Deckwright / Walkthroughs / The Bridge

The Bridge walkthrough

Video: the-bridge.mp4 · 9:39 · recorded from the operating product, July 2026. This page is the textual record of that video for readers and agents that do not process video: what it demonstrates, a timestamped transcript, and what it does not establish.

What this demonstrates

This is the operational Deckwright The Bridge, not a concept animation or a design mockup. Polyphonic, multi-officer context development: five named role-scoped officers, individual and all-hands threads, context injection without triggering replies, dispatching research, forwarding one officer's reasoning to others for additions, synthesis across threads, and attaching the result to the Mission Board as inherited downstream context.

Capabilities demonstrated, with timestamps

What this walkthrough does not establish

Timestamped transcript

Machine transcription (Whisper small.en), reviewed; product names normalized (Deckwright, Mapmaker). Otherwise verbatim, including asides.

[00:00] Deckwright is a multi-agent context development harness.

[00:04] Five named agents organized like officers on a ship support the user,

[00:09] positioned as captain in developing software.

[00:12] This isn't mere decoration as the LLMs like to say.

[00:16] We use this naval metaphor strategically

[00:19] throughout the platform as a tool

[00:22] to shape better agent coherence. Deckwright

[00:26] has three surfaces, the bridge for multi-agent chat,

[00:30] the bench for spatial capture,

[00:32] and the mission board which is the production pipeline.

[00:35] Together they all form one deck.

[00:38] You can have multiple decks for multiple projects

[00:41] and you work across these three surfaces to develop raw context.

[00:46] Ideas or comps, outline specs, animations, anything.

[00:51] And all of it's aligned, curated and finally transformed into

[00:55] deployable code.

[00:57] So let's take a quick look at each of the three surfaces individually.

[01:03] So first let's look at the bridge.

[01:05] The bridge is our polyphonic chat interface.

[01:09] This is the ops panel where we do multi-officer or polyphonic chat.

[01:13] It's less like a solo and more like a choir or something.

[01:19] Here's where you can speak to or text to

[01:24] any of the officers individually or you can interact with them

[01:29] all at one time through the all hands thread.

[01:34] Anything that happens on this surface can be forwarded, attached,

[01:37] pruned, curated, added to the pipeline.

[01:41] And here's how it works.

[01:43] Since we want all the officers to be operating

[01:47] from the very beginning with the same basic understanding

[01:51] of the problem space,

[01:54] I'm going to make an all hands announcement to them,

[01:56] describing the project that I want to start making.

[01:59] Although I haven't decided what that's going to be yet.

[02:02] So I can do that here in the all hands thread.

[02:05] And I can do it one of two ways.

[02:07] I can do it either as a message, which will broadcast a post

[02:12] to all five officers, which they will respond to,

[02:17] or I can do it as context, which is a way of simply introducing

[02:22] important context into the chat threads of all of the officers at one time.

[02:32] They don't respond to it, but the next time they start up to respond to something,

[02:37] that's the first thing that they see.

[02:39] You can do that at any time all through the development.

[02:42] We could do an industry standard demo like a Tetris clone,

[02:46] but Deckwright really shines brightest as a development tool

[02:50] to shape a unique vision, to parse out the subtle details of something

[02:57] that a creator specifically wants to make, not recycling someone else's old ideas.

[03:03] Of course, you could do that too.

[03:04] So instead of remaking Tetris, I know, okay.

[03:10] So this is as context.

[03:13] All hands, today we're starting a new project,

[03:16] which is I want to design and build the world's worst game,

[03:22] the worst game that ever existed.

[03:24] It has to be playable.

[03:25] That's all.

[03:25] That's the only design idea I have right now.

[03:28] So together, we'll develop our idea into a truly,

[03:34] like a historically horrible computer game.

[03:41] And then I send it off.

[03:45] So now this is in there as context in each of these separate

[03:50] chief officer, secretary, librarian, scribe, Mapmaker.

[03:53] They all have that, but they haven't responded.

[03:55] It's just there.

[03:58] Okay, now let's start getting individual officers involved.

[04:02] First, let's get the librarian in.

