MC.54: OpenAI's o1 dominates + Alter out of beta!

Hey everyone,

I've got some exciting news to share about my MacOS app Alter. We've officially moved out of beta and are now fully available to everyone!

Here are a few cool things you can now do with Alter:

• Create mini-apps called Artifacts - like interactive charts, diagrams, and presentations. You can even share them online with just one click. Watch the video 🎥

• Use AppSense to grab images and content from your browser and use them in your AI conversations. It's like giving Alter x-ray vision for your apps.

• Your chats are now encrypted and stored locally. Finding old conversations is super easy with the new search feature.

• Add screenshots to any action with just a click. Works great with the latest AI models that can analyze images.

• Pin your favorite actions to the top of the menu for quick access.

• There's a new minimize mode so Alter can keep working in the background while you do other things.

We've also made Alter way smaller - it's now just 45MB instead of 500MB! Loads faster too.

I'd love to hear what you think about these updates. Alter is all about making AI work seamlessly on your Mac, so your feedback helps make it even better.

👉 Give the new version a try and let me know how it goes!

Now let’s talk about OpenAI and their latest achievement: models o1

Both of their models, o1-preview and o1-mini, just stormed into the Chatbot Arena and snagged the top spots. They're not just good - they're scary good at math, hard prompts, and coding. They are almost 100 points in rating ahead of Claude 3.5 Sonnet.

o1 uses something called "chain-of-thought reasoning." It's like the AI is thinking out loud, step by step. This lets it tackle super complex problems that other AIs struggle with.

Sam Altman, was excited about how well o1 did on their "goal 3" - which is agents solving complex tasks. He said it outperformed way beyond what they expected.

As o1 models were rate limited until now, I haven’t been able to feel such a step forward. On coding tasks where Claude 3.5 Sonnet, o1-mini failed just as well. Now that OpenAI is raising the rate limits, more people will be able to test on real use cases.

Sam

Enjoyed this newsletter? Forward it to a friend and have them sign up here.

Until next Thursday 🎉