Min Choi Profile picture
AI Educator. 𝕏 about AI, solutions and interesting things. Showing how to leverage AI in practical ways for you and your business. Opinions are my own.
25 subscribers
Dec 13 12 tweets 4 min read
Anthropic literally dropped the Prompt Guide to 10x your Claude Opus/Sonnet 4.5

10 best practices you should know. Bookmark this. 👇 1. Be Painfully Specific

Say exactly what you want (format, tone, length, audience). Claude 4 follows instructions literally. Image
Dec 12 8 tweets 2 min read
Google Labs just dropped Disco + GenTabs yesterday

Disco is an AI browser and GenTabs looks at your open tabs + chat, then builds a "mini app" to finish the job (trip planner, study tool, meal plan, etc.)

6 examples + how to join the waitlist👇

1. Trip + activity planner 2. Interactive learning tool - solar system
Dec 11 9 tweets 3 min read
This is wild.

OpenAI just dropped GPT-5.2

A frontier model built for actual work, not just shiny demos.

It beats human experts on ~70% of real job tasks, hits 100% on AIME 2025, and is built for long-running agents. 🤯

People are already creating wild examples👇 Image 1. Ocean simulation in one shot

PROMPT: "create a visually interesting shader that can run in twigl-dot-app make it like an infinite city of neo-gothic towers partially drowned in a stormy ocean with large waves,"
Dec 11 11 tweets 3 min read
Nano Banana Pro is the best AI image model right now

People are finding insane ways to create 3D renderings, impossible shots, and ultra realistic AI influencers.

10 examples + prompt. Bookmark this 👇

1. Nano banana -> World Labs -> WebAR (threejs)
2. Ultra realistic AI influencers
Dec 7 12 tweets 3 min read
Nano Banana Pro is pretty wild.

People just dropped some new insane use cases

10 examples. Bookmark this. 1. 3D visuals on urban LED display
Dec 2 11 tweets 6 min read
It's so over

These are not real people

100% AI Nano Banana Pro

1. "Generate a photorealistic image of a cafe terrace in the Marais district of Paris on a Wednesday morning in March 2025. It is a crisp, cool spring morning with clear skies. Locals are drinking coffee. In sharp focus should be a young woman with a pixie cut wearing a scarf, stirring a cappuccino and looking thoughtfully to the side; the waiter and street traffic behind her are blurred. The photo should have the candid, natural morning light feel of an iPhone image."Image 2. "Generate a photorealistic image of an outdoor farmers market in Phoenix, Arizona on a Saturday morning in June 2025. It is already very hot and bright desert sun beats down. In focus should be an elderly man wearing a straw hat, examining tomatoes at a stall; the other shoppers and vendors are blurred. The photo should look like a bright, slightly overexposed iPhone photo taken in harsh sunlight."Image
Nov 29 12 tweets 3 min read
Nano Banana Pro by Google is insane.

But it doesn't just generate AI images, it can do so much more🤯

10 wild use cases. Bookmark this🧵👇 1. Fake Instagram photos
Nov 28 6 tweets 2 min read
AI is unlocking pro filmmaking.

This Heineken spec ad is 100% AI by Simon Meyer.

No shoot, no crew, just with Agents & Models on Invideo

5 wild examples + how to: 👇 2. Generate, camera control, characters with Agents & Models
Nov 27 11 tweets 4 min read
Google Nano Banana Pro realism is going viral

This is 100% AI

10 wild examples:

1. New Year's Eve that hasn't happened yet Image 2. PROMPT: "Generate a photorealistic image of Time Square on a saturday in Fall 2006, it's a beautiful late October day, people are busy. in focus should be a young blonde - rest can be blurred. the photo should be reminiscent of that a digital camera from 2006 would take, with a timestamp like a printed photo would have."Image
Nov 25 11 tweets 3 min read
It's only been just 26 hours since Claude Opus 4.5 dropped.

And people can't stop getting creative with it.

10 wild examples:

1. Opus 4.5 + CUA: "file my taxes end to end make no mistakes"
2. Vibe code app to show real-time arrival of next metro
Nov 18 11 tweets 3 min read
Ok Gemini 3 is insane.

People can't stop just build things

10 wild examples:

1. Theme Park game from 90s remade in hours
2. Build fun games in minutes
Nov 16 12 tweets 3 min read
Invideo just dropped 30+ Black Friday ad trends.

No shoot, no editor, no prompt, just a photo.

Drop your brand pic and get a full viral video ad in seconds.

10 wild examples + how to try: 👇 2. Hot Surprise
Nov 7 9 tweets 3 min read
Qwen Image Edit w/ Camera Control is wild 🤯

Quickly rotate the camera, switch between bird's eye and worm's eye views using just clicks.

Here's how plus 7 wild examples:👇 2. Cinematic shots
Oct 14 13 tweets 3 min read
OpenAI literally dropped ChatGPT Prompt Packs for Any Role

11 roles. Bookmark this. Image 1. ChatGPT for any role
academy.openai.com/public/clubs/w…
Oct 10 7 tweets 3 min read
Ok, this is wild.

Invideo + Sora 2 recreated this iconic KFC-style spec ad.

No camera, no crew. Just prompts. 🤯

Step by step. Bookmark this. Step 1: Create the full base video

In invideo, use Sora 2 to create a full video from a prompt.

Add reference files.

In this example, 4 prompts were used. One base timeline plus clips to fill it.

Cost: ~$100 in credits.
Oct 9 12 tweets 3 min read
Higgsfield AI Trends just dropped.

Powered by Sora 2, now anyone can make viral celeb look-alikes, spec ads videos even easier.

10 wild examples + how to try: 2.
Oct 6 10 tweets 3 min read
Ok this is insane.

AI social apps are exploding.

Pika Predictive Video just dropped.

Type a thought → get an AI video. Pika fills script, music, lighting, camera moves. On iOS now.

8 wild examples + prompts & how to try:

1. "im rapping in a studio, fish eye camera" 2. "im an olympics swimmer, doing an olympic dive"
Oct 5 12 tweets 3 min read
Sora 2 videos are going crazy viral.

Insane cameos, anime episodes, vlogs from text & images

10 wild examples: 1. Mario Kart police chase, Remix to continue next scene
Oct 5 9 tweets 3 min read
Sora 2 Pro can generate some wild YouTube and TikTok influencer videos.

These are not real 🤯

8 examples + prompts:

1. Selfie vlog, girls' night out, Vegas 2. Pet vlog
Oct 3 10 tweets 3 min read
This is wild.

Hume AI just dropped Octave 2.

Ultra realistic text-to-speech AI model in 10+ languages with multi-speaker + voice cloning.

100% AI

8 wild examples + how to try: 2. Sarcasm
Sep 30 11 tweets 3 min read
This is wild.

OpenAI just dropped Sora 2 that generates realistic, controllable video with lipsync and sound.

The Videos look insane.

100% AI

1. Step into any world 2. Sora 2 can do gynmastics