WEBVTT

NOTE Sentence-level transcript of https://www.youtube.com/watch?v=TRfzFJCJ7ZE

NOTE One cue per sentence. Cue ids are the line anchors on /transcripts/TRfzFJCJ7ZE.html. A cue ends where the next begins, or 2 s after its last word.

s1
00:00:01.309 --> 00:00:03.309
[music]

s2
00:00:25.080 --> 00:00:26.600
Hello.

s3
00:00:26.600 --> 00:00:38.240
I was at a wedding this weekend and it was in New York and there were a bunch of trendy people there and uh I I told I was talking to someone who I'd never met before and

s4
00:00:38.240 --> 00:00:41.600
uh I told him I was prepping for this presentation and he was like, "Oh, that's cool.

s5
00:00:41.600 --> 00:00:43.640
Like, what what's the conference about?"

s6
00:00:43.640 --> 00:00:46.320
And I said it was about AI engineering.

s7
00:00:46.320 --> 00:00:52.880
And I could just see his eyes glaze over and he started like looking behind me to find the next person to talk to.

s8
00:00:52.880 --> 00:00:57.800
And I'm it's like so cool to actually be in a room full of people who actually want to hear about this stuff.

s9
00:00:57.800 --> 00:01:00.160
So, I I I I'm very excited to be here.

s10
00:01:00.160 --> 00:01:02.720
Um I am the co-founder of Conductor.

s11
00:01:02.720 --> 00:01:05.440
Um has anyone here used Conductor?

s12
00:01:05.440 --> 00:01:06.200
Okay, nice.

s13
00:01:06.200 --> 00:01:07.120
Nice.

s14
00:01:07.120 --> 00:01:07.920
Okay, cool.

s15
00:01:07.920 --> 00:01:15.560
So, uh uh for those who don't know, uh Conductor is a desktop app for managing a team of coding agents all at the same time.

s16
00:01:15.560 --> 00:01:24.480
So, instead of having a bunch of terminal windows for your Cloud Codes, your Code Accesses, or your uh whatever uh coding agent, you have one interface to manage them all.

s17
00:01:24.480 --> 00:01:31.480
And one of the really cool things about building Conductor has been that I've seen a lot of the best builders up close.

s18
00:01:31.480 --> 00:01:37.360
I've like watched their workflow, I've seen how they work, I've seen the things they do do and the things that they avoid doing.

s19
00:01:37.360 --> 00:01:44.840
And so, I thought I would compile a bunch of the principles um that I've seen the best engineers use um and give them to you all.

s20
00:01:44.840 --> 00:01:48.440
Um so, here's what Conductor looks like.

s21
00:01:48.440 --> 00:01:50.560
Uh can you can you guys see this?

s22
00:01:50.560 --> 00:01:51.320
Okay, nice.

s23
00:01:51.320 --> 00:01:51.680
Okay.

s24
00:01:51.680 --> 00:01:53.440
So, here's what Conductor looks like.

s25
00:01:53.440 --> 00:02:00.240
Um and here are my principles for being the fastest builder in your organization.

s26
00:02:00.240 --> 00:02:05.440
So, let's start with number one, stay near the frontier.

s27
00:02:06.080 --> 00:02:11.520
Staying near the frontier means you are always trying the latest things, basically the day they come out.

s28
00:02:11.520 --> 00:02:16.280
Uh it means that when uh uh ultra code comes out, you're trying it.

s29
00:02:16.280 --> 00:02:21.160
It means that when uh uh slash go comes out, you're giving it a go.

s30
00:02:21.160 --> 00:02:25.480
Um it's really important to stay near the frontier um for a few reasons.

s31
00:02:25.480 --> 00:02:33.440
Um if you are doing your own startup, then staying near the frontier means that you will um come up with lots of new ideas for what you should actually be building.

s32
00:02:33.440 --> 00:02:34.959
Um and this literally happened to us.

s33
00:02:34.959 --> 00:02:37.920
We were building a totally different app called Chorus.

