imagine you are a railroad engineer standing at the helm of a train that’s just reached the end of its tracks. you look out front, staring at an open landscape.

you gather your crew to start placing rails ahead of you so the train can move forward. you decide whether it will be a straight track, a bendy one, or one made out of steel or wood.

though, all of a sudden, you encounter a hill at such a steep angle that your existing tracks simply will not work. you’re stuck! but luckily, you have access to a library of existing track types, some basic building materials, and know-how to try something new.

you work with your team to invent a new category of rail, one that summits any hill with ease. its fabricated and placed ahead.

the train moves forward. did it climb the hill? yes, fantastic! but was the climb optimal, efficient, repeatable? maybe there’s room for improvement. you iterate until the train smoothly navigates the path. move forward, and repeat

eventually, the path from start to destination is established, tested, and refined. future trains can easily follow the track, reliably repeating the journey. though, later trains may have to extend or adapt the route, handling new edge cases or conditions (what happens when its snowing? or when the train is carrying bulkier cargo?)

this is what we’ve been calling an agentic system. leveraging the generative, logical, and exploratory nature of LLMs to systematically enumerate structured workflows that solve previously unknown tasks.

rather than simply “ai agents” (LLMs that call tools), this approach is more deterministic, reusable, symbolic, and explicit. it’s an LLM-powered framework for designing structured workflows - repeatable pipelines from point A to point B - while capturing and evolving the best of previous implementations for other tasks.

thinking abt something similar? we'd love to talk! email me