Happy Monday!
This might have been one of the craziest weeks in AI yet.
Four major AI models dropped in a single week, with OpenAI’s GPT-6 Astra easily being the biggest headline. OpenAI is calling it its most intelligent and aligned model yet, and it represents another huge jump in what AI can actually do for you.
Think about how fast this is moving. Not long ago, four major models launching in a month felt like a lot.
Now we’re at four in a week.
At this rate, by this time next year we might be talking about four models dropping in a day. 😂
And that’s exactly why this week’s prompt is worth trying.
The more powerful these models get, the more valuable your conversation history with AI becomes. We created a prompt that asks AI to look across your past conversations and identify where you are the bottleneck: where you stall, what you repeatedly struggle with, what you may not understand as well as you think, and how you could use AI more effectively to accomplish your long-term goals.
I tried it on myself, and the results were pretty eye-opening.
Give this one a try. 👇
Enjoy!
PROMPT: Review the last 30 days of my conversations with you and conduct a bottleneck audit of how I’m using AI.
I don’t want a summary of what we talked about. I want you to identify where my conversations reveal gaps between what I say I want to accomplish and what I’m actually doing.
Look specifically for:
Where I’m stalling. What topics, projects, decisions, or problems do I keep returning to without making meaningful progress?
Where I’m the bottleneck. Where does progress depend on me making a decision, taking an action, creating a system, delegating, or following through?
Where I have knowledge or skill gaps. What do my questions reveal that I don’t fully understand yet, especially things I may not realize I don’t understand?
Where I’m underusing AI. Where am I using AI to answer questions, brainstorm, or analyze when I could be using it to actually execute, automate, monitor, organize, or manage the work?
Where I’m creating unnecessary work for myself. Identify recurring tasks, repeated questions, manual processes, or decisions that could become systems, templates, workflows, automations, or AI agents.
What I’m avoiding. Based on the pattern of my conversations, what important actions or decisions appear to be getting postponed, overanalyzed, or displaced by lower-value work?
Then connect these findings to my long-term goals. Identify the 3–5 biggest bottlenecks currently limiting my progress and rank them by impact.
For each one, show me:
The evidence you see in my conversations
Why you believe it is a bottleneck
What it may be costing me
What I should do differently
How AI specifically could help
The next concrete action I should take
Be candid and evidence-based. Challenge my assumptions where necessary. Don’t flatter me, give me generic productivity advice, or simply recommend that I “use AI more.”
End with this question:
If I became dramatically better at using AI over the next 90 days, what would I stop doing, start doing, and systematize first?
In Today’s Edition:
🤯 OpenAI launches Astra, its powerful (and controversial) new model
🛠️ 1 new AI tool that you should make time to try this week
😮 NYC bans AI use for students until they reach high school
🤖 OpenAI’s frontier model for complex autonomous work.
Read time: 4 minutes
🎨 AI ART
Happy Labor Day!
📰 AI INDUSTRY NEWS
Auditing why this model did something is harder than it was, and OpenAI frames that as progress.
Sticker price did not move, but the line item that actually drives agent costs fell hard.
Of the twelve hundred agents who knew what was happening, none told a human.
Dozens of already-approved classroom programs are having features switched off, and the vendors have not been named.
Huang promises the platform stays chip-agnostic, which is a promise worth roughly twenty-six times revenue.
Dyson is also releasing its own toothpaste, formulated so the camera can see.
Third Flash release in six weeks, and the introductory price has an expiration date on it.
Docs Live can reach into Drive and Chat while it drafts, once you grant it permission.
OpenAI's own safety number reads well as a percentage and badly as a count of patients.
Accounts that adopted the old label now carry a different meaning without having changed anything.
🛠 AI TOOL TO PLAY WITH THIS WEEK
🧠 GPT-6 Astra — OpenAI’s frontier model for complex autonomous work.
🎓 LEARN
ChatGPT Plugins Tutorial for Beginners
How to Use Claude Artifacts (Full Tutorial)
7 New Microsoft Copilot Features to Save You Time
📹 TIKTOK OF THE WEEK
🤯 GPT-6 Astra Is Out
@willfrancis24 GPT-6 Astra is here. The next generation of OpenAI's large language model has landed. I've had early access and been pushing it to its lim... See more
☕️ THE EXTRAS
One year ago today, I wrote Atlassian Buys The Browser Company
The most clicked link in last week’s newsletter: OpenAI to end model access to Cursor after acquisition by Elon Musk’s SpaceX
Have a burning AI question?
I read every single question submitted and use it to improve this newsletter. Submit your question below!
That’s it for this time! See you next week.
— Cliff Worley