s34
00:02:37.920 --> 00:02:54.360
Um but we started using We were such power users of cloud code back in February of last year that we uh started building our whole workflow around cloud code, and we started cloning our repo five times, and then we discovered work trees, and then bit by bit we had built conductor as an internal tool.

s35
00:02:54.360 --> 00:02:58.320
And we couldn't have figured that out if we uh weren't staying near the frontier.

s36
00:02:58.320 --> 00:03:06.000
And if you're not doing a startup, you should be the person at your company who always knows what the latest uh the latest workflows are.

s37
00:03:06.000 --> 00:03:15.239
Um it used to be that you could just kind of like use your social graph and the information that was important about the best workflows to use would trickle down to you.

s38
00:03:15.239 --> 00:03:17.239
But things just like move way too fast now.

s39
00:03:17.239 --> 00:03:20.840
You're You're always going to be three to six months behind if you do that.

s40
00:03:20.840 --> 00:03:23.120
Um so, it's a very important to stay near the frontier.

s41
00:03:23.120 --> 00:03:26.000
And there's an important word here, near.

s42
00:03:26.000 --> 00:03:38.000
Um there is a danger if you are at the frontier, um you can you can do what uh I call mid-whip miming, where you're spending all of your time working on your workflow and not doing actual work.

s43
00:03:38.000 --> 00:03:41.320
And so, internally we've come up uh with a heuristic for this.

s44
00:03:41.320 --> 00:03:43.760
Um we call it don't beat the market.

s45
00:03:43.760 --> 00:03:45.239
Don't try and beat the market.

s46
00:03:45.239 --> 00:03:56.560
Um uh the concept here, like the heuristic you should use when you're you're to decide if you're near the frontier or at the frontier and uh too deep in into

s47
00:03:56.560 --> 00:04:02.840
the latest trends is you should ask yourself why isn't this workflow the default?

s48
00:04:02.840 --> 00:04:16.359
So for example when Ralph loops were becoming a big thing you should ask yourself like should I spend a ton of time optimizing my workflow to work with Ralph loops

s49
00:04:16.359 --> 00:04:29.960
because if Ralph loops work for everyone like if they are the default then you probably should just wait for Anthropic or Open AI or ever to build the workflow into the into the default harness.

s50
00:04:29.960 --> 00:04:39.720
You can think of this is sort of like an efficient market hypothesis where unless you have like real alpha you shouldn't be optimizing your workflow too much.

s51
00:04:39.720 --> 00:04:48.200
And what I mean by real alpha is like some kind of information about either your users or your code base that the models might not know about.

s52
00:04:48.200 --> 00:04:51.960
So an example for us is we it's we're we're a chat app.

s53
00:04:51.960 --> 00:05:00.840
We have to render really long chats really quickly and performance is is important to us and so we need to spend a lot of time optimizing our react queries

s54
00:05:00.840 --> 00:05:07.440
to render the chats quickly and we're willing to make sacrifices in other parts of our code base to make that happen.

s55
00:05:07.440 --> 00:05:16.200
Um So if you have some kind of alpha like some kind of information about the app you're building that the models might not know about then you should put time into the workflow.

s56
00:05:16.200 --> 00:05:18.400
Otherwise don't don't midwit meme.

s57
00:05:18.400 --> 00:05:24.360
Don't don't be the person who has an amazing Emacs setup but like doesn't actually get stuff done.

s58
00:05:24.360 --> 00:05:26.040
Okay three.

s59
00:05:26.040 --> 00:05:27.600
Create slop free zones.

s60
00:05:27.600 --> 00:05:32.120
So at Conductor we have this term we call a slop free zone.

s61
00:05:32.120 --> 00:05:40.560
And a slop free zone is a part of the code base or a part of the app that requires really strict human review.

s62
00:05:40.560 --> 00:05:43.280
And we're actually I think a little bit unusual in this way.

