@GergelyOrosz "loop" means repeat but it differ from the cron jobs
I use an Agent loop to help me fix some minor work items assigned to me every Monday. It would be some dependency packages bumping up. They have similar pattern but it depends case by case
@antirez Agree.
With the new model getting more and more smarter, the behavior of it also getting more and more predictible and accuracy.
We need a higher level abstraction instead of reviewing each line of codr
@creepydotorg WRONG
She dressed that way to prove she had the freedom to dress as she pleased, and that no one had forced her to wear anything.
The male commentators next to her later also achieved the same freedom of dress.
If you don't know how to make your first loop, just copy the link of this article to your agent and ask it to review your session history. You will get the first one!
introducing pi-dynamic-workflows
This is probably going to be a bigger token burner than pi-goal, BUT, dynamic workflows is the first implementation of subagents that i don't hate, mainly because it's "code mode" for subagents. agent writes a js-based workflow DSL into a dedicated tool, engine parses the workflow code and runs it.
the dsl implements some primitives for the agent (agent(), parallel(), pipeline(), phase() and log()) to keep it as simple as possible.
now available in @badlogicgames pi!
pi install npm:pi-dynamic-workflows
Now, every night before bed, I have an extra task: assigning to-do tasks to my Pi agent using `/goal`.
During the day, I often have ideas about my side project. Sometimes minor fixes, sometimes experiments. And I collect them all.
Before going to sleep, I compile them into a file called `goal_tasks.md`.
Most people use `/goal` to execute a single, very long task, but it can also be used to execute a batch of multiple smaller tasks.
The only extra step is that we need some additional prompts to better track, summarize, and review the complete status of each smaller task the next day.
Below is my prompt template; I hope it helps!
Please complete all the non-strikethrough tasks below according to the requirements.
Requirements:
1. For each task below, break it down into multiple commits that are easy for humans to review, as much as possible. Unless a commit's changes are very easy to review, a single commit should not change more than 10 files.
3. After completing the task, strikethrough it in the last commit.
4. Some project-specific requirements xxxxx
5. Before starting the following tasks, please review the logical dependencies between these tasks as a whole to determine if the implementation order needs to be adjusted. Each task should be handled in a separate branch. I will review your work based on the adjusted order, and you need to avoid conflicts between branches as much as possible.
6. While you are working, generate an HTML document to track your implementation progress, providing a brief description of your changes to help the reviewer quickly understand your real-time changes and the summary report after all tasks are completed.
Tasks:
1. Task 1
2. Task 2
3. ...
n. Task n
Suzhou is starting to use robots to help police officers direct traffic.
Not replacing the cops β just giving them a robotic assistant at busy intersections.
Honestly, itβs wild how quickly robots are starting to show up in everyday life. The future is getting real.