agree its not a drop in replacement i actually bet codeact would use less tokens. function definitions and generations count as tokens. also codeact can potentially require fewer LLM calls which massively saves on tokens. Why do you think otherwise? why do you think ReAct is
PROMPT ENGINEERING
-
Jonny Miller Uses ChatGPT as Personal AI Coach with Codex Vitae
By
–
Jonny Miller (@jonnym1ller) uploaded his entire life to ChatGPT to use it as the ultimate AI coach.
— Dan Shipper 📧 (@danshipper) 26 mars 2025
He created what he calls a Codex Vitae—with core personality traits, values, goals, burnout signals and more to load into ChatGPT. It hyper-customizes his responses, to help him… pic.twitter.com/8u6mxcJMKMJonny Miller (
@jonnym1ller
) uploaded his entire life to ChatGPT to use it as the ultimate AI coach. He created what he calls a Codex Vitae—with core personality traits, values, goals, burnout signals and more to load into ChatGPT. It hyper-customizes his responses, to help him -
Using ChatGPT 4o to Generate YouTube Thumbnails
By
–
ChatGPT 4o can now generate INCREDIBLE YouTube Thumbnails! Prompt I used:
"Create a 16:9 YouTube thumbnail of my mascot with his hands on his face acting surprised and on the right side ChatGPT window with image editing features, with big red text saying "MASSIVE ChatGPT 4o -
Vibe Coding Deathmatch: Bolt, Replit, Cursor, Windsurf Compared
By
–
Vibe Coding Deathmatch
— Ben Tossell (@bentossell) 26 mars 2025
wednesday afternoon full video
Bolt, Replit, Cursor, Windsurf head to head building the same project
simple prompts across all, the results were:
🥉 Bolt + Windsurf
🥈Cursor
🏆 Replit
Reasons:
– Bolt was quickest, got into a 'no-audio' issue early… pic.twitter.com/2DEPJVgmU7Vibe Coding Deathmatch wednesday afternoon full video Bolt, Replit, Cursor, Windsurf head to head building the same project simple prompts across all, the results were: Bolt + Windsurf
Cursor Replit Reasons: – Bolt was quickest, got into a 'no-audio' issue early -
ChatGPT Image Generation Upgrade and Prompt Example
By
–
BREAKING: Image generation in ChatGPT just got a HUGE upgrade! Prompt I used:
"Create image of phone photo of friends chilling at a party, taken with a film camera" -
AI Model Prompt Adherence with Large Scale Objects and Text
By
–
The prompt adherence is carzy, you can just throw an insane amount objects, and still get it right, plus handling text.
-
Motorbike Example Outputs and LoRA Customization Options
By
–
Here's one of the motorbike example outputs you can tweak:https://t.co/PabxOUElcy
— Replicate (@replicate) 26 mars 2025
And those popular squish and rotate wan effect loras work too. Grab the safetensors links from HuggingFace and use them in the lora_url fieldhttps://t.co/l767ZYm5AQ pic.twitter.com/4sq08bG9woHere's one of the motorbike example outputs you can tweak: https://
replicate.com/p/54ck050b41rm
c0cnskrsg4xjqc
… And those popular squish and rotate wan effect loras work too. Grab the safetensors links from HuggingFace and use them in the lora_url field https://
huggingface.co/collections/Re
made-AI/wan21-14b-480p-i2v-loras-67d0e26f08092436b585919b
… -
Detailed Prompt for Photorealistic Ramen Image Generation
By
–
A photorealistic close-up of a bowl of ramen with steam rising, tiny droplets of broth splashing as chopsticks pick up noodles. Include detailed reflections on the bowl and accurate lighting on the hand holding the chopsticks.
-
Detailed Prompt for AI Image Generation
By
–
An old wizard reading a glowing spellbook in a dark forest. The book should have readable text in an ancient-looking font, surrounded by floating magical symbols. Realistic lighting from the spellbook illuminating the wizard’s face and hands.
-
Prompt engineering and output of a generative AI image
By
–
A hyper-realistic photo of an astronaut standing in a medieval castle courtyard, holding a glowing sword. One hand visible with accurate fingers, sunlight casting long shadows. Fusion of sci-fi and fantasy with depth and contrast.