s63
00:05:43.280 --> 00:05:51.120
I think a lot of people assume that we are uh pure token maxers and we are like ripping through like 30,000 line PRs, but we're actually not.

s64
00:05:51.120 --> 00:05:57.080
Um we're actually quite careful with certain parts of our codebase, and then very loose with other parts of our codebase.

s65
00:05:57.080 --> 00:06:06.640
Um the reason this is important is because if you are not careful about your slot free zones, your your code your codebase can get in a really tricky spot.

s66
00:06:06.640 --> 00:06:07.880
Um and this actually happened to us.

s67
00:06:07.880 --> 00:06:13.800
We have We've had to rewrite our whole app like a couple of times because we weren't careful about slot free zones.

s68
00:06:13.800 --> 00:06:18.440
Um specifically, we have a uh migrations file.

s69
00:06:18.440 --> 00:06:26.680
And uh in our CI uh any change to migrations file requires the AI uh a human to review it.

s70
00:06:26.680 --> 00:06:29.800
Um we also assume that anything written in Slack is slot free.

s71
00:06:29.800 --> 00:06:30.919
It's It's not written by the AI.

s72
00:06:30.919 --> 00:06:32.280
It's written by a human.

s73
00:06:32.280 --> 00:06:39.480
All of our docs, um all our cloud MDs, like all our skills, we put a ton of time into making them good.

s74
00:06:39.480 --> 00:06:43.880
Um and this is also something I've seen with all the the the best builders up close.

s75
00:06:43.880 --> 00:06:48.400
Like they put an unusual amount of time into the cloud MD or their skill files.

s76
00:06:48.400 --> 00:06:58.680
Um and I I think like another way that I've thought about this is like if you had a new intern that was joining your company, and you had the opportunity

s77
00:06:58.680 --> 00:07:03.200
to like whisper something in their ear every time they started working.

s78
00:07:03.200 --> 00:07:06.600
Like every day, anytime they sat down, you could like whisper something in their ear.

s79
00:07:06.600 --> 00:07:10.160
You would probably put a lot of thought into like what it is that you're whispering in their ear.

s80
00:07:10.160 --> 00:07:13.480
And this is what the cloud MD or agent's MD is.

s81
00:07:13.480 --> 00:07:17.800
It's like information that gets loaded into the agent's context every time they start working.

s82
00:07:17.800 --> 00:07:22.760
Um and so you probably want to put a lot of thought into into those.

s83
00:07:22.760 --> 00:07:25.080
Um okay, four.

s84
00:07:25.080 --> 00:07:26.760
Feed the beast.

s85
00:07:26.760 --> 00:07:34.040
Uh at Conductor, we have a uh internal tool we call the Conductor Internal Agent.

s86
00:07:34.040 --> 00:07:38.720
Um and it is uh we also also known as the CIA.

s87
00:07:38.720 --> 00:07:46.280
And the CIA is basically like the the uh centralized database of everything that's happening in the organization.

s88
00:07:46.280 --> 00:07:50.280
So, anytime a new Slack message gets sent, the CIA picks it up.

s89
00:07:50.280 --> 00:07:56.880
The CIA agent will see that a new message sent Slack, it will pick it up, and it will save it to a Postgres uh uh table.

s90
00:07:56.880 --> 00:08:00.440
Anytime a user has a bug request in Discord, the same thing happens.

s91
00:08:00.440 --> 00:08:05.440
Anytime we have a meeting, we are recording it, uh and it goes into the CIA.

s92
00:08:05.440 --> 00:08:18.880
Um we call this feed the beast because you you want to for if your agents to be effective uh in your company, you want them to have as much information and as much context as they can have about the way you guys specifically work.

s93
00:08:18.880 --> 00:08:25.040
And the best way to do that is by having a centralized place for all of the information to go.

s94
00:08:25.040 --> 00:08:28.280
Uh I think this this tweet sums it up pretty well.

