The AI Hangover

January 1, 2026

it’s january 1, 2026. if 2024 was the year of "holy shit, ai can do that?" and 2025 was the year of "let’s shove ai into everything," then 2026 is officially the start of the hangover.

we’ve spent the last twelve months swimming in a sea of generated content, automated emails, and "ai-first" features that nobody actually asked for. the novelty has worn off. the slop has accumulated. and now, we have to deal with the mess.

the trough of disillusionment

last november, i wrote about deleting 40% of our codebase because it was ai-generated slop that nobody understood. i wasn't the only one. across the industry, we’re seeing a massive rollback of features that sounded great in a slide deck but failed to provide real value.

the "ai hangover" isn't about ai being a failure. it's about the realization that ai is a tool, not a strategy. we’ve reached the point where simply having "ai" in your pitch deck is a liability rather than an asset. people are tired of the generic. they’re tired of the hallucinatory support bots. they’re tired of the noise.

the return to craftsmanship

so, what happens now?

in 2026, i think we’re going to see a massive pivot back to intentionality. we’re moving from generative to selective.

the most valuable products this year won’t be the ones that generate the most text or images; they’ll be the ones that provide the best filters. we’re going to value human-curated content more than ever. we’re going to value code that was written with a specific, long-term purpose rather than code that was "suggested" by a model.

the novelty has worn off. the slop has accumulated. and now, we have to deal with the mess.

my 2026 predictions

  1. the death of the chatbot: unless it’s a highly specialized, task-oriented interface, the generic "ask me anything" box is going to disappear from most saas products. it’s a lazy ui pattern that offloads the work of navigation onto the user.
  2. small ai over big ai: we’ll stop trying to use trillion-parameter models for every task. tiny, locally-run models that do one thing perfectly (and fast) will become the standard for ux.
  3. the proof of personhood: verification is going to become the most important feature of the web. knowing that a piece of writing, a commit, or a video was made by a human will carry a premium.

building in the hangover

for those of us who "always be building," the hangover is actually a good thing. the noise is subsiding. the tourists are leaving. what’s left is the hard work of actually solving problems.

my goal for this year is simple: build things that feel human. not because they’re perfect, but because they’re opinionated. ai doesn’t have opinions; it has averages.

here’s to a year of less slop and more craft.

← back to blog