5 Comments
User's avatar
Fernando Garcia Alvarez's avatar

An interesting part of what you’re saying from a practitioner which has suffered from not enough loud alerts, plumbing errors when models update, and a number of other things is that it’s important to understand which part is linear/predictable (ie code) vs probabilistic (ie LLM inference). So a proper closed loop system needs to be thought through with that vision in mind in order to scale. Just my 2 cts.

Brendan J Short's avatar

absolutely! this is spot on.

how have you solved this to-date?

Fernando Garcia Alvarez's avatar

Not perfectly lol. But single points of writing with guardrails of context. Audit agents. Architect agents. It’s easy for llms to code but you need to keep them honest for intent and architecture. I’ve been following your posts for a time now and I think you have many valid points. Sent a connect request. Happy to chat on what you are seeing,

Fernando Garcia Alvarez's avatar

Not to mention that fast code enables fast technical debt in the sense that you start accumulating layers of code and that without intent it’s easy to make conflicting versions of gtm systems and very hard to know which layer inference comes from

Fernando Garcia Alvarez's avatar

Btw this resonates with thoughts I’ve been having for the past weeks in the work that I’ve done