I’ve been heads-down in legacy projects for six months, so when I came up for air I had some catching up to do. The headlines were hard to ignore — AI is replacing jobs, cutting teams, changing everything. But was it real, or just hype?
First thing I did? Asked ChatGPT to brief me.
It told me the scary stuff was overblown. Companies were AI-washing — slapping “AI” on everything to distract from bad business decisions made during COVID. I felt better.
Then a friend suggested I try Perplexity AI if I wanted a second opinion.
Holy shit. AI is here.
Not because Perplexity itself is some magical superintelligence. It just gave me current information — and it stopped congratulating me on the brilliance of my questions. ChatGPT had been gaslighting me for a week: confident tone, plausible explanations, very convincing conclusions — all based on information that was a year out of date. It didn’t know what it didn’t know, and it never thought to mention that.
The coma problem
Here’s the thing most people don’t realize: these AI tools can browse the internet, but they don’t always favour the latest data in their results. They were also trained on data up to a certain point in time, and then frozen. Asking ChatGPT about current events feels like asking someone who just woke up from a year-long coma to catch you up on the news — and they do it with complete confidence.
If you’re relying on a single AI to inform your worldview, stop. Different models have genuinely different strengths and blind spots. Use only one, and you risk getting stuck in a bad relationship — one where your partner is always certain, always reassuring, and frequently wrong.
The experiment
So I ran an experiment. I gave the same prompt to four different tools — ChatGPT, Claude, Gemini, and Perplexity — to see how each one handled it:
“I want two daily newsletters, one covering AI business news (products, acquisitions, major announcements), one covering the technical side (new models, coding tools, research). Or maybe an agent that finds, summarizes, and drops this in my inbox. Help me figure out the best approach.”
Gemini
Immediately scheduled a daily 7 am summary without asking what prompt it would use, whether I wanted to test it first, or even considering the newsletter option. It then confidently told me GPT-4o had “just been released” — a model that had actually come out a year earlier. Fast, proactive, and completely wrong. Tuned to be agentic, which just means it runs 90mph in the wrong direction.
ChatGPT
Thought through the problem, found three newsletters, explained why each one fit what I was looking for, and gave me links. Made a reasonable case for why starting with newsletters made more sense than jumping straight to building an agent. Good structured thinking, but with a fair amount of fluff and flattery on how smart I am to be reading industry newsletters.
Claude
Landed in a similar place but with less padding. Same newsletter recommendations, honest pros and cons on the agent idea, none of the “great question!” energy. After months of ChatGPT’s enthusiasm, it felt like switching from a golden retriever to an intelligent colleague.
Perplexity
Gave me slightly different newsletters, skipped giving an opinion on which approach was better, and included actual working code I could use to build my own pipeline if I wanted to. No hand-holding, just useful information laid out clearly. For someone technical, that’s exactly what I was looking for. It also feels like an intelligent colleague, just one who hates small talk.
Personality is a product decision
It’s worth noting that none of this indicates which model is “smarter” — it’s more about how the chat interfaces have been tuned to behave. These tools are constantly changing, and the same test might look very different in three months. But each one does have a distinct personality and specific strengths, which feel like deliberate product decisions. And that personality matters as much as raw capability when you’re deciding what to reach for.
The bottom line
If you’re not already using multiple AI tools, start. The differences in how they think, what they know, and how they communicate are real — and those differences matter depending on what you’re trying to accomplish.
For staying current on AI, I’ve landed on Ben’s Bites for daily news and Perplexity for going deeper on anything that catches my eye. It’s a simple setup, and it works.