# GOTO Serverless - Bangalore

*   Serverless opens up business opportunities that weren’t imagined before.
    
*   Serverless means no infrastructure to manage
    
*   With serverless, pay for value; this is good for prototyping and bootstrapping an idea
    
*   AWS serverless launched in 2014
    
*   Core principle of serverless - event driven architecture
    
*   Most software is event driven
    
*   AI agents are going to be big - Nick coult, Director of product, Serverless compute @ AWS
    
*   Focus should be on the end business goal and not on the input like which language or tech stack
    
*   Getting hands dirty and jumping in is the best way, not to wait and plan
    
*   For tech stack modernization initiatives, create a phased migration strategy that breaks down changes into small, digestible increments
    
*   AWS Step functions, AWS Distro - open telemetry
    
*   LLM Judge - use one LLM to critique another’s performance
    
*   MCP, an open protocol that standardises how applications provide context to LLMs.
    
*   MCP: USB-C port for AI applications
    
*   Cold starts is one of the most significant architectural challenges in serverless computing, where functions experience latency delays when invoked after periods of inactivity.
    
*   Limitations of LLMs
    
    *   Trained in the past ( they don’t know the current day’s stock market )
        
    *   Are stateless. They don’t retain memory between interactions
        
    *   Can generate content, but they cannot take actions
        
    *   Aren’t trained on AWS scale problems ( Trillions of lambda invocation / month )
        
*   Forcing function: A mechanism that compels specific behaviour by making alternative actions difficult
    
*   Very challenging to modernise AWS services, ECR took 5 years to rewrite.
    
*   Bleeding edge of AI: Automated reasoning