ChatGPT Agent is here

Welcome back to Building AI Agents, your biweekly guide to everything new in the field of agentic AI!

I’m slowly making peace with the fact that I’m soon going to be spending as much on AI each month as on food.

The highlight of the week in agent news was definitely OpenAI’s release of ChatGPT Agent, but AWS and Google’s announcements are not to be slept on.

In today’s issue…

• OpenAI launches ChatGPT Agent

• Google makes it easer for agents to build agents

• Get 80% off on Building AI Agents’ course and community!

• How to get an agent startup to $10 million in revenue

• Predicting the future with agents

…and more

🔍 SPOTLIGHT

The “intelligence” in artificial intelligence is like beauty—it’s often in the eye of the beholder.

Last Thursday, OpenAI unveiled ChatGPT Agent, a system designed to unify its Operator and Deep Research agents into one general assistant for users which can plan and book trips, research and buy products, perform financial analyses in spreadsheets, and more. A demo video released by OpenAI showed Sam Altman and some of the company’s engineers putting it through a variety of tasks, such as making plans for a wedding, creating and ordering laptop stickers, and building an itinerary for a roadtrip to all 30 Major League Baseball stadiums—including one which is apparently in the middle of the Gulf of Mexico (Gulf of America?). Flawless it is not.

The agent was immediately available to ChatGPT Pro users, who will get 400 requests per month, with the lesser Plus and Teams tiers getting access “in the coming weeks” at 40 requests per month.

Immediate reviews have been mixed, with some users raving about it, some satisfied, and others laughing it out of the room. This tends to be the case with nearly any new AI product—there’s much more randomness to their performance than deterministic tech like an iPhone, meaning that they will inevitably perform brilliantly on some tasks and faceplant on others. Combine that with the preset biases that reviewers bring to the table, from “the singularity is near” hype to “AI is a Tech Bro™ scam” cynicism, and you have a perfect Rorschach test. The only true way to get a good sense of the agent’s capabilities is to try it, giving yourself enough time to get used to using it—working around LLMs’ foibles takes practice.

My own experience? When OpenAI rolls out access to Plus users, I’ll get back to you. I can say, however, that this is one of the first times I’ve been tempted to cough up $200 a month for the Pro tier. Here’s an easy heuristic: multiply the number of hours a month you spend working on the computer—on anything—by what you make in an hour, and what felt like a ripoff before starts to seem much more reasonable. This is the value proposition that OpenAI and Anthropic are pitching with their premium tiers: if their agents can do anything on a computer that you can, accessing them for $200/month is a steal.

The flaw, of course, is that they can’t do everything you can, unless you’re incapable of recognizing that there are no baseball stadiums 100 miles off the coast of Louisiana. Saying “there’s no way AI will ever solve this”, though, has been a badly losing bet over the last decade, and particularly the last 2-3 years.

It’s not clear yet whether ChatGPT Agent is worth $200 a month. But it won’t be long before we have agents for which that question is a no-brainer.

 🚀 $50 COMMUNITY ACCESS FOR JUST $10

We’re offering this deal to prove how much value you’ll get, but it ends Tuesday, July 22nd at midnight!

Why join right now? 🤔

• 80% OFF: Join for just $50 $10, ends Tuesday at midnight.

• Build real AI agents, fast: Get hands-on in our step-by-step n8n Masterclass, no coding needed.

• Plug-and-play agent templates: Launch your first agent with proven workflows.

• Live support & builder community: Get real help from founders and fellow builders.

• Your wins = our proof: We want to feature your results as our next success story.

Don’t wait, this deal ends Tuesday at midnight. Lock in your spot and help us prove what’s possible.