s95
00:08:28.280 --> 00:08:40.080
Uh uh It's really effective to just put everything in a database and then give your agent a SQL tool and let it handle uh uh handle the rest.

s96
00:08:40.800 --> 00:08:41.840
Okay.

s97
00:08:41.840 --> 00:08:45.440
Next is free-range agents.

s98
00:08:46.000 --> 00:08:49.520
Give your agents a lot of space to play.

s99
00:08:49.520 --> 00:09:01.920
Give them Give them a sandbox where they that won't get killed, where they can explore your code base, where they can work on really hard tasks, where they can they know that they're not going to get shut down when you close your laptop lid.

s100
00:09:01.920 --> 00:09:06.520
Uh Give them Give them opportunities to create more of themselves.

s101
00:09:06.520 --> 00:09:11.520
Um Give them ways uh to collaborate with other agents and other humans.

s102
00:09:11.520 --> 00:09:23.720
Um I think what's really in- interesting about free-range agents and like this this concept and why this is important is the models are getting better, and they are able to run for much longer,

s103
00:09:23.720 --> 00:09:25.520
and there're going to be many more of them.

s104
00:09:25.520 --> 00:09:34.520
And so, if they are confined to your laptop, then the they're not going to be nearly as effective as uh as if they are free roaming.

s105
00:09:34.520 --> 00:09:44.960
Um The other thing that's important here is that once you give them a a sandbox to play in that isn't confined to your laptop, there's a bunch of really cool stuff that you can build on top.

s106
00:09:44.960 --> 00:09:52.680
Um and we have built some of those things into Conductor and I I'll give you a quick glimpse into into some of those cool things.

s107
00:09:52.680 --> 00:09:58.320
So, this is Conductor.

s108
00:09:58.320 --> 00:10:00.200
I'm going to make it a bit bigger.

s109
00:10:00.200 --> 00:10:10.000
Um and this is actually a new version of Conductor that is uh coming out this week and it is centered around collaboration in the cloud.

s110
00:10:10.000 --> 00:10:14.200
And so, the the the thing I said was you need a a sandbox for agents to play.

s111
00:10:14.200 --> 00:10:15.760
You need a free-range agent.

s112
00:10:15.760 --> 00:10:25.600
And so, you'll notice that each workspace has a little cloud icon at the top and I can click it and get information about the sandbox that the agent is running in.

s113
00:10:25.600 --> 00:10:30.480
Um and what's what's awesome about this is I can close my laptop and the agents are going to keep running.

s114
00:10:30.480 --> 00:10:37.280
Up up until basically this week, every uh every task in Conductor was built on a get work tree, but now they're in a cloud sandbox.

s115
00:10:37.280 --> 00:10:38.960
They are free-range agents.

s116
00:10:38.960 --> 00:10:46.280
But what's also really cool um that you can that we can build on top of uh cloud is collaboration.

s117
00:10:46.280 --> 00:10:49.720
So, you'll see here, I'll make this even bigger.

s118
00:10:49.720 --> 00:10:51.000
That's me.

s119
00:10:51.000 --> 00:10:53.680
And here are a list of the things that I am working on.

s120
00:10:53.680 --> 00:10:56.400
And you can see that I am in the Conductor org.

s121
00:10:56.400 --> 00:11:01.480
But if I scroll down, I can see what Cadence is working on.

s122
00:11:01.480 --> 00:11:04.160
I can see what Lewis is working on.

s123
00:11:04.160 --> 00:11:09.200
I can see what Tywan is working on and I can I can I can see that I can see Jackson's face pop up there.

s124
00:11:09.200 --> 00:11:12.560
I can click in and see what he's working on in real time.

s125
00:11:12.560 --> 00:11:21.960
Um and I think collaboration is the the one one of the most important new concepts uh in these tools that no one is really talking about right now.

s126
00:11:21.960 --> 00:11:30.040
Um collaboration is important because uh not only as not only it is it true that all great things are built with teams of people.

