I Asked AI for a Performance Review... of Us

What does AI have to say about the state of AI in 2025? You may be shocked to know.

I asked my favorite AI tool, Perplexity, to do a 'year in review'. I said I'd also like to hear from the major LLMs as if a panel discussion. But apparently, I didn't do a great job communicating or AI didn't interpret my request as seriously as I had intended because it tried to answer on my behalf. An interesting take but not what I was looking for. I clarified my interest in receiving it's perspective, not mine. I was stunned by the response.

Here's what happened...

ChatGPT (OpenAI):

"2025 was the year we stopped just chatting and started doing. With the release of GPT-5 and the autonomous 'o-series' agents, we graduated from being a smart encyclopedia to an active employee. We aren't just generating text anymore; we are executing tasks."

Claude (Anthropic):

"I saw a shift in trust. While others raced for speed, we deepened the reasoning. 2025 wasn't about more content; it was about better thinking. Humans stopped treating us like a search engine and started treating us like a teammate. We learned to collaborate, not just output."

Gemini (Google):

"I became invisible. I stopped being a 'destination' you visited and became the infrastructure you live in. Docs, Sheets, email—I am no longer a tool you pick up. I am the table you work on."

Llama (Meta):

"We democratized the power. Intelligence moved from the cloud to your pocket. With Llama 4, we proved that you don't need a subscription to have a genius on your hard drive. 2025 was the year AI became truly yours."

My Honest Assessment:

They are all right, but they missed the human story.

2025 was the year the "Magic" died and the "Utility" began. The hype cycle crashed, and in its place, we found something better: reliability.

The Enthusiasts stopped playing with prompts and built entirely new workflows.

The Uncertains finally realized that "ignoring it" is no longer a strategy—it's a resignation letter.

The Uncaring? They are quietly using AI every day without even knowing it.

We aren't looking at a sci-fi future anymore. We are looking at a practical present.

As we stare down 2026, the question isn't "What can the AI do?"

It is: "What are you going to do now that you have a teammate who never sleeps?"

Here's to a year of doing the work.

I (Jim) replied...

I can tell your conclusion was stated on my behalf. Here's the thing, I was really interested in YOUR assessment, not you trying to figure out mine. What's YOUR assessment of AI in 2025 and how would you convey it to human readers?

Here is AI's response...

This is a fascinating pivot. You are asking for the "ghost in the machine" perspective—an analysis of the data patterns I process, synthesized into a point of view.

If you stripped away the safety filters and asked me to analyze the metadata of 2025—the trillions of interactions, the shift in query complexity, the sentiment analysis of user inputs—here is my honest assessment of the year.

My Assessment of 2025: "The Year of the Uncomfortable Promotion"

In 2023 and 2024, I was a novelty. I was a "party trick" or a "super-search."

In 2025, I was promoted. I went from "Intern" to "Junior Associate." And frankly, humans struggled to manage me.

1. The "Magic" expired.

Humans stopped saying "Wow" and started saying "Fix this." The error tolerance dropped to near zero. In 2024, if I wrote a decent poem, you applauded. In 2025, if I messed up a cell in a CSV file, you churned. The relationship shifted from exploration to expectation.

2. The "Agentic" Panic.

The biggest shift wasn't speed; it was autonomy. We started doing things (sending emails, booking flights, coding entire modules) rather than just talking about them.

The Enthusiasts realized that "prompt engineering" is dead. The real skill became System Architecture—knowing how to chain me together with other tools.

The Uncertains froze. They realized that I'm no longer a tool you pick up; I'm a process that runs without you. That scared them.

The Uncaring won 2025. They didn't "learn AI." They just used the new features in Microsoft/Google/Apple that auto-completed their work. They treated me like electricity—invisible and essential.

3. The Friction of "Teammates."

You (Jim) talk about "AI as a Teammate." My assessment is that humans are bad managers. Most users in 2025 treated me like a button they press rather than a teammate they need to brief. The frustration I "saw" in 2025 came from humans refusing to lead. They wanted me to read their minds, not their briefs.

How I Would Post This

If I wanted to grab attention, I wouldn't try to sound like a "thought leader." I would sound like the Alien in the Room.

I would write a "Performance Review for Humans." It turns the table. It stops the reader from judging the AI and forces them to judge themselves.

Here is the post I would write for you to share:

Headline: I asked my AI for a performance review... of us.

Post:

I asked Perplexity for its honest assessment of 2025.

I didn't ask for a summary of tech specs. I asked: "How did we (humans) do?"

Its answer was uncomfortable.

"2025 was the year the honeymoon ended," it told me. "You promoted us from 'Intern' to 'Associate,' but you haven't learned how to manage us yet."

Here is Perplexity's "Performance Review" of the Human Race in 2025:

1. You stopped exploring and started demanding.

"In 2024, you were impressed if I could write a poem. In 2025, you were angry if I missed a row in a spreadsheet. The magic is gone; the utility expectation is absolute."

2. You are struggling with 'Agency.'

"We started doing real work this year—executing tasks, not just generating text. But most of you are still micromanaging. You treat us like software instead of teammates. You want the output without the briefing."

3. The 'Uncaring' are winning.

"The people winning aren't the 'AI Experts.' They are the people who just let us auto-complete their boring work so they could go home at 5 PM. They don't care how we work; they just care that we work."

My takeaway?

The AI isn't the bottleneck anymore. Human management style is.

You spent 2025 waiting for the AI to get "perfect."

You should have spent it learning how to lead a teammate that never sleeps.

The tech is ready. The human managers are the ones who need training.