[04:05] If we're trying to make the world's worst game,

[04:09] we should know what our competition is, I guess, for worst.

[04:13] And the librarian's the officer in charge of formal research.

[04:17] So once I send him out on an official research project,

[04:20] he'll start building out the stored corpus of data specific to this one deck.

[04:25] And every deck should have its own little reference library,

[04:29] like a little Wikipedia, but it's a ship, so it's a sea chest.

[04:33] So here we go.

[04:37] I'm gonna choose librarian.

[04:39] And I'm gonna say, librarian,

[04:44] I have an MCP research project for you.

[04:48] Go out and find out what are considered to be the five worst games

[04:52] ever made by humanity, I mean computer games.

[04:56] And figure out what made each of them so bad, and

[04:59] assemble your findings here as your tools direct you.

[05:06] And I send that out.

[05:10] Now we can go to the CO, Chief Officer.

[05:15] And we can say, Chief, I wanna make a manifesto to guide our development of this

[05:25] game project so that all the downstream agents who get our specs and

[05:30] detailed build instructions, they understand the high level scope of

[05:33] what we're building, a legendarily horrific computer game.

[05:36] We're stiffer.

[05:37] So in your opinion, what are the top three things that make a game a great game?

[05:43] Like what are the things that all great games have to have?

[05:50] And we will send them off.

[05:54] So he's working on that, librarians working on this, and in a moment he'll come back.

[06:01] I'm gonna chop this thing together so

[06:03] we don't have to wait around for all the LLMs to do their work.

[06:09] So I could have asked all the officers at once through an all hands post,

[06:15] what do you all think is the top three things that make the game great?

[06:19] But that's a waste, because the top three things are low hanging fruit and

[06:23] all the models are all gonna tell us pretty much the same thing.

[06:25] But now that we have Chief Officer's response, we can force a deeper scope.

[06:32] And this is where we get into the polyphonic flow.

[06:34] Bunch of officers saying things at the same time, and

[06:37] you as a captain are conducting them.

[06:40] Now I'm gonna forward Chief's list of three game features to the other officers.

[06:47] Well, not librarian, he's researching, but

[06:51] I'm gonna say secretary, scribe, and Mapmaker.

[06:55] Each of you add two different new things to this list,

[07:11] and I'm going to forward it to them.

[07:14] So all the officers responded, each in their own thread.

[07:18] And now, I'm gonna tell Chief Officer,

[07:23] who is the officer who sees every post on every thread and can make combinations and

[07:30] synthesize the information across all the surfaces.

[07:35] I'm gonna tell him, make a master list out of all of the editions that all of the other officers

[07:45] have added to your initial three.

[07:47] And post it all back here framed as a manifesto of the most important things that we do not

[07:54] want our game to do or be.

[07:57] This is the counterfactual.

[08:00] If we're careful not to do any of the things that any of the officers have suggested make a great

[08:06] game, our game has a much better chance of being really, really bad.

[08:09] And we'll send that off.

[08:12] Okay, here's Chief Officer's response.

[08:18] Yes, this isn't gonna be funny.

[08:20] Okay, so the last thing that we will look at here on the bridge is really the first essential

[08:30] step towards the development process, which is whatever decisions that we've made here

[08:35] we need to push it into the pipeline and we do that by attaching.

[08:38] So you can attach anything, individual sections of a post, entire posts, ultimately on other

[08:46] surfaces, animations and pictures, etc.

[08:48] It works like this.

[08:50] So because we haven't yet set up the full pipeline, all we have is the top level of the

[08:59] pipeline, which is the mission board.

[09:01] So we're gonna call this the label, we'll call it the manifesto.

[09:07] And we're not gonna put any notes and we're gonna attach it.

[09:11] So now that whole list has become context that every downstream officer building or working with

[09:20] our game ultimately and definitely in the future, we're gonna have to do that.

[09:24] Building or working with our game ultimately and down to the officers ultimately get the spec

[09:31] and create the code understands.

[09:35] Okay, now let's take a look at the bench.

See also: Product evidence ledger · Technical diligence brief · Multi-modal capture · Production pipeline