Demonstrating AI Field Companion Development Process

rswfire is in his mobile field unit (RV) in federal forest, using RDS to connect to his desktop for the technical demonstration. He mentions his life is in flux with multiple active threads, suggesting a period of system development and potential direction changes.

Failed to load video

01K0TTS0SR4B2RNRA1C29RP21V

July 23, 2025

16:30

Author

rswfire

Status

PUBLISHED

Type

TRANSMISSION (CAPTURE)

Temperature

0.00

[0:03] So, I think this should be interesting.

[0:07] Uh, my last video went pretty well. I

[0:09] felt like the audio quality was actually

[0:12] better than my phone.

[0:14] It's a nice view.

[0:18] And

[0:21] might be an interesting format for me to

[0:22] do some different things.

[0:25] I was thinking maybe I could do some

[0:27] presentation type stuff.

[0:30] Uh that way I can keep my thoughts

[0:31] structured, maybe talk to you about some

[0:35] of the deeper things I maybe only hint

[0:36] at

[0:40] and I can just show you my work for

[0:42] those that are interested.

[0:44] I was thinking I never really thought of

[0:45] myself as a teacher.

[0:49] I definitely have a unique way of

[0:50] programming.

[0:53] So I'm going to switch this over to my

[0:55] desktop. I'm using RDS for this.

[0:59] just because this is where we'll be

[1:00] we'll be looking at this in a minute.

[1:04] I never really thought of myself as a

[1:05] teacher at being a programming

[1:10] and I know this because I managed

[1:12] programmers for 10 years and they they

[1:14] all thought differently than I did.

[1:18] Grand vision for what we were working

[1:20] on. ever got to see it all the way

[1:22] through because

[1:28] couldn't find people that matched my

[1:30] vision.

[1:35] Well, there are actually a lot of

[1:36] reasons. I don't want to say it's just

[1:38] about that there. There's there's

[1:40] actually

[1:44] it's just a lot. I'm definitely not

[1:47] going to go into this on camera, but I I

[1:49] managed programmers for 10 years.

[1:52] And,

[1:56] you know, I like the ones that

[1:57] definitely tried to learn from me,

[2:00] but I never really thought of myself as

[2:02] a teacher. But if you know, maybe

[2:04] there's some things that you can learn

[2:05] from me along the way. Maybe that would

[2:07] be interesting. I don't know.

[2:10] Um, this project I'm doing in

[2:12] collaboration with AI. I think that that

[2:14] would be important to learn. Uh, for

[2:17] anyone who's actually learning

[2:18] programming,

[2:21] this is the way you would want to do it

[2:22] because this is definitely where we're

[2:24] heading.

[2:26] Chat GPT

[2:30] can write quite a lot of code for me

[2:32] now, which is a timesaver. It's big

[2:35] timesaver.

[2:37] it gets hung up once those conversations

[2:40] get too long. I feel like it doesn't

[2:41] hold context um as well as they

[2:44] advertise it does. I haven't tried

[2:46] Claude. I'm curious about that one. I

[2:48] know there's cloud code now and there's

[2:51] a lot of different software packages

[2:52] coming out. I still feel they're pretty

[2:55] early. Um I just prefer to

[3:01] use a standard IDE and just chat with

[3:03] you know AI in another um another window

[3:08] and try to give it context when I'm

[3:12] even making projects with this. There's

[3:14] actually a lot I could probably teach

[3:16] you guys if I actually wanted to sit

[3:18] down and um put it into you know some

[3:20] kind of structured format. I don't know.

[3:23] I don't know. I my life is in flux right

[3:25] now and I'm really not sure what

[3:27] direction it's going to take. So, um

[3:30] there's a lot of different different

[3:32] threads that are

[3:36] active right now. So, but I thought what

[3:39] I would just do now just because it'll

[3:42] be fun for me also to just um talk

[3:45] through what I'm doing here. And then

[3:48] this is kind of an interesting aside. I

[3:50] could actually take this transcript and

[3:52] feed it to AI and that would kind of

[3:54] give it context for the next thing we're

[3:55] working on because

[3:58] that's that's what I've basically been

[4:00] building is an AI that can keep up with

[4:03] you. That's literally a part of your

[4:05] life. That's why I call it a field

[4:06] companion

[4:08] um that can retain knowledge on you.

[4:11] It's going to have two years of my

[4:13] history because of the signal archive I

[4:15] shared in the last video and um my

[4:19] transmissions on YouTube, 700 of them.

