Silicon Valley has welcomed a new genius. At just 25 years old, Michael Truel has created a code editor valued at $300 million, making GitHub Copilot look like a toy. By eliminating the tedious and repetitive parts of coding, he enables developers to complete in minutes what used to take days. This is his story: 🧵

Meet the MIT dropout team leading the Cursor revolution: Michael Truel, Suare Asif, Aman Sangh, Avid Lunemak This team of around 60 people has built a programming technology company valued at $2.6 billion. They developed customized AI models that have increased engineers' productivity by 10 times.

The CEO of Cursor, Michael Truel, is not just looking to create a better code editor. His goal? To invent a whole new way of programming. "A post-code era... in this era, engineers are more like doing logical design."

But the biggest misunderstanding from the outside about Cursor is: This is not just a tool wrapped in a ChatGPT shell. Every magical moment in Cursor is backed by their independently developed AI model — something they never expected to do when they first started their business.

The current AI programming field is trapped in two erroneous thinking frameworks: 1. "The future is just like the present, only with AI assistance added." 2. "If you just chat with the robot, it can automatically build everything." Cursor believes that both of these assumptions are fundamentally wrong.

The programming of the future will not be languages like TypeScript or Python. Nor will it be chatbots (because they are too imprecise). It will be a brand new way of expressing software logic, more like English — a high-level pseudocode that can be freely modified and controlled.

In this new world, what ability is the most important? It is "taste." "The work of engineers will be more like that of logical designers... you need to clearly express how you want each part to operate."

What is most unexpected? Engineers can actually be less cautious. Current programming requires extremely high accuracy and attention to detail. In the future, engineers will focus more on overall design decisions rather than on the details of implementation.

So where is the moat for AI? "This might be a bit harsh for people like us, but it's a good thing for the world: there are still many technologies in this industry that can be breakthrough." This market is not suitable for those who are content with the status quo, as there is simply too much room for development.

Why hasn't Microsoft's Copilot dominated the market? 1. This market is not friendly to existing giants. 2. Users can easily try other tools and determine which is better. 3. The original Copilot team has already dispersed and moved to other companies.

What is the biggest lesson they learned? "Many companies hire too quickly, while we actually hired very slowly at first." They intentionally delayed expanding the team to find the truly right people — those who are both curious and rationally honest.

Will AI replace engineers? Not in the short term. "The demand for software is long-term... If you can reduce development costs by ten times, we can do more on computers."

Want to make good use of tools like Cursor? 1. Break tasks into smaller chunks instead of throwing a large block of prompts at the AI all at once. 2. Clearly test the boundaries of the AI's capabilities (preferably on side projects rather than in production environments). 3. Cultivate judgment to understand which methods are effective and which are not.