Question: What type is the following skill? Is it a topic, note, etc?
So suppose I create a responsibility, that I want to become a better software engineer. I want to record some skills as a part of this responsibility that I ensure I perform when doing software engineering planning.
For instance, I want to record that I need to practice high-level tasks breakdown and to offer trade-offs on the basis of this breakdown. What type are these skills? Are they topics, are they notes? They do belong to a responsibility. And these skills can be a potential operation if needed to fulfill a responsibility.
If this breakdown is recurring work that happens in one sitting, I would model it as an operation.
In the operation note I would write down the mental models, steps, and criteria you use to perform the operation, which may also be useful eventually to turn this into something the AI does on your behalf
The Operations type doing the work as the agent surface is the underrated piece of this spec. If procedures with stable inputs and outputs are where AI agents actually deliver value today, then encoding them as first-class objects in a knowledge base puts the human and the agent in the same schema. Two relationships only, belongs-to and related-to, also lands on the right level of abstraction. Graph-style knowledge bases tend to drown in custom edge types, and constraining to two forces the modeling discipline that makes the system legible to both a person and an LLM.
Awesome! Love that the name (almost) rhymes with import"e"nt.
A suggestion, if I may: have an FAQ or How to section that shows a few common note types and how they would map to (1) the Portent schema and (2) which relationships to use.
For example: "I have a product roadmap document. Which note type do I pick and which relationships do I add".
This could even be turned into an interactive playground where the user could pick from a few basic sample starting points (e.g. "I'm a developer working on X", "I'm a product manager at ACME corp.").
Then the user has to pick a note type & relationships for a number of example scenarios driving to drive home the distinction between the various types and relationships.
This could start with very obvious scenarios ("File a note for project Moon Landing"), getting more and more advanced.
Thank you for the helpful article! I am currently building a knowledge system for a local AI node and this is timely.
Quick question: for AI governance and workflow rules that act as rules for the system - things like system API budget caps, ethical boundaries, code sandboxing and testing hurdles, etc., do you try to keep these in the Portent framework or an entirely different framework / layer?
Question: What type is the following skill? Is it a topic, note, etc?
So suppose I create a responsibility, that I want to become a better software engineer. I want to record some skills as a part of this responsibility that I ensure I perform when doing software engineering planning.
For instance, I want to record that I need to practice high-level tasks breakdown and to offer trade-offs on the basis of this breakdown. What type are these skills? Are they topics, are they notes? They do belong to a responsibility. And these skills can be a potential operation if needed to fulfill a responsibility.
If this breakdown is recurring work that happens in one sitting, I would model it as an operation.
In the operation note I would write down the mental models, steps, and criteria you use to perform the operation, which may also be useful eventually to turn this into something the AI does on your behalf
The Operations type doing the work as the agent surface is the underrated piece of this spec. If procedures with stable inputs and outputs are where AI agents actually deliver value today, then encoding them as first-class objects in a knowledge base puts the human and the agent in the same schema. Two relationships only, belongs-to and related-to, also lands on the right level of abstraction. Graph-style knowledge bases tend to drown in custom edge types, and constraining to two forces the modeling discipline that makes the system legible to both a person and an LLM.
Thank you Alec!
Awesome! Love that the name (almost) rhymes with import"e"nt.
A suggestion, if I may: have an FAQ or How to section that shows a few common note types and how they would map to (1) the Portent schema and (2) which relationships to use.
For example: "I have a product roadmap document. Which note type do I pick and which relationships do I add".
This could even be turned into an interactive playground where the user could pick from a few basic sample starting points (e.g. "I'm a developer working on X", "I'm a product manager at ACME corp.").
Then the user has to pick a note type & relationships for a number of example scenarios driving to drive home the distinction between the various types and relationships.
This could start with very obvious scenarios ("File a note for project Moon Landing"), getting more and more advanced.
Couldn't resist to quickly give https://portent.md/agent-one-pager to ChatGPT and have it quiz me. Pretty neat.
Thank you for the helpful article! I am currently building a knowledge system for a local AI node and this is timely.
Quick question: for AI governance and workflow rules that act as rules for the system - things like system API budget caps, ethical boundaries, code sandboxing and testing hurdles, etc., do you try to keep these in the Portent framework or an entirely different framework / layer?
Appreciate the insight. Thank you!