At work, many people are being given an initiative to learn AI and innovate with it. One person brilliantly took a simple assessment and made it into her own web app. Another person took the output of several assessments and had Claude merge them into one super assessment.
I thought both were brilliant and wanted to take it another step further … of merging these related ideas, using Agentic Engineering, to build an assessment that merges the assessment concepts of DiSC, Myers-Briggs, Clifton Strengths and Working Genius. There should be some overlap, and Generative AI should find it…
Check out the result: your friendly AI-Robotic PERSONAL BRAND ASSESSMENT
But what I want to harp on a little is something YouTuber Alberta Tech addressed in her video “Do Google engineers actually vibe code?“.
In her video she talks about the differences between vibe coding and agentic engineering.
Agentic Engineering is “Engineering using agentic AI systems.” It asks, “How do I use AI to build systems that can reliably do the work I design it to do, and can I add to that some unit tests and continuous integration?”
Vibe Coding is “Using AI to write code that I might not understand but seems to work.” It asks, “How quickly can I build something useful with AI helping me write the code (that I don’t need to understand, update, maintain or manage by hand)?”
| Aspect | Agentic Engineering | Vibe Coding |
|---|---|---|
| Primary Goal | Build reliable autonomous systems that accomplish business tasks | Rapidly create software by collaborating with AI |
| Success Metric | Task completion, accuracy, reliability, ROI | Speed, creativity, working prototypes |
| Planning | Explicit workflows, orchestration, state management | Minimal upfront planning; iterate as ideas emerge |
| Human Role | Designer, supervisor, evaluator, governor | Co-creator, experimenter, director |
| AI Role | Autonomous worker that can plan and act | Coding assistant that generates code on demand |
| Complexity | Often involves multiple agents, tools, APIs, and memory | Usually focused on a single application or feature |
| Failure Cost | Can affect operations, customers, or business processes | Usually limited to bugs, rework, or technical debt |
| Testing & Validation | Formal evaluation, monitoring, guardrails, observability | Often “run it and see if it works” with lighter testing |
| Similarity: Iterative Development | Iteration is structured and measured against objectives | Iteration is intuitive and driven by exploration |
| Similarity: Heavy AI Usage | AI performs delegated work within engineered constraints | AI accelerates coding through conversational collaboration |
“Agentic Engineer” is one of those new roles that has evolved from the existence of AI. It fits along the same lines as “AI Workflow Architect” and “Human-AI Operations Manager”.
For the past 25 years we had IntelliSense built in a VisualStudio.NET IDE. Some loved it. Others hated it. But over the years from 2001 and 2021 it got better and mind-melded itself as part of coding life. And actually, around 2017, Microsoft had expanded it to IntelliCode.
Vibe coding was coined in 2025 and took off for a few months, but it was effectively … bad.
Bad at unit testing – it would create false positives. Bad maintainability – it would create more technical debt. Bad coding practices – it would cause patchwork code bases (i.e. spaghetti code) . Bad security – it would hardcode secrets and introduce bugs that allowed injection attacks. But most grievous of all it changed the behavior of coders and introduced…
The Erosion of Developer Skills
The over-reliance of AI caused a dumbing down of junior and mid-level software engineers. They would gloss over the hundreds of lines of newly generated AI slop and not take the time to do a thorough code-review to determine if it broke LEAN design practices or if it disrupted the orthogonality of the existing system – i.e. introduced new code that unintentionally affects other code. Juniors typically don’t look for this in their code on a good day – but for AI to offer code slop made it all much worse.
The Time-Suck of Senior Developers
Which takes Alberta (and many of us senior-level engineers) into problems that hit closer to home. We are the gatekeepers responsible for making sure new code is following the design patterns and is bug-free and contains unit testing code-coverage to prove it. When hundreds or even thousands of new lines of code … or so much worse … lines of changed code get submitted for approval, our weekends are shot. There’s no way we can go through all that… so we just hit the “reject” button and walk over to the junior’s desk to have a talk… which doesn’t do much good because the manager just had a talk with them earlier that day and ordered he (or she) use more AI to vibe code faster.
Vibe Coding Isn’t Great, But Agentic Engineering Has Promise
Now we can create agents that do some gatekeeping for us (senior folks). We can give full instructions on what tests to see, what coding practices to adhere to, and what design patterns to implement across the code to make it consistent, clean and coherent. When people check in slop, the agents can at least do a quick check and flag if anything looks off … like more than 15 lines of code change (I’m not naming who, but it was a manager in my past who really didn’t know better). So it now becomes this Agentic Bureau of Acceptable Code As-a Service…
That’s right! I’m cracking out an ABACAS. I want to bring history full circle!
Now go try that PERSONAL BRAND ASSESSMENT … it actually is pretty useful!
You must be logged in to post a comment.