RadReads​ - Is ChatGPT pro worth $200 a month?

Hey friend,

I bit the bullet.

Just for you guys.

I wanted to see if ChatGPT pro is worth the absurd price

I'd been stalling for a while.

First, that's a lot of money.

But more importantly, I didn't know how I could justify the ROI.

What's included in Pro

Some of the most vocal advocates of AI (like the economist Tyler Cowen) insist that if you're not using the latest models, you're missing out.

The model I always hear people rave about is GPT o1 Pro - the most advanced reasoning model.

So I wanted to get my hands dirty with it.

Here are the other differences between the Plus ($20/month) and the Pro ($200/month) plans.

Operator and Sora - cool, but meh

I was excited to try Operator which basically is an agent that can conduct tasks for you on the web.

I wanted to run an analysis on my 2024 spending, so I asked:

This was WILD. And a glimpse into the future.

It was clicking through each statement, saving the pdfs - exactly like I had asked.

It was quite easy to see a world where rote and repeatable tasks (on real web pages and real software) would just be automated away.

However, it really struggled with the last mile. Operator had saved all the files and I could see them. Yet for some reason, I couldn't download them.

After 3 different attempts - I gave up.

(A quick note on Sora: It's a video generation tool, and it works well. But unless you're a videographer or in marketing, it's probably not that useful for you.)

Next up: Deep Research

Deep Research is a tool that does exactly what its name implies.

It goes deep.

Websites. Academic papers. Books.

AI pundits speculate that this product will eliminate the need for a lot of entry-level (and potentially more advanced) knowledge work.

(Note: The $20 version now lets you do some Deep Research queries, so I encourage you to try it ASAP.)

In my case, I'm preparing for a talk on the psychology of money called "How much is enough?"

So, I fired off my first Deep Research Query to find all the research combining money, happiness and well-being. Here's the prompt:

I then did another search for all the stories on the same topic. I wanted to capture fables and parables, yet also more contemporary stories from blogs and current events. The prompt was similar:

Now each query took about 15 minutes (so I did a bunch of push-ups while I waited) and the results were damn good.

The research query caught the classics (Daniel Kahneman on $75k and studies on lottery winners).

And the stories query had breadth and really incredible stories (Chuck Feeney, Maupassant and Siddhartha).

If you're curious, here are the full results.​

Synthesizing this all with GPT o1-Pro

Armed with a massive set of data, my next task was to make sense of it all.

I wanted to organize these outputs into an outline. One that would serve as the basis of my talk.

For that, I went to the most powerful reasoning model (aka "thinking model") with the data from the prior step.

Here's the prompt (notice that I voice dictate these and I don't worry about punctation or formatting):

This query takes advantage of GPT o1-pro's massive context window (aka you can paste a lot into the prompt, roughly 133,000 words).

It took ~ 3 minutes to run.

I can't share the full output (but you can find it here). Here's a sample:

• Opening (Hook + Core Question)
  • Open with a vivid anecdote: briefly recall Adolf Merckle—a billionaire who lost a fraction of his net worth and ended his life despite still having immense wealth.
  • Pose the central question: “If a billionaire can feel it’s not enough, how much is truly enough?”
  • Frame the talk: we’ll explore where money does increase happiness, where it doesn’t, and how envy or comparisons twist our perceptions.
• Section 1: The Emotional Pull of “More”
  • Brief Highlight of Sam Polk (former Wall Street trader): he called his obsession “wealth addiction,” always wanting a bigger number.
  • Show how social comparison fuels the chase: “1 or 2 million doesn’t look sweet if the person next to you has 10.”
  • Transition: Our culture often says more is better; but does data support that?

It's not great, IMO.

It used too many stories and failed to create a compelling narrative arc. But if you look at 30 minutes of Deep Research and 3 minutes of o1-pro, that's a lot of work that was done in 33 minutes.

It would have taken me at least half a day to replicate this output.

Last Step: Writing out our talk with the GPT 4.5 model

Now, in full disclosure, I haven't spent much time with the 4.5 model.

But from what I've read, it's great for writing and has a lot more personality.

And since we're paying for it, we might as well use it!

Once again, you'll recognize the prompt (which includes the outline from the prior step):

The result was mediocre. Garbage-in, garbage-out. (You can read 4.5's full write-up here.)

Will I use this as my talk? Absolutely not.

Does it get me 25% of the way in 35 minutes by doing the heavy lifting?

Absolutely.

Verdict: Is it worth $200 a month?

If you're not using Operator and Sora and only do a few Deep Research queries, the answer is no.

But I suspect that the models at the $200 price point are much better. (I just can't prove it... yet.)

More importantly, here's why I'm going to keep paying for it:

You may have missed

• ​8 ways to get started (if you're feeling overwhelmed)​
• ​AI is much more than a Google replacement​
• ​All the paid (and free) AI tools I use​
• ​How I got AI to write like me​
• ​How I created custom math quizzes for my daughter using Claude Artifacts​

See you next week!

Khe

PS Did a friend send this to you? Sign up to Future-Proof your Career with AI, you won't regret it (and it's totally free).