How is PAI designed to do this? Capture and hierarchise personal archive, ideas, manage todos and schedule prioritisation...

[stephtara]

As I work on trying to get my DA to do what I'm hoping it'll do and workaround some release gaps (amnesia and other broken workflows -- not criticism, I know PAI has holes) I find myself often asking myself (and my DA, to the extent I suggested we build a PAIExpert agent) "are we building a parallel system to PAI here or is there already something in PAI that should be able to deal with this?"

I'm not a developer. I'm more of a thinker and systems-fixer, with some tools and skills to actually help me do that. I also have ADHD and I'm recovering from a brain injury, which means I struggle with a certain number of challenges: a lot of hyperactivity means I have tons of ideas all of the time, difficulty with task-initiation and follow-through, prioritisation, difficulty keeping track of my todos and ongoing projects. After approximately 30 years of rich and intense online life where I have pretty much thrown nothing away, I have a huge personal digital archive containing a lot of raw material I always knew I'd one day have tools to help me manage. It feels like the day has come.

So here are the things I'm hoping my DA will help me with, in addition to adding development skills I'm missing to actually bring into reality some of my digital projects.

• keep my digital archive organised and usable ("I wrote something about xyz at some point, didn't I?" or "I know I heard something about this in a podcast, what was it?" and "ok, now let's pull together all of my creating writing snippets related to grief over three decades, and let's see if there's something to do with all that today")
• make links between what I'm doing now and things from the past that might be relevant
• stay on top of my "open loops" and reduce friction to task initiation or getting back onto something by making it easy for me to recall what I did last and what were the next steps when I left it
• help my prioritise (both micro and macro) by taking into account dependencies between open loops (whether tasks or projects) and deeper purpose behind them ("why is this important and how does it align with my mission")
• help me manage motivation and energy on a day-to-day basis

So, some of the things here are obviously already thought for in PAI, even if they may not work yet at the level I'd need. But for others, I don't know if I'm looking at a proper gap in PAI (and therefore, I just go ahead and design the system I need from scratch) or if there is actually a place for that "ready-made" (or close) in PAI.

Examples:

• all my ideas, ranging from "oh I should do this" (on a sliding scale from "just an idea" to "actually intend to") to "here's how the world works": do they go in KNOWLEDGE, or is that twisting the purpose?
• my personal archive (which is a mess right now): do I "absorb" it into the memory system? do parts live there and others elsewhere? There is unprocessed raw material, e.g. voice recordings, videos, years of geotagged photographs which tell the story of where I was when and what I was doing, IRC chat transcripts, courses I wrote, writing snippets, things I said in conversation or facebook comments that is not just chatting but gives insight into important issues for me, etc), there are 26+ years of blog posts, social media exports, etc -- or do I just organise this stuff in my own way?
• tracking todos (projects are taken care of by the system, that's pretty clear) -- do I build something ad hoc or is there already functionality inside PAI that could do that?
• keeping track of how everything I'm "doing" fits together, linking project A to goal x inside TELOS, understanding that task Y actually unlocks 5 different stalled projects even though it seems to have nothing to do with my wider mission... it feels like PAI "should" tackle this, but it doesn't seem to for me right now – how do I know if it's because I haven't been using it long enough, or if it just doesn't have that in the system, so I had better build it?
• keeping a timeline of my health (symptoms, medications, possible impacting events): is the health stuff inside PAI supposed to do this, or is what I'm thinking of different?
• I could go on... ;-)

I keep stumbling upon these questions because:

• I don't like reinventing the wheel and wasting energy when I already don't have enough for everything that is really important
• I believe that "going with the system" will give more ROI in the long-run than patching on ad hoc workarounds or duplicating functionality which may not be properly integrated with the rest, now or in the future.

Who else has been thinking about this kind of thing?
Have you had similar needs, and how did you choose to deal with them?
Do you have any answers or thoughts about this to help me move forward?

[nick-walt]

A good breakdown of most prolific peoples' lives and what seems Iike an ideal use case for PAI. Have you been using PAI with Claude and have you asked Claude to help create a prompt for each question that it can then use to find the solution in PAI's sc Scaffolding? Basically bootstrapping PAI.

