Luca Moretti

Before building an app, ask Claude to find any MIT GitHub repos for the features.

90% has already been built for you.

This community is going to be so cool because every proposal from the members will turn into a project within it.

Absolutely anything available on Whop—trading, football, AI models, social media automation, business automation, starting a company, managing a SaaS, building an app or website, running a faceless YouTube channel—whatever exists and whatever you propose, I will teach you step-by-step until you make it happen.

My goal is for this community to cover everything, so you only have to pay in one place to get all the information you need.

*I don't like to lie, so I can't guarantee you'll make a specific amount of money, but I can guarantee that if you actually apply what I'm telling you step-by-step, you will inevitably make money.

At one point, I examined absolutely every platform offering free models and concluded that 99% of them:

1. claim otherwise, yet they are actually just old open-source models

2. say they don't collect your data, but inevitably save it

3. look identical to one another and lack quality

4. simply mirror what OpenRouter publishes (there are legitimate platforms like Kilocode—recently acquired by Anaconda—Vercel, Novita, OpenCode, and Cline; Blackbox used to be one before shifting focus solely to enterprise; now there's Nous Research with the trending Hermes model, plus many smaller ones offering a few minutes or a few million tokens to play with)

5. offer far too little volume to actually matter

I should also mention Nvidia's build here; it features top-tier models, but they simply don't work—or rather, they respond incredibly slowly.

It’s pointless showing me Opus 5.5 if I’m accessing GLM 5.2, and seeing Kimi K3 doesn't help if I'm limited to one request per minute or a maximum of 1 million tokens.

For context, sites like codecraftapi.com and https://cleanapis.com/ have only been around for a few days, yet they are being heavily—and pointlessly—promoted on X.

Yesterday, Muse Wonder was released in two versions; one is called "Full" and was marketed as being slightly less powerful than Opus 5.5, yet extensive user testing shows it doesn't actually live up to that claim.

In reality, it falls far short of Astra and Opus 5.5, and Meta’s $200 subscription offers roughly 12 blocks of 5 hours per week—meaning a maximum of four days of usage, whereas actual work requires two.

The design aspect is a disaster; Meta seems to be focusing heavily on coding, much like OpenAI focuses on the hard sciences and Anthropic on the anthropomorphization of AI.

The price is far too high for this model—exceeding that of Sonnet 5, which, in my view, offers the best value for money.

Google has released Argon, though it isn't public yet; we’re all waiting to test it next month. In the meantime, we’ll likely see many open-source models reaching the same level.

As of today, October 13, only Anthropic and OpenAI have public models with an intelligence index above 50 (as I said, Meta and Google aren't worth mentioning; they’re useless).

In the 40–50 intelligence index range, we find the famous Muse Spark 1.3, Step 5 Preview, Ling 3.1 Flash, DeepSeek V4.1 Flash, Kimi K3, Grok 4.7, Mimo V2.6 Pro, GLM 5.3, and Qwen 3.8—plus various "Flash" and "Pragmatic" variants—spanning about 8–10 companies (including Mistral or Nvidia, which are biding their time, and other companies aiming directly for ASI).

Over the next two and a half months—and by the end of the year—these approximately ten companies will release at least one model with an intelligence index exceeding 60 as open source, capable of running on a personal PC or laptop.

My goal is to post content daily and be available to members 24/7; I am currently trying to document at least 1% of what I know—knowledge that represents 99% of what isn't shared publicly.

My aim with this community is for you to build friendships and make money—first individually, then together—and to grow into much larger, more prosperous groups.

Do you know the most-watched show on the planet?

Joe Rogan has spoken with approximately 1,200 people across 2,560 episodes, likely covering every conceivable human category. I will teach you how to create bots that embody the person in question—using all their accumulated knowledge to respond to you exactly as that person would.

This applies to anyone with online or physical content (whether they’ve made videos or written books).