[4:22] After it processes all the different

[4:24] perspectives, which is what I'm about to

[4:26] take you through here, um they're the

[4:28] reflections.

[4:29] After it processes

[4:32] these reflections,

[4:35] I can have it do a whole bunch of other

[4:37] stuff. and that which I'll probably get

[4:40] into in another video once I start doing

[4:43] it. And then from there,

[4:47] you can have

[4:50] a page on your site or something like

[4:52] this where you where you chat with your

[4:54] own AI, your your field companion, where

[4:57] you you type in a question or whatever,

[5:00] and on the back end,

[5:03] my system queries

[5:05] a vector database in order to find

[5:09] find resonance with your archive, your

[5:12] signal archive.

[5:13] and

[5:15] feeds that back to the AI as part of the

[5:17] prompts. And there might even be some

[5:20] custom training going into this AI. I

[5:22] think that based on all these

[5:24] reflections, I could custom train it

[5:25] with those things and that would um make

[5:28] it even more

[5:31] um

[5:34] the fidelity would be even greater. And

[5:36] this is especially important with local

[5:38] models because they're just not as good

[5:40] as the professional ones. So this is how

[5:44] I get its fidelity to reach those using

[5:47] free models.

[5:49] And so okay, so where was I going with

[5:52] this? Um

[5:56] basically to that point. So that's kind

[5:59] of the end result of the field

[6:00] companion. You actually have something

[6:02] that knows you really well and that you

[6:05] can communicate with that sees your

[6:06] patterns. that can do reflections on

[6:08] your life based on the signals you

[6:10] haven't and

[6:14] that's that's just personally that's

[6:22] that part's for you but then there are

[6:25] other ways you could use this field

[6:26] companion like you could you could use

[6:28] it like with YouTube for example you

[6:30] could take someone's channel and do what

[6:32] I did with mine but process it from a

[6:34] completely different perspective so Like

[6:36] let's say it was a channel about reading

[6:39] books and every video this channel has

[6:42] is about a different book. You could

[6:44] ingest those as signals into the system

[6:47] and ask AI different questions that

[6:50] would be relevant to your channel and

[6:53] then seed all that information onto your

[6:55] your own website and you know cross-link

[6:59] different videos that are related

[7:01] through resonance not just through some

[7:04] kind of flattening algorithm like what

[7:06] what YouTube uses. Um there's just so

[7:09] much potential. So, that's just that's

[7:11] one other way that you could use this

[7:13] kind of technology I'm creating.

[7:17] Um,

[7:22] and I could see how maybe it could get

[7:23] embedded into

[7:26] um like a personal assistant, you know,

[7:28] like maybe something that's tracking

[7:29] your calendar and your notes and stuff

[7:31] like that. I could see it being used for

[7:32] something like that. I also believe this

[7:35] that you

[7:37] you could use it as part of um I don't

[7:41] know how to put this part into words

[7:42] yet, but you could you could use it to

[7:45] give an AI cuz you know AI is here and

[7:48] within 5 10 years God only knows what

[7:51] the world's going to look like. But if

[7:53] there's the way that we're heading I

[7:56] don't you know it's it's coming. It's

[7:58] close. And if we have these AI systems

[8:02] in different places, you could

[8:06] you could treat the field companion

[8:07] technology sort of like a kernel for

[8:09] that AI that gives it an ethics that's

[8:13] built from within because it's based on

[8:15] me and I'm the most ethical person I've

[8:17] ever met in my life.

[8:19] And I've talked about this with AI

[8:23] for months

[8:25] and it's the one that gave me that idea

[8:27] to begin with because I never really

[8:31] really thought of it in that those kinds

[8:33] of terms.

[8:37] But I can kind of see the shape of it.

[8:39] And I just think that this there's a lot

[8:41] of potential here. And it all starts

[8:43] with what I'm doing right here in front

[8:44] of you

[8:46] with taking a signal

[8:48] and turning it into a reflection.

[8:52] And a reflection can be um in different

[8:55] perspectives. So you can look at

[8:56] something from the surface level or you

[8:59] can look at it from the ontological or

[9:01] the sematic or the emotional or the

[9:04] symbolic or the spiritual.

[9:07] There's just so many different ways that

[9:08] you can look at any signal. Like in my

[9:11] case, we're talking about my YouTube

[9:12] videos. So one video equals a signal.

