I love to use a paper notebook for fast notes, random thoughts, and to capture tasks (something similar to a GTD inbox). With it, I use a small system of symbols (some degree of similarity with the bullet journal method).
The only issue with paper is the edition. My solution has been to use a compact, slim ring binder with a notebook-like feel. If required, you can move sheets to reorganize content, move the most important ones to another place, etc. Until now, the best alternative that I have found is the Kokuyo Campus Smart Ring Binder.
No practical code example, sorry. The post is based on my own experience using agents, and I haven't reached a reusable generalization yet.
That said, two cases where I noticed the pattern:
Meal planning: I had a weekly ChatGPT task that suggested dinner options based on nutritional constraints and generated a shopping list (e.g. two dinners with 100g of chicken -> buy 200g). After a few iterations, it became clear that with a fixed set of recipes and their ingredients, a simple script generating combinations was enough. The agent's reasoning had already done its job — it helped me understand the problem well enough to replace itself.
QA exploration: I was using an agent to explore a web app as a QA tester. It took several minutes per run. After some iterations, the more practical path was having it log its explorations to a file, then derive automated tests from that log. The agent still runs occasionally, but the tests run frequently and cheaply.
Regarding your point about tasks that need individual reasoning every time — I think you're right, and that's actually the core of the idea. Not every task matures into a script. Extracting structured data from images probably stays deliberative if the images vary significantly. The cycle only applies to tasks that, after enough repetitions, reveal a stable pattern. The agent itself is what helps you discover whether that pattern exists.
These past weeks I finally organized some ideas I'd been sitting on and wrote two posts:
From Agentic Reasoning to Deterministic Scripts: on why AI agents shouldn't reason from scratch on every repeated task, and how execution history could compile into deterministic automations
> The second is absorption: mental models form, edge cases become intuitive, architectural relationships solidify into understanding. ... . The friction of implementation creates space for reasoning.
> This gap between output velocity and comprehension velocity is cognitive debt.
I have felt that lack of absorption during the last months, adding doomscroolling to the equation, I have felt how my thinking is disappearing.
I tried to speculatively expand that idea in this post
I agree, the LLM support we lose some level of manual work, maybe that is useful to be more familiar with the information, although I am not sure if that is a critical step. The LLM gives us management support. We can use a command to collect incomplete tasks, the LLM is doing that with extra steps, and it shows us the collected information, but in the end, we decide which tasks to prioritize.
Inbox Processing Workflow - GTD-style system to keep your inbox at zero, with AI classification of notes to projects, resources, or someday/maybe lists.
Incomplete Tasks Review Workflow - weekly review that analyzes your tasks, identifies gaps, detects duplicates, and generates a focused "plan of attack" with your highest-priority actions.
Both workflows combine classic productivity principles (GTD, weekly reviews) with AI assistance to reduce cognitive load and maintain system trust. They use command-line tools (find, ripgrep) to analyze your vault and generate processing recommendations in an easy-to-review card format.
Use at your own risk. Validate each command before executing it. It is strongly recommended to use a version control system (Git) in your Obsidian vault to visualize and revert changes made by the agent.
Any suggestions to improve these workflows are appreciated.
Remote: Yes
Willing to relocate: Yes
Technologies: Python, Go, Elixir, Docker.
Résumé/CV: https://www.linkedin.com/in/juanpabloabarzua/
Email: jpabloaj(at)gmail.com