Hey Reader,

Quick note before we dive in: Happy holidays. Hope you get some rest. You deserve it. 🎄

Now let's talk about something that might ruin your holiday spirit.

Opus 4.5 Just Got Expensive

The promotional pricing has officially ended.

That means Opus is now 2-3x more expensive than it was a few weeks ago.

If you're using it inside Cursor, Windsurf, or Augment Code with thinking mode enabled?

Your wallet is about to start crying.

Here's the reality: The only way to keep using Opus without breaking the bank is to grab a Claude Max subscription and use Claude Code directly.

I currently have one again, but mostly for Claude itself, not just for building.

The alternative is, and that's what I do, burn money on Cursor.

Because the ability to switch models on the fly and use plan mode is worth every penny to me.

But I get it. Not everyone wants to pay premium prices just to build their ideas.

Is GPT-5.2 Any Good?

I teased this in my last email.

I've now tested it properly.

And I have screenshots that show exactly how expensive this thing can get. One of my plans the other day? Over $8. For a single plan.

That was the extra-high effort variant in fast mode.

But here's the thing…

The model is genuinely great.

It can go head-to-head with Opus 4.5.

These are the only two models I actually use right now.

Everything else? Not worth your time.

The Truth About Other Models

Gemini 3 Pro is honestly trash. I won't sugarcoat it.

Yes, it has its uses. But it makes so many mistakes and completely nukes your app in the process that I can't recommend it to anyone.

And everything else isn't even worth mentioning.

My Recommendation

Stop. Overthinking. This.

Opus 4.5 and GPT-5.2 are both phenomenal. Both can build whatever you want.

Here's how they differ:

Opus 4.5

• Feels more interactive
• It's faster
• More fun to use
• Best tool calling (editing files, MCP servers, etc)

GPT-5.2

• Much more autonomous
• Requires almost no babysitting
• More reliable/trustworthy
• Smarter overall (especially for planning)
• But slow as hell

I use both because I'm in Cursor all day.

But if you're watching your budget?

Pick one lane and stick with it.

Like Claude and want to use Claude Code? Get a Claude Max plan.

Prefer the OpenAI ecosystem, the image generation, the ChatGPT experience? Use Codex instead.

Both will get the job done, stop overthinking this.

What's Coming Next

The model race isn't slowing down.

An update to Gemini 3 Pro is rumored to come sooner than we think.

Another rumor mentioned Sonnet 4.7 might be here before year's end too.

But here's my advice:

• Use what works today
• Don't wait for the "perfect" model
• Ship your ideas with what's available now

That's exactly what I teach inside the AI Builder's Blueprint.

Which by the way is currently 20% OFF for the Holiday Season! 🎁

If you want to start 2026 building apps instead of dreaming about them, this is your sign.

​

Finally, let me know if you're building anything over Christmas. I'm reading every reply.

Talk soon,
Rob

P.S. This might be my last email of the year. Or maybe second-to-last. Either way, thank you for reading every week. It means more than you know. Now go enjoy the holidays. And if you want to start the new year with a new skill, the Blueprint is waiting for you.​

​

​