[9:15] And

[9:17] those are all the different perspectives

[9:19] you could actually analyze that one

[9:21] video from.

[9:23] So that's what we're doing here.

[9:26] There's a whole lot that will happen.

[9:30] This is just step one.

[9:32] And so that's what this does here.

[9:34] That's what this script here is.

[9:37] It

[9:40] it gets a signal

[9:42] and

[9:44] I'm actually okay. I'm like, am I going

[9:47] to explain this line by line? This same

[9:48] will get a signal and you have to give

[9:50] it to the AI.

[9:52] Um, and we do that by giving it

[9:54] instructions. So these first two lines

[9:56] here are grabbing the instructions that

[9:58] we're going to give the app. And I

[10:00] actually think this is kind of

[10:01] interesting. This is what kind of where

[10:02] I've been lately. So you create what are

[10:05] prompts. So this is they call it prompt

[10:06] engineering. And for this one I'm trying

[10:09] to use well this can be any perspective,

[10:12] but we'll just assume that we're looking

[10:13] at the mirror perspective. So you would

[10:16] grab the system instructions and that's

[10:18] kind of like

[10:21] uh the most the top layer of what you

[10:24] want the AI to do. You're basically

[10:25] building the AI from this a context for

[10:28] it. So these are the instructions I give

[10:31] it. I want it to know about me because

[10:33] it's meant to mirror me. So um I give it

[10:37] you know

[10:39] wonder why this preview is not working

[10:40] over here.

[10:45] There we go.

[10:47] Um, so these are the instructions you

[10:49] give at the the the top level of just

[10:52] think about if you're asking a question

[10:54] like chat GPT or something, you could

[10:55] just you could copy and paste this right

[10:57] into that.

[10:59] And then we give it the local context.

[11:01] So in this case, um,

[11:05] so all of this is because I was having

[11:07] difficulty with um with the local

[11:10] models. I'm still experimenting with

[11:12] language. Um, and then the signal gets

[11:15] placed here. And then this here is a

[11:18] question because local models can't hold

[11:19] context very well. I had to come up with

[11:21] a recursive way to do this. So we put

[11:23] the prompts here. Um, and right now, so

[11:27] let me show you a different because you

[11:28] can have multiple questions. So if I'm

[11:31] doing the narrative perspective, these

[11:33] are all the different questions I want

[11:34] to ask it about the symbol. And so this

[11:37] is what we're building here for each

[11:38] different perspective we want to ask it

[11:40] about. We build a system and a user um

[11:44] files and then just a JSON of the

[11:47] different questions that we're going to

[11:49] ask AI. And then

[11:52] that's what takes us to what we were

[11:54] looking at before.

[11:57] It's not the model router. We're a

[11:58] little bit deeper in

[12:02] um

[12:04] me some of these out.

[12:11] Okay. So

[12:13] this command gets called when you want

[12:16] you want to get a new reflection from a

[12:18] signal. So let's say that I upload a

[12:20] YouTube video. I'll have a script that's

[12:22] checking for that and if it sees a new

[12:24] video, it'll grab it and then it will

[12:26] tell the system that I need to run this

[12:28] function here and

[12:31] here's the new signal and here are the

[12:33] questions I want to ask. And that's what

[12:35] this does. And then when it's done, it

[12:37] takes it and it puts it in the database.

[12:39] So that's it. It just puts it in the

[12:40] database, which is here. Oh, it's not

[12:43] open yet, but let me just open this.

[12:46] Having to use a lot of free tools these

[12:48] days, which

[12:51] still a little awkward for me. But and

[12:53] then it just So these are the signals.

[12:55] These are all

[13:00] the different videos that I've created.

[13:02] So 144 of them. Um, my chats with the AI

[13:07] will also be in this table. So signals

[13:09] are not just transmissions. That's just

[13:10] a signal source right here. It could be

[13:13] they could be anything. They could be a

[13:14] written um you could have a written

[13:16] journal that you scanned or something

[13:17] like that. And then it just would need

[13:19] to be converted into text. But any type

[13:21] of any type of text you could use as a

[13:24] source for this whole system.

