In the process of making something useful, there are multiple kinds of abstractions created, use them but know the fundamentals. Abstractions make usage easier but knowing fundamentals always help you reason from first principles.
In physics, Newton’s laws of motion and gravity are the fundamentals. Theorems on energy conservation and planetary mechanics etc are just high-level derivations – abstractions built on top of them.
In computing, assembly and machine code are the fundamentals used to interact directly with hardware. High-level languages like COBOL, C++, and Python act as abstractions. They make programming user-friendly, but that convenience costs efficiency.
In GenAI, for example, MCP is also a kind of abstraction layered over raw APIs. It makes tool usage seamless for LLMs, but that convenience has a price tag : it can bloat the context window, burning through tons of extra tokens.
To understand the crux of any system and avoid getting allured by every shiny new theorem/tool, we can master the fundamentals and weigh an abstraction’s cost against its convenience. Only then can we build and choose better paths – e.g. knowing when to let an agent call an API directly from bash instead of relying on MCP every time etc.
Hope this helps someone.
About me : I am a staff product analyst having interest in Product, Analytics, ML, GenAI, DE, Physics, Art, Literature. You can take mini AI course or read my mini blogs here : https://aisimplified.live
답글 남기기