Daniel is working on improving PAI's Scaffolding / Harness to improve Claude's ability to help the user customize / bootstrap PAI.

[stephtara]

Thanks! I'm not sure what you mean by "using PAI with Claude"? As I've understood now, PAI "is" Claude once installed. So yes, I have definitely been "using Claude=PAI" to try and answer all this, but I'm stumbling upon things like broken memory integrity (translation: recurring amnesia), system structure drift, etc, which are making my progress very slow. I'm finding that where I make best progress is when I guide more, so that's why I'm interested in understanding more about how the system is intended to work and how other people have done things.

[Drizzt321]

Briefly scanning your post, I think what I'm hearing (and what I've considered with my own PAI agent) is executive function support. I'm sort sort of neurodivergent/ADHD or the like, and that's what it seems I'm starting to build for myself.

Part of that is having my own way of organizing a knowledge base. This is all where I have project directories each with a doc index, backlog, brief (what is this), journal, references, plans, archive, etc. And then some on my own, but mostly my agent, files get created/added/updated/etc within each project dir within that structure. I then just tell my agent "in this project, XYZ" and as part of my system prompt I have it know about the knowledge base dir and a brief of how it's organized, and then it goes in and looks at the brief & index of that project, and then finds the file(s) it needs to answer your question, or what next thing to do, or comes back and asks me a question to clarify, etc.

I then have a template directory that when I have it create a new project it'll copy over the template, and start from there.

I need to go through and revise/refine the prompt, it's pretty old from when I first started using it.

[stephtara]

Thing is, PAI already seems to have some idea about how to organise certain types of information. I have mine, but I'm trying to see how to work "with the system" as much as possible for the moment. I've got a project directory set up with a readme and that works quite well, but I'm wondering if there is a more "PAI-internal" way of keeping track of the state of moving targets, for example.
Thanks for sharing how you do it!

[jdrolls]

You've got quite a few examples in there. Some of those I use personally and some aren't that important to me. I wouldn't necessarily wait for the features if these are things you would use. I'd add one thing at a time, point your DA at the task, see what it comes up with, and use the feature for a few days / weeks and see if its something to keep or retire.

I've built out some custom things that are "so" custom they aren't worth sharing with the group. Things like health related tracking - synching with various health data points that I have (weight, steps, sleep scores, workouts, food tracking, blood work, any medications / supplements I'm taking etc) and providing that context has been helpful when I have questions before / after a doc visit, but it did take quite a bit of work to compile it all and a small amount of work to keep what isn't automatically synced with an api up to date. My own stuff is so unique though, I don't know that a community built thing would be as good. I plan to do a similar thing with finances once I am in a place to do that.

File organization is something I haven't completely figured out yet. I'll sometimes point my DA at my massive google drive, local fire archive, or obsidian vault files (the three places I store things) and ask it to find stuff but it doesn't have it all archived and organized. Some are organized more than others. The geotagged photos is an interesting idea.

All that said, the concept of "LIfe OS" and a "Scafolding" all points to this being a foundation that you build on, and what it doesn't provide is essentially your own custom software. I try not to reinvent the wheel either, and it's easy to get sucked into spending a ton of time over engineering, but at this point I work it into the DA's prompt and ask if there's a native PAI feature that will do the same thing, or if I should build something adjacent or in addition to PAI. Hope that's helpful.

[stephtara]

An example: health-related tracking. That seems to exist in some form inside PAI now, right? So, build own system from scratch, or try to use (and maybe expand) the internal system? That's always the kind of thing I bump into.
But yeah, what you describe is reasonably close to how I'm trying to approach it. I do think that until recently I may have been having a harder time than necessary because of a certain number of things that were "broken" in 5.0 (skills not loaded, memory feedbacks going down a black hole, etc). Will see how things go now I've got some more of that working properly. But it's tricky when you're new to it and don't know how it's "supposed" to work.

[stephtara]

Follow-up: I'm starting an inventory of the stuff I managed to fix, which is making it easier for me to figure out "how things are supposed to be dealt with with PAI". #1264
As things get clearer, I'll try to summarise my "solutions" in a similar manner. Maybe it will be helpful to somebody!