Welcome Back Everyone! It’s Sunday, Aug 17th. (& I’m exciteddd)

Do You Know?

GPT-4o Returns After GPT-5 Backlash

OpenAI has reintroduced GPT-4o following backlash over GPT-5. Users criticized GPT-5 for its formal tone and shorter responses, preferring the warmth of GPT-4o. In response, OpenAI CEO Sam Altman announced GPT-4o’s return for Plus subscribers, with GPT-5 limits also increased. This move highlights the importance of user experience over mere technological advancements.

On that note, Here is today’s menu:

1. AI Spotlight: GPT-5 One Week In 🤯

2. Tool Talk: Luminar Neo 🪄

3. Prompting Curiosity: Metaprompting with GPT-5 📝

GPT-5 One Week In 🤯

Link: Sam Altman Announces Major Changes to ChatGPT | Sam Altman on X​

Why it matters:
​GPT-5's launch has been rocky to say the least, with the general consensus being that GPT-5 shows improvement in key areas like coding, but it falls short in too many other areas.

OpenAI listened to the feedback, and on Tuesday of this week, Sam Altman posted on X announcing some major changes to ChatGPT.

First up: ChatGPT now has a new setting that lets users toggle between three GPT-5 models. ChatGPT will still use the "Auto" setting by default, but now users can manually select the "Fast" or the "Thinking" versions of GPT-5 for any prompt.

In his post, Altman clarified that the Thinking model has a 196k token context window and now has a 3,000 message weekly cap. Once users hit that limit, ChatGPT will switch to a lightweight variant called GPT-5 Thinking Mini.

Altman also had great news for those mourning the loss of the legacy ChatGPT models like GPT-4o and o3: paid ChatGPT users can now find those legacy models in ChatGPT by enabling the "Show Additional Models" setting in ChatGPT. (GPT-4.5 remains a Pro-only model)

Finally, Altman teased a new personality update coming soon to GPT-5 that promises to make it warmer and less robotic than the current default, but not quite as verbose and sycophantic as GPT-4o. He mentioned that OpenAI's long-term goal is to move towards full per-user personality customisation.

So if you miss the old models like GPT-4o, go check out your ChatGPT settings to get your old favorites back!

And stay tuned, because all signs point to OpenAI continuing to refine and improve GPT-5 until it matches the lofty expectations placed on it.

Tool Talk: Luminar Neo 🪄

Why It’s Worth Your Time

• AI sky replacement & reflections swap skies with realistic lighting and easy fine-tuning

• Generative AI tools remove or replace objects seamlessly for creative edits and repair

• Portrait enhancements refine eyes, face shape, and lighting with intelligent selections

• Relight AI adjusts lighting based on 3D image depth for natural exposure fixes

• One-click presets speed up edits with ready-made looks tailored for landscapes and more

• Extensions for pros include HDR merge, upscaling, and focus stacking

Prompting Curiosity: Metaprompting with GPT-5📝

Prompt Formula:
​When asked to optimise prompts, give answers from your own perspective - explain what specific phrases could be added to, or deleted from, this prompt to more consistently elicit the desired behaviour or prevent the undesired behaviour.
Here's a prompt: """[PROMPT]"""
The desired behaviour from this prompt is for the agent to [DO DESIRED BEHAVIOuR], but instead it [DOES UNDESIRED BEHAVIOuR].
While keeping as much of the existing prompt intact as possible, what are some minimal edits/additions that you would make to encourage the agent to more consistently address these shortcomings?

Quick Description:
​Ever write up an incredible prompt, send it to an AI, and it just...doesn't do what you asked it to do? That's a very common occurrence, and often fixing the issue is as easy as adding or removing a few words or phrases from a prompt.

As it turns out, ChatGPT with GPT-5 is a great prompt engineer! You can use this metaprompting formula to have ChatGPT fix any prompt that isn't quite up to par.

How to Use:

1. Copy the above prompt formula into ChatGPT.

2. Replace [PROMPT] with the prompt that you want to improve. (Do not remove the triple quotation marks!)

3. Replace [DO DESIRED BEHAVIOuR] with a description of what you want your AI to do.

4. Replace [DOES UNDESIRED BEHAVIOuR] with a description of what you want it to do instead.

5. Send the prompt!

ChatGPT will output an improved version of the prompt that more closely aligns with your goals.

This prompt comes directly from OpenAI's new GPT-5 cookbook, which is an incredible resource for improving your communication with modern AI systems. It's dense and fairly technical, but it's a must-read if you truly want to improve your prompting skills with GPT-5!

That's all for this week folks…

What did you think of today's email?
Your feedback helps me create better emails for you! 😃

Stay Curious, Hustlers!

See you next Friday.
Much Love, Vasu