Lab31 released 31 AI projects in 31 days — Here’s how

Two creative leads, Jenny Nicholson and Allister Hercus, met on LinkedIn and decided to do something ambitious: release a custom ChatGPT every day in January 2024.

This year, they released 31 projects made with OpenAI’s suite, Anthropic’s Claude, Google Gemini 2.0, DALL-E 3, Grok by X, Luma, and more.

The project, known as Lab31, includes: 

• Shame Roulette: Anonymously share a secret and receive one from another participant.
• Creative Seance: Summon the AI spirit of a noted creative for input on your ideas. 
• Golf Sim 2.0: A golfing simulator game. 
• Life Stats: An immersive experience that estimates how many times someone’s fallen in love with you or how many bugs you’ve eaten. 
• The Attention Audit: A browser extension that analyzes your history to figure out where your focus went each day. 
• Pocket Guide: An audio assistant that tells you about your surroundings so you’re not stuck staring at your screen. 

The hardest part…

… isn’t working with AI, the pair said, but coming up with an idea. Once you have that, Hercus and Nicholson suggest explaining that idea to an LLM, then asking it — as you might a human collaborator — how to execute it.

For example, you might be a writer or an artist, but not an app developer. An LLM could help you figure out what software tools could expedite tasks and how to use them, cutting the time it takes to make a project to mere hours.

Some more takeaways

• Get weird. Maybe the model can’t help you, but give it a shot.
• Models return precisely what you ask for — and if not, it’s a sign that you didn’t ask for what you thought you did.
• You can never give a model too much info.
• The more creative, helpful, and invested you are when communicating with an LLM, the more that will reflect in its response.

“Our shorthand is always, don’t use AI to do what you’ve always done, just faster and cheaper without humans, [but] what happens when you use AI to do things that you couldn’t have done before?” Nicholson said.

“We feel… that the most important skillset of the next several years isn’t going to be knowing [specific AI tools or having a specific technical background, but] having a mindset of, ‘I don’t know, but I bet I can figure it out.’”