The 2025 AI predictions that aged terribly

DEEP DIVE
The 2025 AI predictions that aged terribly

A few weeks ago, the infamous “AI 2027” quietly updated its predictions.

This was the same report from April 2025 that described a full-blown catastrophe if AI progress kept accelerating.

Then reality hit, their model fell apart almost instantly, and they had to tack on a note saying:

“Hey everyone, chill, this was just our opinion.”

And now that we’re entering peak “Here are my predictions for 2026” season, I want to go the other direction.

I want to round up the predictions people made for 2025 that completely face-planted.

But before pointing fingers, I’ll start with my own L.

My Failed Prediction: “AI Agents will compete with SaaS”

Back in November 2024, I said:

Looking back, this wasn’t the boldest claim, but even then, I was too early.

If anything, 2025 showed me how much infrastructure still needs to be solved before agentic workflows become stable enough for non-technical users.

Alright. Enough about me.

Prediction #1 — Elon Musk: “AI smarter than any human by 2025.”

I don’t think anyone would argue that this is even close to a reality.

To reach that bar, AI needs deep world understanding, grounded reasoning, and reliable planning, not just better benchmark scores.

And this year showed the limits of the “just scale it” approach every lab has been relying on.

Prediction #2 — Sam Altman: “OpenAI will achieve AGI by 2025.”

Sam has a habit of pushing big predictions that sound like vision but read like fundraising. The cynical interpretation is that you need those promises when you’re burning billions and need more capital.

But the reality at the end of 2025 is that ChatGPT costs a ton to develop, and it still isn’t profitable.

And OpenAI lost its clear lead to other players: Gemini, Claude (at least on coding), and DeepSeek, who came out swinging this month (again).

Whatever roadmap they drew up last year didn’t land where they said it would.

Prediction #3 — Dario Amodei: “AI will write 90% of code by mid-2025.”

This is hard to measure precisely, but even by vibe alone, it’s clear it didn’t happen.

Most developers do use AI regularly now, but nowhere near 90% of functional and active code is AI-generated.

The hype around “vibe coding” cooled off fast.

Teams realized AI can scaffold and refactor, but it still produces bugs, dead ends, and everything people now call AI slop. We even saw a drop in day-to-day use by the end of the year.

Prediction #4 — John Zimmer: “Private car ownership will end by 2025.”

He gave himself nine years to be right…

Today, fully autonomous cars make up less than 0.5% of all vehicles.

Even optimistic forecasts say only 4% of new cars sold by 2030 will have Level 3 automation or above.

Autonomy didn’t collapse car ownership, it barely dented it.

Prediction #5 — Emad Mostaque: “Indian outsourcing will be destroyed in 2025.”

Not only did that not happen, but the sector actually grew 5% this year.

Once again, bold predictions from founders who benefit from hype don’t match what’s happening on the ground.

Replacing giant industries doesn’t happen because a new model dropped on a benchmark leaderboard.

A Closing Thought as We Head Into 2026

My hope for next year is simple: fewer dramatic claims, fewer “AGI is here” headlines, fewer predictions designed to farm engagement.

This space is starting to sound like the “crypto bros”:

Every year someone shouts “This is the year it all happens — trust me!”

Then the year ends, and everyone quietly moves the goalposts.

AI is still moving fast, but it’s moving in the normal direction: improvement, not revolution.

No AGI.

No extinction.

No overnight job collapse.

No instant utopia.

Just better tools, better workflows, and a lot of noise from people who want attention.

And honestly?

I’ll take that version of 2026.