The problem is in the room, when four people hear the same words and take away four different versions of it.
Everyone nods. Then the meeting ends.
A few days later the notes are a slide. A few weeks later the slide is a requirement.
Fig. 1 — This is where a lot of translation loss begins.
Vibe coding has had quite a year. Andrej Karpathy coined the term in early 2025; by the end of the year it was Collins’ word of the year, and this spring he joined Anthropic. Purely hand-written code is becoming rare indeed, coding with assistants is the standard.
Vibe coding for business takes a different focus: prototyping. It is now as easy as speaking to Lovable to put a first working version on the table, made by the people who know the work. Everyone can understand it, bend it, test it. The prototype is disposable. The shared understanding is the deliverable.
It is a bit like having all IKEA modules at your disposal and building furniture with the person who will actually use it standing next to you.
You read the manual, build the thing, and only then find out it does not fit the room. By that point, pride, time and budget are already invested.
- The correction arrives early
- Before pride is invested
- Before time is spent
- Before budget is committed
Fig. 2 — The thing changes while it is still cheap to change. That is the value.
A working sketch in the room gives everyone something concrete to react to, and each person reacts from where they stand.
Fig. 3 — Five positions, one object.
The prototype becomes what sociologists long ago named a boundary object: a single thing that people standing in different parts of the organisation can all work on, without first having to agree on the words.
Fig. 4 — People can take different positions on the same object and still be talking about the same thing.
That is why I like the word superglue.
Service design has always worked through objects people can stand around — journey maps, blueprints, storyboards. Vibe coding adds a new material to the practice.
Fig. 5 — The newest material lets the intangible appear earliest — not perfect, not polished, but concrete enough to make the conversation sharper.
It also changes the rhythm of a sprint. Making moves toward the research and sense-making: you can test an assumption while people are still in the room.
Fig. 6 — You can expose missing data, unclear ownership or broken handovers before they get buried inside a delivery backlog.
But only with structure. Without it, vibe coding becomes theatre: something appears on screen, everyone gets excited, it looks like progress because the prototype is moving, and nobody has agreed what question it is answering.
Fig. 7 — Movement on screen is not the same as progress. Structure is what tells them apart.
Engineering got its harness. Design already has one. The questions do not slow the making down; they tell you what the making is for.
Fig. 8 — AI can help us make faster. It cannot decide what is worth making.
For designers this changes the job. We used to carry the translation alone: listen, synthesise, return weeks later with artifacts. Now some of it happens live, with the people who know the domain shaping the first version directly. That requires letting go.
Fig. 9 — Letting go of these is what makes room for something better.
Fig. 10 — Shared understanding moves from the end of the process to the middle of the conversation.
Vibe coding leaves research, design judgement, architecture, governance and delivery discipline exactly where they were. It changes one thing: when shared understanding becomes possible.
A sketch made inside the conversation shows where the process is unclear, where the source system is missing, where a feature request is actually a broken handover.
Before anyone prompts anything in your next session, write the question the prototype must answer where everyone can see it. If nobody can phrase it, you are about to make theatre.
Sources: Vibe coding — origin, Collins Word of the Year, criticism, CodeRabbit — AI co-authored code, PR analysis, The state of vibe coding in 2026, TechCrunch — Karpathy joins Anthropic, Star & Griesemer (1989) on boundary objects