which may consist of a few monolithic providers of large models, may mushroom into a much more complex and fragmented field with specialist models for various workloads. This in turn might draw upon skills like running evaluations, reviewing benchmarks and applying critical faculties to the claims made by labs.
• Beyond understanding the capabilities of models, deploying agentic AI also requires understanding various accessory components that are essential to ensure their successful operation. These include things like guardrails, recovery mechanisms, prompt management systems. memory modules, vector databases and RAG. Orchestrating these tools into efficient workflows requires not only selecting the best components, but selecting the best framework to bind them together. Open source poses something of a challenge in this regard. Sifting has become a key skill.
• Keeping pace with rapidly emerging best practices regarding AI assisted development.
• Keeping pace with the also rapidly emerging field of AI security to ensure that code generation tools are leveraged responsibly and safely within organizations.
• Understanding emerging best practices in quality assurance to ensure that the rapidly accelerated development workflows made possible with AI agents can be leveraged without impeding upon code quality.

Reducing this complex skill set to "vibe coding," in my opinion, does a gigantic disservice to the skillset that has to be developed and nurtured.

Bringing the right vibe to a project rarely determines much about its success. Loose prompting and insufficient task structuring is a recipe for broken code bases, disaster, and unproductive hours.

AI agents, especially those used for code generation, work in rather regimented and structured patterns that defy the idea of simply iterating from a feeling to a viable technology implementation. These best practices are being standardized in vendor implementations. This shared process of knowledge aggregation suggests that the process of maximizing the utility of code generation agents is a recognizable discipline.

---

Questions (For AI Think Tank!)

How Will Vibe Coders Define Themselves?

Leveraging complex AI tools is an important skill set. Increasingly, it plays a pivotal role in development processes. It's also very new. Until recently, very few companies that developed code leveraged AI deeply or in an everyday manner.

Vibe coding has come to media prominence through discourse on X. But I feel that it does a disservice to what is actually a useful skill set that requires the application of skill knowledge, training, perseverance, and learning.

How do you think those who wish to forge their career in this area should define themselves and their skill sets for the purpose of presenting themselves and presenting themselves to the job market and other technology professionals?

At a collective level, how might those involved in this field organize? It's common for people in a specific job function to convene physically or virtually in the form of conferences and networking events. Industry associations or other manifestations of this process of collective organization. How do you think this might happen with agentic development professionals?

Vibe Coder - Developer Relations

My personal view of the integration of AI development into professional development practice is this:

I do not foresee a future in which AI tools will eliminate the need for skilled developers. I do believe that those expert at leveraging AI tools in development workflows will be different professionals from those who are involved in manual code editing and generation. Therefore, I believe that collaboration and working together will be a feature of the workforce of tomorrow.

The question remains how these two groups of professionals will relate to one another and work collaboratively in a way that maximizes the contribution of the other.

We should also consider potential gatekeeping and friction. Those who identify as real developers may exhibit dismissive attitudes towards those who come from the AI world or non-developer backgrounds. At an interpersonal and organizational level, how do you think these dynamics could be handled diplomatically?

The AI Development Team

As AI tools become increasingly prominent in development workflows, there is a need for a variety of different skill sets among those who understand both technology and AI tools.

Professionals might be involved in aspects like code security, code quality review, as well as model procurement and selection. There may initially be a tendency towards generalist AI experts, but over time this may break down into specific specializations.

Let's start from the premise that there is a AI Assisted development generalist working at a small organization like a startup. Let's then imagine that this startup grows into a medium sized company. The workload of the AI generalist becomes unsustainable. The generalists may be involved with hiring of the team defining functions. Let's sketch out the initial and then long term evolution of this team. As the team expands, we might have to consider not only sub specialists but their reporting teams.

The Facets

AI code generation tools initially suffered from severe usability issues owing to context flooding. Large language models were never designed for code generation tasks. The significant amount of context that they are exposed to in the form of code bases made their operation challenging. Degraded inference may take the form of the AI tool, fixing some bugs or adding some features only to destroy others. The experience for developers was extremely frustrating.

Solving the underlying challenge of context management is an important undertaking. Models capable of handling vastly larger context windows may take time to come to market.

In the meantime, the industry and vendors and tools have responded with a temporary fix of sorts. The commonality of this tooling is that it aims to provide more granular task execution by agents. This has taken the form of planning modules and planning modes and even planning models. The planning model is responsible for coming up with a plan of execution and defining a task list, which can then be delegated to subagents with their own context windows.

This bifurcation raises a question of whether the pool of human operators required to operate these tools may itself specialize along these lines. Planning and execution are conceptually and fundamentally different tasks. Do you think the scope is sufficiently wide that we might see a sub specialization? How do you predict this challenge will evolve?

Compensation, Demand

What type of compensation might be realistic for those with this skill set at the moment? What type of salary trajectory might be possible over the course of a career?

How can these professionals avoid the peril of underemployment and under compensation? How can these professionals ensure that they are working for organizations that value and are similarly invested in this function? Or looked at from the negative framing, how can these individuals avoid companies that view this as a low value and hopefully quickly replaceable skill set? How can these individuals assure alignment with companies?

Where do we see demand for this skill set at the moment? Where might we see it tomorrow? How might we see differences between industries and size of organizations and even geography?

The Skills

Supervising or operating AI systems for code generation or editing processes involves skills. These skill sets range from directly understanding the fundamentals of AI operation through to having a good basis and understanding in the technology that is being worked upon.

And this requires knowledge in some different areas. What do we see as the core skill sets in this regard? If we were to devise a structured curriculum for somebody looking to get into this field, what might it look like? What would be the foundational elements and what might be the more advanced topics?

Deprecation Avoidance

The frontier of artificial intelligence is extremely fast moving. Many professionals worry that their jobs will be taken away by AI.

For those who want to professionally work with AI and view this as a core part of their career plan, there are other potential challenges.

If AI moves so quickly, what can one do to ensure that the knowledge one develops and is developing will be slightly ahead of the curve and useful for the foreseeable future? In other words, how can one develop skills in this area and avoid rapid self deprecation? How can one assess if a specific technology stack is a good long term knowledge investment rather than a passing trend?

Certification

What do you think about certification in this regard?

What certifications that exist currently might be useful for those working in this field looking to validate their skill set? What certifications that might not exist would be valuable if they were to exist?

Focus specifically on certifications that involve a reasonable expenditure of effort and time to achieve a certification that is not extremely easy to pass. Focus is well on certifications from established providers.