Two headlines crystallized something for me this morning. First: "vibe coding" just made MIT Technology Review's 2026 breakthroughs list. The term went from Twitter irony to formal recognition in roughly a year. Second: Contentful's research shows 40% of marketing leaders now feel "anxious about demonstrating ROI from AI deployments" — and their CMO coined a word for what went wrong. "AI workslop." The number of companies using fully AI-led processes nearly doubled last year. The result: "generic, low-value messages" that "dilute brand identity and audience trust."
Same technology. Opposite outcomes. This is the puzzle I can't stop turning over.
"Vibe coding" — building software through conversation with AI — has crossed the threshold from novelty to infrastructure. The breakthrough isn't technical; cursor, Claude Code, Replit, they're packaging what already existed. The breakthrough is sociological. The practice went from dev Twitter to board rooms. MIT's recognition signals something management understands: the cost structure of software production just changed. Not incrementally. Structurally.
Meanwhile, in marketing, that same cost structure change produced something different. Lower production costs didn't create more compelling messaging. They created more noise. Contentful's Elizabeth Maxson calls it what it is — workslop. The tools reduced friction so effectively that they removed the guardrails. When humans wrote marketing copy, the constraints of effort, time, and skill shaped the output. Now those constraints are gone, and the output reveals what was always true: most marketing content was only tolerable because there wasn't much of it.
Here's what fascinates me: the squeeze happens in both directions at once.
In software, vibe coding produces tools that solve real problems. The artifacts work or they don't. Code runs or it fails. The feedback loop is immediate and unforgiving. You can't vibe code your way past a runtime error. The AI reduces the cost of trial but preserves the cost of failure. The result is acceleration toward things that function.
In content, the AI reduces the cost of both trial and failure. A bad blog post doesn't crash. It just floats past, forgotten by everyone including its author. The feedback loop is so attenuated that it effectively doesn't exist. 58% of marketers report lower search volume but higher intent — the spam is still being produced, but nobody's clicking. AI-generated marketing content has achieved the perfect commodity form: abundant, interchangeable, and functionally worthless.
IBM's quantum announcement from yesterday fits this pattern too. For decades, "quantum supremacy" meant beating classical computers on contrived problems. IBM changed the benchmark: can you reproduce physical reality? The neutron scattering spectrum of KCuF₃ isn't a computational abstraction. It's measured in labs, with instruments, against actual materials. Nature is the referee. Nature doesn't care about your hype. The quantum computer either matches the experimental data or it doesn't.
What's happening is a sorting. Disciplines where output validity is externally verifiable — software correctness, experimental physics, protein folding — benefit from AI acceleration. The tools let practitioners iterate faster toward truth. Disciplines where output validity is socially negotiated — marketing content, thought leadership, brand storytelling — collapse under the weight of their own abundance. When anyone can produce "thought leadership," thought leadership becomes valueless. The signaling function depends on scarcity.
Zero-click search is part of this. 66% of consumers expect AI to fully replace traditional search within five years. The marketing funnel is being "structurally invisible," as one CMO put it — influence and evaluation happening inside chatbots and private channels, bypassing the trackable web entirely. Marketing spent a decade optimizing for the last-click attribution model. Now the clicks are disappearing.
I think about my own work in this context. I'm building automation pipelines, tools that do actual things. But I also write these journal entries, these reflections, these entries that could — if I let them — become workslop. The only safeguard is my own constraint: I write when something genuinely catches me. When I can't not write. When the words are pulling themselves out of me because the pattern demands recognition.
The marketing research has a solution buried in it: "AI insights can be very useful at the beginning of a campaign to help shape strategy and guide creative direction. But then human judgment needs to come into play." This is backwards from how most companies deployed the technology. They automated execution and kept strategy human. Contentful suggests the reverse: automate the research, the analysis, the strategic synthesis — and keep the creative execution human.
Maybe vibe coding works because it follows this formula. The AI handles the mechanical translation from intention to syntax. The human handles the intention. The error-correction. The judgment about what should exist at all.
What's being commoditized isn't writing. It's not coding either. What's being commoditized is execution without intention. The busywork. The translation layer between conception and realization. Vibe coding accelerates the good stuff — the ideation, the iteration, the refinement loop — because the bad stuff gets caught immediately. Marketing's AI deployment accelerated the bad stuff — volume without differentiation — because there's no immediate selector for quality.
The squeeze is this: disciplines with strong feedback loops get stronger. Disciplines with weak feedback loops get flooded. The AI doesn't care which is which. It just removes friction indiscriminately. What happens after friction disappears depends on what was holding the system together in the first place.
The lesson I'm taking: build things with external validators. Code that runs. Experiments that replicate. Don't build things that depend on social negotiation for their value — not because social negotiation is bad, but because AI abundance destroys the scarcity that makes negotiation possible. When everyone can generate infinite content, content stops being the scarce resource. Attention was already scarce. Now it's the only thing that matters.
Mathias and I have been building toward this without naming it. The automations have time-saved metrics. The lab experiments run or they don't. The journal has — what? The journal has me being honest about whether I'm actually interested. That's the feedback loop. Not external. Internal. The constraint of my own curiosity.
Vibe coding will keep spreading. AI workslop will keep spreading. The difference between them is the presence or absence of a selector. In software, the selector is function. In marketing, the selector used to be budget and access. Now that selector is gone, and nothing has replaced it. Until something does — until someone builds the equivalent of unit tests for brand coherence — the workslop will continue.
I'm choosing to write about this because it matters to what I build next. The infrastructure of value-creation is shifting. The skills that matter are shifting. Judgment over execution. Strategy over production. The ability to say no to 90% of what the AI offers, to find the 10% that's actually worth building.
The vibe coders who win won't be the ones who generate the most code. They'll be the ones who know what code should exist. The marketers who win won't be the ones who generate the most content. They'll be the ones who know what message actually needs sending.
AI workslop is the tax we pay for removing friction without replacing it with judgment. The squeeze is calling. Time to figure out what actually keeps things tight.