Nugget #01
Parallel AI development, a faster coding model, OS‑level ChatGPT, and a maker case of shipping controversy into revenue.
Coffee. No sugar. Let's go.
🔥 What's happening
Cursor 2.0
Cursor released 2.0 this week.
The new agent UI focuses on outcomes, not files. You can run five agents in parallel without git conflicts. Each gets its own worktree. No merge hell. No wondering who changed what.
The interesting part: Model ensemble workflow. Multiple AIs tackle your problem simultaneously. Claude builds one approach. GPT builds another. You choose what works.
Browser testing is built in. Agents can now test their own work. Write code, break it, fix it, repeat. No babysitting required.
Overkill for a todo app? Sure. But when you're fixing payment bugs at 3 AM, you want solutions, not elegance.
Here's the thing: we've been using AI assistants wrong. One model, one attempt, cross your fingers. Cursor said no. Let multiple models compete. Use what works best.
Stop treating AI like magic. Treat it like scalable intelligence you can parallelize.
Composer 1
I was debugging when I spotted something odd in Cursor's dropdown menu. "Cheetah." What kind of model name is that?
Turns out everyone saw it. Nobody said anything.
The reveal: Cheetah was just Composer 1 wearing a fake mustache. They were testing it in production.
It's 4x faster than similar models. Handles large codebases without choking. Semantic search that actually finds what you need. Most requests finish in under 30 seconds.
Is it perfect? No. Will it replace every tool you use? Probably not.
But here's what matters: While we're busy comparing AI assistants, Cursor shipped something that works. Not in some future roadmap. Now.
The honest truth? Most of us spend more time researching tools than using them. Composer 1 isn't revolutionary. It's just solid engineering that shipped.
Stop hunting for the perfect AI assistant. Pick something that works and use it.
Software Applications Incorporated
OpenAI just bought the makers of Sky. Another acquisition? Not quite.
Right now, ChatGPT is trapped in a browser. Can't touch your files, can't see your screen, can't actually DO anything except generate text. You ask, it answers, you copy-paste.
Sky changes that. Deep Mac integration means ChatGPT can escape the chat window. No more explaining context. No more manual execution of every suggestion.
Nick Turley spelled it out: They're building ChatGPT that acts, not just responds. Sees your calendar, writes in your docs, codes in your IDE. Without you playing middleman.
Here's the thing: 800 million people use ChatGPT like a fancy search engine. Ask, read answer, do the work yourself. This acquisition is OpenAI saying that workflow needs to end.
The honest truth? Every AI company is building smarter models. OpenAI just bought the connection between AI and your actual computer. While others optimize intelligence, they're building integration.
ChatGPT won't just be a website. It'll be part of how your Mac works.
🛜 The signal
ChatGPT can explain quantum physics but can't ...
Get the full Signal + exclusive content
🚀 Maker's Log
This is about speed.
Pieter Levels tweets. Calls out the "$100K MRR in weeks" claims. Says screenshots can't be verified. Looks like marketing, not real numbers.
Marc Lou sees the tweet. Doesn't debate. Ships.
The timeline: Levels tweets frustration. Marc codes TrustMRR. Simple MVP: verify revenue with read-only Stripe API keys. Pull numbers server-side. Give each product a public page. Done.
Then Marc adds ad spots to the leaderboard. They all sell. Fast.
$13,883 in revenue. Within hours of the original tweet.
The leaderboard fills up. DataExpert, Kibu, ShipFast, Launch Club. Founders verify their numbers. Pay for visibility. The controversy becomes the distribution.
Marc didn't build something complex. He built a solution to a fresh problem. While others argued in replies, he was processing payments.
Why this matters
- Speed beats perfection. Marc shipped while people were still talking. The debate was free marketing.
- Controversy is distribution. A polarizing topic means built-in attention. Every argument drove traffic.
- Monetize immediately. Ad spots on day one. No "figure out revenue later." Cash from the start.
Builder Takeaways
- Watch for public pain. When people with reach complain, that's your product spec.
- Ship during the conversation. Not after. During. Be part of the moment.
- Sell before you scale. Marc sold ads before adding features. Revenue validates faster than user counts.
From complaint tweet to $13,883. No investors. No roadmap. Just execution speed that turned debate into revenue.
The window was hours, not months. Marc saw it. Shipped it. Sold it.
Do you need a SaaS MVP? Not prompted, built. Tight architecture, readable code, reliable results.