Pinned Tweet
Introducing Codex for marketing
This November, I’m building what I hope will become the best AI marketing agent in the world.
The goal is simple: one agent that can help founders with everything marketing.
Today, we’re launching V1. It can create:
• Slideshows
• Wall-of-text videos
• Hook + demo videos
This is just the beginning.
I’ll be shipping a new feature every day until the end of November, and I want to build this with you.
What marketing feature would you want me to add next?
What would make you switch from Claude Code to Codex?
1. Better code quality
2. More generous limits
3. A lower subscription price
6
4
359
What would you most like to see improved in GPT-6.2 Sol?
1. Better coding ability
2. Faster responses
3. Higher usage limits
1
352
If OpenAI could improve only one thing in GPT-6.2 Astra compared to GPT-6 Astra, what would you choose?
- Better reasoning
- Stronger coding ability
- Faster responses
7
9
1,856
Stop telling Claude Fable 5.2, “build my app.”
Stop telling it, “get me users.”
Stop telling it, “make this better.”
Give it the context and a clear goal instead.
Here are 4 prompts you can copy and paste:
1
3
396
1. Turn your idea into a real MVP
“Help me turn this idea into a first version I can actually launch.
Idea: [describe it]
Target user: [who it’s for]
Time and budget: [your limits]
Identify the core problem and one workflow that solves it. Separate the features I need for launch from the ones that can wait.
Give me an ordered building plan, a simple architecture, and a way to test whether the first version works. Flag your assumptions and any security risks I shouldn’t ignore.”
1
87
If OpenAI releases GPT-6.2 Sol and Astra soon, the comparison with Anthropic is going to be interesting.
But I don’t just want to see benchmarks.
Give the models the same real-world tasks: build a feature, find a difficult bug and work through a large codebase.
Then compare how long each takes, how many mistakes they make and how much human help they need.
That would tell us a lot more about which model is actually better for builders.
What’s the first task you’d use to compare them?
2
17
2,170
Anthropic just released Claude Fable 5.2 and Mythos 5.2.
Anthropic made the first move. Now all eyes are on OpenAI.
There’s been a lot of speculation about GPT-6.2 Sol and Astra, but OpenAI still hasn’t confirmed when they’ll be released.
Will OpenAI respond with its new models soon, or will we have to wait longer?
I’m especially curious to see what Sol can do for coding. Can it handle larger codebases, follow instructions more reliably and build complex features with fewer mistakes?
And how will Astra compare when it comes to reasoning and real-world tasks?
The AI competition is getting more interesting by the day.
Which are you more excited to try: Anthropic’s new models or GPT-6.2 Sol and Astra?
GPT-6.2 Sol and Astra could drop tomorrow.
Nothing has been officially confirmed yet, but if OpenAI releases both, this could be a huge week for AI.
There’s also speculation that Fable 5.5 is ready and that Anthropic is waiting for OpenAI to release its models first.
Meanwhile, some think OpenAI is waiting for Anthropic to make the first move.
If that’s true, both companies could be waiting on each other right now. But for now, that’s just speculation.
I’m especially curious to see how much better Sol will be at coding. Can it handle bigger codebases, follow instructions more reliably, and build an entire feature without constant fixes?
And how will Astra compare to Anthropic’s best models when it comes to reasoning?
Benchmarks are interesting, but what matters most is how these models perform in real work.
If Sol and Astra do drop tomorrow, which one are you trying first?
3
18
4,766
If you want to compare Codex and Claude Code, give them both the same task.
Same repo context. Same requirements. Same tests.
Then compare the result, the time you spent reviewing it and what still needed fixing.
2
1
549
A $50 Codex Pro plan sounds interesting.
But what would it actually need to include to make you upgrade from Plus?
1
2
350