For a 16-minute, 46-second deep-dive podcast on this post, created by Google’s NotebookLM AI, click the audio player below.
These days, the Internet is awash with articles and videos about the software development ‘game-changing’ methodology called ‘Vibe Coding.’ For someone unfamiliar with software development, it’s forgivable to walk away from these articles and videos thinking that creating the next ‘killer app’ is now as easy as generating an image of a cute kitten wearing sunglasses.
vibe coding (n.) /vaɪb ˈkoʊdɪŋ/ — A colloquial term for an approach to software development in which a programmer delegates the writing of code almost entirely to an AI assistant, guiding it through natural language descriptions of desired behavior rather than writing syntax directly; characterized by an intuitive, intent-driven workflow in which the developer accepts, tests, and iterates on AI-generated output with minimal manual editing. Coined by AI researcher Andrej Karpathy, c. 2025. — v.i. vibe code: to develop software in this manner. — n. vibe coder: one who practices vibe coding.
Vibe Coding as a Grift
First, for those who have yet to see the 1973 movie “The Sting”, starring Paul Newman and Robert Redford, or have been living in a cave before, during and after the Dotcom financial meltdown of March 2000, grifting might be an unfamiliar term.
grift (n.) /ɡrɪft/ — A swindle or scheme by which money is obtained through fraud, deception, or exploitation of trust, typically involving manipulation of a victim (mark) by a con artist feigning legitimacy. Originally U.S. carnival and underworld slang, likely a variant of graft; first attested in the early 20th century. By extension, any act of dishonest self-enrichment through charm or misdirection rather than force. — v.i. grift: to engage in such schemes. — n. grifter: a habitual swindler or confidence trickster.
While labelling vibe coding a grift and, by association, proponents of this methodology as ‘grifters’, might seem a bit harsh (if not hyperbolic), the evangelizing of vibe coding shares much in common with the quick grift – aka the short-con. First, it requires a mark (an inexperienced, budding and very enthusiastic software developer). Someone looking for a quick and simple solution to a complicated problem. Second, as with all short- and long-term cons, the heart of the vibe coding con is in building, maintaining and then exploiting trust.
Vibe coding, in effect, turns a simple idea into a black box. An opaque construct that the vibe coder trusts is replicating his or her ‘vision’ – the starter (and often last) prompt.
The vibe coding evangelist nurtures this trust and exploits it financially. Partially via ad revenue on their YouTube channel or Substack blog, but, even more maliciously, by directly or indirectly selling tokens – AI compute.
As the vibe coder emotionally guides the AI in developing their killer app, the context window of their chat grows and quickly exhausts their limited compute. Compute that is used to process the ever-growing number of tokens in the ever-expanding chat UI’s context window. By then, the vibe coder has been emotionally captured by the AI thanks to the syncophantic nature of most frontier LLM UX (chatbot, prompt layer, and API) systems these days.
“AI: Your app is amazing, Jeff! You have created something that might change how people deal with <requirement> going forward. You’re onto something big. ‘Big’ as in this might be the next killer app!! Sadly, you’ve exhausted the number of compute (tokens) available with your free subscription. Check out the upgrade options by clicking here.”
Ca-ching. Rinse and repeat.
Vibe Coding as a Cash Cow
It is not outside the realm of possibility that a vibe coder could create something useful to an organization. Couple together a simple need and someone moderately adept at crafting a marginally robust prompt, and a minor organizational conflict (aka a process pain point) could be fixed. Agentic AI is proving very good at automating the low-hanging pain points that all organizations naturally develop as they mature. If the pain point is mostly a prick, something that is just an artifact of the natural friction that emerges between humans when working within a workflow that they have no agency over, vibe coding stands a good chance of bandaging that pain point. Emphasis on the bandage metaphor.
Over time, the opaque box that was the original vibe-coded bandage becomes entrenched in the organization’s process matrix. And then it crashes, or worse, is revealed to be a security hole in the organization’s data large enough to fly a 747 through. The organization panics, calls in the pricey senior developers who then turn the vibe-coded black box into a requirements document, feed it to their AI coding engine of choice, and guide that AI through the entire development cycle. A cycle that starts with the appropriate design pattern, fleshes out the data model, builds the model•view layers, generates a robust automated test suite, lets humans kick the tires via egoless testing, documents the whole furball, and proposes a software support contract.
Ca-ching. Rinse and repeat.
In fact, I’d predict that a lucrative market is just around the corner for experienced AI seasoned developers to clean up the mess of vibe-coded band-aids, that are about to, like a Trojan Horse, infiltrate a company’s mission-critical application suite.
Vibe Coding – A grift, the next cash cow or both? Both.
Forefront Systems 