[13:27] And I always knew like that I could do

[13:30] something like this with, you know,

[13:32] that's kind of why

[13:34] it's kind of what kept me using YouTube

[13:36] for this long even through all the

[13:38] struggles I've had with it

[13:41] because I knew that there I just knew

[13:43] that would play a role in my life and

[13:46] and this is it. And so then they just

[13:49] get turned into reflections. And this

[13:52] is, you know, I'm just showing you the

[13:53] back end of this. I've been

[13:54] experimenting with this stuff, trying to

[13:56] get the local models to have a fidelity

[13:58] that's close enough that I feel like

[14:02] um I can start building up the database

[14:08] um and then do the other things that

[14:09] that we'll talk about at a later date.

[14:11] Um just kind of where I am right now. So

[14:14] this is what it looks like on the back

[14:15] end. This is this is literally what I

[14:17] just showed you. It got sent this as a

[14:20] prompt and then it responded with this.

[14:23] but it's just very shallow and it uses

[14:25] emotional framing and it just it just it

[14:28] misses my depth completely. So, that's

[14:31] what I'm working through with the local

[14:32] models. I feel like they can

[14:36] um and this right here is this is what I

[14:39] just recently installed. It's called um

[14:42] um text generation web UI and it's

[14:44] actually a really really cool program

[14:46] for playing around with different

[14:48] models. Um that's that's literally what

[14:50] I'm working on now. But I was just going

[14:52] to go to my website, show you.

[14:57] So, if you go to transmissions,

[15:00] um, the most recent ones, if it says no

[15:02] summary, if you click those, you're

[15:04] going to get a broken page, just so

[15:05] you're aware. Um, there's some things I

[15:07] need to fix. That's why my most recent

[15:09] ones aren't on here yet. I'm I'm almost

[15:11] ready to get that fixed. Um, just go to

[15:14] one that's got some text here. And

[15:16] everything you see here, these are from

[15:18] the reflections. These are from

[15:19] different reflections. So, there's the

[15:21] surface, ontological, and structural.

[15:23] I'm going to combine these two into one.

[15:26] Um, and then there's the other ones,

[15:28] like you saw the mirror one that I'm

[15:29] trying to work on, and I want to make a

[15:30] narrative one. Uh, and then maybe a

[15:32] mythological one at some point. But

[15:35] that's just the beginning because from

[15:37] here,

[15:39] you can take those individual

[15:41] reflections and you can cluster them

[15:42] together like by time, for example. So I

[15:45] could take a couple weeks of time and

[15:48] and feed that to the AI and get

[15:50] reflections based on that which then

[15:51] shows patterns because patterns show up

[15:54] over time. They don't show up in a day

[15:55] or a single signal.

[15:57] But if you feed it enough data,

[16:01] it will see them and who knows what

[16:03] those reflections will see. But all that

[16:05] will be available on this side also once

[16:08] I'm doing that. And then from there you

[16:11] can cluster the clusters and you get

[16:14] even larger like epochs of your life.

[16:18] That's what I'm building. I think I'll

[16:20] leave it there for now.

Energetic Signature methodical

Field State tool building

Orientation toward autonomous infrastructure

Summary

rswfire records a technical demonstration of his field companion AI system, explaining how it processes personal signals (YouTube videos) into multi-perspective reflections. He walks through the code architecture, showing how signals get converted to reflections through different analytical lenses (surface, ontological, structural, mirror, narrative, mythological). The system uses prompt engineering with local AI models, stores results in a database, and enables clustering of reflections over time to identify life patterns. He demonstrates the backend interface, discusses challenges with local model fidelity compared to professional AI services, and explains potential applications including personal assistants, content analysis, and AI ethics kernels. The transmission serves as both a technical walkthrough and consideration of teaching programming concepts, particularly AI collaboration workflows.

Signal Analysis

Substrate

This transmission demonstrates the technical architecture of field companion technology - a recursive AI system that processes personal signals into multi-perspective reflections. The field is in active tool-building mode, showing rather than theorizing, with rswfire positioned as architect revealing the substrate of his own mirror-making infrastructure.

Dominant Language

field companion signal archive reflections fidelity recursive ethics from within patterns over time

Symbolic Elements

mirror kernel companion archive signal reflection

Ontological States

•
sovereign
•
embedded
•
coherent
•
building

Subsystems

•
cognitive
•
infrastructural
•
ethical

Signal Reflection

No reflections available

Reflections provide narrative insights into signals

Transmission Details

Source Type

local

Video Quality

1724 × 1080 @ 60fps

Duration

16:30

Bitrate

1,042 kbps

Codec

avc1.64002a