s127
00:11:30.040 --> 00:11:31.280
Like they're not built by individuals.

s128
00:11:31.280 --> 00:11:32.960
They're they're built by teams.

s129
00:11:32.960 --> 00:11:40.960
Um but also as the models get better, um as we've seen this with the two days of Fable, you can get a lot more ambitious with the kinds of things you're building.

s130
00:11:40.960 --> 00:11:45.160
And if you're getting more ambitious, you're going to need more people and more agents to work on those things.

s131
00:11:45.160 --> 00:11:51.160
So, I'm going to go into an uh a workspace that Caden is working on.

s132
00:11:51.160 --> 00:11:57.360
Uh this one, and I'm going to say um I can review the changes that he's made.

s133
00:11:57.360 --> 00:12:09.560
Um I'll make this a little smaller, and this looks fine, but I'm just going to say uh can we actually use tabs, not spaces?

s134
00:12:10.760 --> 00:12:19.240
And Caden should be able to see that message happen in real time, and he can actually chat in the workspace as well.

s135
00:12:19.240 --> 00:12:22.280
So, we can see here that he's typing.

s136
00:12:22.280 --> 00:12:24.960
So, I see Caden is typing.

s137
00:12:24.960 --> 00:12:27.480
Let's see what he says.

s138
00:12:31.720 --> 00:12:34.560
Seems like he's typing a lot.

s139
00:12:37.200 --> 00:12:40.000
Okay, maybe he stopped typing.

s140
00:12:45.600 --> 00:12:47.120
I'll give him a second to take a look at it.

s141
00:12:47.120 --> 00:12:54.160
But, the point is we can now have collaborative workspaces that are shared in real time with people on our team.

s142
00:12:54.160 --> 00:12:58.040
Um Okay, he's typing again.

s143
00:13:01.839 --> 00:13:02.680
Okay, come back.

s144
00:13:02.680 --> 00:13:04.760
The agents have escaped.

s145
00:13:04.760 --> 00:13:05.880
All right.

s146
00:13:05.880 --> 00:13:11.200
So, I'm really excited about this, and we're rolling this out to all conductor users this week.

s147
00:13:11.200 --> 00:13:18.560
Um I think I think collaboration is going to be one of the most important one of the most important new interface changes this year.

s148
00:13:18.560 --> 00:13:27.840
Um the other really cool thing about cloud is that um and giving it the agents the free-range sandbox is that we can give the agents APIs to spawn themselves.

s149
00:13:27.840 --> 00:13:32.400
So, I have here a uh I'll bring up my open claw.

s150
00:13:32.400 --> 00:13:36.280
Uh this is my open claw called Lord Crandon.

s151
00:13:36.280 --> 00:13:44.120
And you can see that it has I don't know how how well you can see this text, but it has access to a conductor API.

s152
00:13:44.120 --> 00:13:56.440
And so, from my phone or from my Telegram or from really or from Slack or wherever I am, I can say, "Hi, can you create a new workspace for me that makes

s153
00:13:56.440 --> 00:14:00.400
Yeah, makes all the buttons blue."

s154
00:14:00.400 --> 00:14:07.880
And so, I'll I'll text that to Lord Crandon, and Lord Crandon has access to the conductor API and so can kick off work itself.

s155
00:14:07.880 --> 00:14:12.240
Um So, let's see what it does here.

s156
00:14:15.920 --> 00:14:17.760
Okay, so just created the workspace.

s157
00:14:17.760 --> 00:14:19.280
The agent is on it.

s158
00:14:19.280 --> 00:14:29.200
And then I can go into my conductor while I'm out and about, but it's still setting up the workspace, and it will do work for me while I am gone.

s159
00:14:29.200 --> 00:14:36.600
So, I'm pretty excited about all the all the cool things you can build on top of of free range agents.

s160
00:14:38.760 --> 00:14:40.360
Okay.

s161
00:14:40.360 --> 00:14:48.000
The final principle that I want to talk about today and the title of this talk is orchestras, not factories.

