Secrets Work Exponentially
Knowledge build gradually.
Secrets work exponentially.
The difference between knowledge and secrets defines success at the highest levels. Often, individuals with great talents may acquire a vast amount of tactical information but fail to unlock the unique insights. Knowledge is replicable, while secrets represent unmatchable advantages.
Knowledge, at its core, refers to information that is known by someone and can be obtained through research or conversations with the right individuals. Examples of knowledge include the intricacies of hiring and firing top employees, the methods for building a better software, and the techniques for creating a more cost-effective M4 Max chip. It is often closely guarded and only shared among a select group of individuals. But the information is ultimately discoverable by others who put in similar effort.
Secrets, on the other hand, may not be known by anyone. These are answers to questions that have yet to be asked, or, most important, have not been asked in the appropriate manner. Secrets may also be answers that require deep insight and experimentation to uncover. Unlike knowledge, secrets are not discovered through conversations but through risky experimentation and instrumentation. SpaceX discovered that rockets could land vertically. Apple realized phones didn’t need keyboards. These weren’t just clever conversations.
To excel success at the highest level, learn to see what others miss. The next breakthrough won’t come from known knowledge – it will come from uncovering what others haven’t yet imagined to look for.
Best example in Dec 2024
Deepseek’s V3 model can compete with the world’s most advanced AI systems, with a total training cost of just $5.6 million USD. DeepSeek’s approach shows that building cutting-edge AI does not always require massive GPU clusters – it is more about using available resources efficiently. It alos shows how chip export restrictions can actually drive innovation. DeepSeek’s limited access to high-end hardware forced them to think differently.
it is (relatively) easy to copy something that you know works. it is extremely hard to do something new, risky, and difficult when you don't know if it will work. individual researchers rightly get a lot of glory for that when they do it! it's the coolest thing in the world.
Started a project to scrape exec compensation for ~10,000 public companies from the SEC. Wrote the scraper code fully on Claude, but each filing has a custom format. So I piped it all into DeepSeek V3 to extract structure. Parsed 6M tokens so far and it cost.. $0.50! New era.
on Artifact
