The Signal Layer: What to Build When Anything Can Be Built — Lena Hall, Akamai
In an era of 'drowning in abundance' where AI makes virtually anything technically buildable, the value has shifted from implementation speed to defining and protecting a unique signal. The talk introduces the concept of the 'signal layer,' which involves two parts: clearly defining your unique signal (the 'build side') and ensuring it survives transmission without distortion (the 'ship side'). Since broad taste is trainable by models, true differentiation requires judgment about unobserved relationships or future events—areas where AI cannot yet operate. The ultimate goal is building trust, which must be actively engineered through careful product design and go-to-market strategy.
Key takeaways
-
The Value Shift
2:00
Because automation has driven the cost of average work to zero, the superpower is no longer using AI, but deciding *what* problem deserves an attack. The scarce skill is choosing which problem to focus on.
-
The Signal Layer
5:50
Differentiation requires a 'signal layer'—a deliberate function ensuring the customer's understanding of your product matches your original intent. This involves defining what makes you unique and protecting that message through all channels.
-
Judgment vs. Taste
11:53
Broad 'taste' is merely preference under feedback, which AI systems can learn. True differentiation comes from judgment about things that have not happened yet (no data exists) or insights into unobserved customer relationships.
-
Hamming's Principle
13:54
A problem is only important when there is a reasonable attack on it. Since AI provides an 'attack on everything,' the critical task is identifying which problem warrants that effort.
-
Signal Distortion Fixes
Signal can break in three places: Source distortion (founders compressing context), Organization distortion (signal getting averaged through management layers), and Machine distortion (AI remixing the message into formats like tweets or one-pagers).