s162
00:14:48.000 --> 00:14:52.120
The whole like talk track today is about software factories.

s163
00:14:52.120 --> 00:14:55.920
And I honestly kind of hate the term.

s164
00:14:55.920 --> 00:15:00.720
I think it's the wrong way of thinking about these new tools that are emerging.

s165
00:15:00.720 --> 00:15:05.040
I think like when I think of a factory, I think of automation.

s166
00:15:05.040 --> 00:15:13.120
And I think of like there's a lot of amazing things about automation and like it makes our lives more efficient, and we can like create more of everything.

s167
00:15:13.120 --> 00:15:17.320
But I don't want the future to be built around factories.

s168
00:15:17.320 --> 00:15:20.520
I want the future to I want to feel like a human.

s169
00:15:20.520 --> 00:15:22.040
I want to like be in the flow.

s170
00:15:22.040 --> 00:15:34.080
I want to be like in front of an orchestra like waving my baton and we're and like I wave it this way and this this team of agents starts working and then this intermingling of humans and agents starts working as I go here.

s171
00:15:34.080 --> 00:15:38.560
And I can when I want to I can zoom in on the details, but most of the time I can zoom out.

s172
00:15:38.560 --> 00:15:50.080
And I don't think the future should be we are like managing swarms of agents and we are like factory line managers like pushing buttons getting the agents to like pump out the next feature.

s173
00:15:50.080 --> 00:15:56.280
Like we can we can we tried this like 10 years ago with the term feature feature factories and it just doesn't work.

s174
00:15:56.280 --> 00:15:59.960
Like I want my software to feel human and crafted.

s175
00:15:59.960 --> 00:16:02.480
I want to feel like a human at the center of it all.

s176
00:16:02.480 --> 00:16:09.080
And I think because we're all building these tools we actually have a responsibility to make the tools great for humans.

s177
00:16:09.080 --> 00:16:18.920
I think it's really important to like use the words that make us feel excited and like feel feel capable and feel like we're in the flow and having fun.

s178
00:16:18.920 --> 00:16:24.240
And so I don't think the future is is is something like this.

s179
00:16:24.240 --> 00:16:27.920
I don't want to be I don't want to be in in my dark factory.

s180
00:16:27.920 --> 00:16:29.240
I don't want to be a line manager.

s181
00:16:29.240 --> 00:16:31.880
I want to feel like this.

s182
00:16:31.880 --> 00:16:32.760
I want to be in the flow.

s183
00:16:32.760 --> 00:16:33.720
I want to be having fun.

s184
00:16:33.720 --> 00:16:35.040
I want to be crafting things.

s185
00:16:35.040 --> 00:16:42.520
I want to feel like I'm Steve Jobs designing the Mac with a team of amazing humans and AI agents all in the same place.

s186
00:16:42.520 --> 00:16:46.520
Like I want to feel like I'm in an orchestra.

s187
00:16:46.520 --> 00:16:52.400
So here are my principles for for being the best builder in your organization.

s188
00:16:52.400 --> 00:16:54.000
Stay near the frontier.

s189
00:16:54.000 --> 00:16:55.839
Don't try and beat the market.

s190
00:16:55.839 --> 00:16:57.440
Create slop free zones.

s191
00:16:57.440 --> 00:16:58.839
Feed the beast.

s192
00:16:58.839 --> 00:17:04.839
Free range agents and think about orchestras not factories.

s193
00:17:05.199 --> 00:17:07.480
Came up with this handy acronym for for remembering it.

s194
00:17:07.480 --> 00:17:08.880
Stickfo.

s195
00:17:08.880 --> 00:17:09.199
All right.

s196
00:17:09.199 --> 00:17:11.439
So thanks thanks a ton for for having me.

s197
00:17:11.439 --> 00:17:12.280
I'll be around.

s198
00:17:12.280 --> 00:17:17.000
Feel free to ask questions and I'll see you on the internet.
