As a thought experiment, maybe at untools, for just €6/month, they can!
Are they conceptual frameworks, or maybe, meta-conceptual frameworks? My immediate thoughts are that an effective Agent should:
- Select the best tools based on the information available + the goal of the novice user
- Apply the tool, supervised by either structured algorithm (code shell), or reinforcement loop (LLM shell)
- Explain the tool application process and result to the user (requires meta cognition)
Consideration:
- Most tools require visual + spatial mapping. FigJam AI has a major vantage point
- What properties of the tool determine whether it can be implemented as algorithms and reinforcement loop?
- Does the tool specify deterministic transforms, optimization goals, exit criteria?
- What is the implicit/external knowledge that the tool draws on? Can it be implemented by existing text retrieval technology?
- Can the system apply the tools to meta-analyze the application of the tools? This could be the critical ingredient that allows arbitrary reduction and pivoting of complexity via recursion.
- What differentiates Conceptual Framework from Mental models (i.e. Constructs)? Is the former are easier for novice to use because because they involve nouns + verbs that lead to a workflow through which a person gains knowledge, while the latter is only nouns that requires existing knowledge to activate?
- If the application of the tool is the experience through which learning takes place, how can that knowledge be transferred?