This is so freaking true Rock! Reading this has been like a big “Duh” with a slap on the forehead. My efforts on the NIST CSF MCP fork I’ve been working on is a case study in what not to do and when not to do it. Costly but valuable in that I’ve been seeing exactly what you describe here and grappling with context bloat, context deviation, and sending way too much information in prompts. Very small scale of impact for me personally as a researcher but expanded to an enterprise level development team and the cost explodes exponentially and doesn’t even factor in user level use of the tools and the prompts they’ll be executing. This is a timely and incredibly relevant article for developers, engineers, and enterprises. As usual, well done sir!
This is so freaking true Rock! Reading this has been like a big “Duh” with a slap on the forehead. My efforts on the NIST CSF MCP fork I’ve been working on is a case study in what not to do and when not to do it. Costly but valuable in that I’ve been seeing exactly what you describe here and grappling with context bloat, context deviation, and sending way too much information in prompts. Very small scale of impact for me personally as a researcher but expanded to an enterprise level development team and the cost explodes exponentially and doesn’t even factor in user level use of the tools and the prompts they’ll be executing. This is a timely and incredibly relevant article for developers, engineers, and enterprises. As usual, well done sir!
Thanks, Trace. I had the EXACT same experience developing it!
Thanks! This is so true. It’s non-intuitive. I’m just now discovering it the hard way. Really appreciate this article!