WEBVTT

1
00:00:00.171 --> 00:00:01.172
You need a frontier team.

2
00:00:01.532 --> 00:00:01.973
Here's why.

3
00:00:03.114 --> 00:00:05.937
Every single person in your company is caught in this dilemma.

4
00:00:06.617 --> 00:00:15.766
Do I get my work done or do I stay up to date with AI and figure out what all these new tools and workflows mean for what work I should do?

5
00:00:16.107 --> 00:00:16.947
That's just crazy.

6
00:00:16.968 --> 00:00:17.748
It's so hard to figure out.

7
00:00:17.828 --> 00:00:19.770
Every single company is struggling with this right now.

8
00:00:20.611 --> 00:00:32.419
Because, I mean, on the one hand, you could ignore the frontier and that'll probably work for a little while until you get disrupted by some agent native version of your product that you didn't see coming.

9
00:00:32.919 --> 00:00:37.522
Or you can keep up with the frontier and just chase shiny objects and never build anything that lasts.

10
00:00:37.863 --> 00:00:40.865
The reason why you need a frontier team is it separates things out.

11
00:00:41.305 --> 00:00:48.309
So you get this nice separation of concerns where you're both keeping up with what's going on on the frontier and you're actually getting your work done.

12
00:00:49.270 --> 00:00:50.931
And I had my co-host, Natalia.

13
00:00:50.951 --> 00:00:51.632
Welcome, Natalia.

14
00:00:52.512 --> 00:00:53.013
Hey, Dan.

15
00:00:53.033 --> 00:00:53.573
Good to be here.

16
00:00:53.993 --> 00:00:56.835
Who is our head of enterprise AI adoption.

17
00:00:56.915 --> 00:01:01.038
Come in and do a little bit of digging into our Frontier team.

18
00:01:01.058 --> 00:01:06.041
So we've been running the Frontier team at Every for, I guess, a couple of months now.

19
00:01:06.241 --> 00:01:08.863
And it has been kind of transformative.

20
00:01:09.303 --> 00:01:15.487
And I think Natalia is one of the best people to come in and take a peek and be like, what are we actually doing?

21
00:01:16.211 --> 00:01:21.343
So actually, I might turn the tables a little bit to you because it was your idea to have a frontier team.

22
00:01:22.134 --> 00:01:22.294
Yeah.

23
00:01:22.354 --> 00:01:29.440
I mean, yeah, the frontier team, I think has been a huge unlock and it's because of this separation that, that I've been talking about.

24
00:01:29.620 --> 00:01:34.364
I think there's so many people at every like me who just love playing with new things and that's super important.

25
00:01:34.744 --> 00:01:38.528
But if that's all you're doing, then you're, you're not going to make progress as a business.

26
00:01:38.548 --> 00:01:43.692
Like you actually do need to go execute on products and services, things that people can rely on.

27
00:01:44.252 --> 00:01:46.254
And it felt like we needed to choose one or the other.

28
00:01:46.274 --> 00:01:46.374
Yeah.

29
00:01:46.454 --> 00:01:53.840
either just constantly be playing with new stuff or make a product that we just build once and then we add features to and we support forever.

30
00:01:54.520 --> 00:02:03.447
And the Frontier team was in a sort of an attempt to say, okay, what if we take all the people like me who love all the new stuff and we just unleash us to just run experiments?

31
00:02:04.067 --> 00:02:08.831
And at the low end, the experiments are just, they're just really good content.

32
00:02:08.851 --> 00:02:12.914
They just teach us something interesting and they become good tweets or YouTube videos or anything like that.

33
00:02:13.534 --> 00:02:28.541
And then on the high end, they get pushed up from the frontier into the products that we build, into the services that we offer, without the rest of the company feeling like they always had to be paying attention to every single model that drops in every single new experiment that we're running.

34
00:02:28.601 --> 00:02:42.388
So I felt like it's allowed me and the other people on the frontier team the space to just figure out what the future looks like without forcing that on the rest of the company until it's actually ready for them to pay attention to it.

35
00:02:43.076 --> 00:02:51.402
Now, one of the things that you seem to have figured out pretty early in the formation of the Frontier team is that you wanted to have both pirates and architects.

36
00:02:51.422 --> 00:02:52.723
Can you describe what that is?

37
00:02:53.303 --> 00:02:57.767
I think there's two archetypes that work really well together, pirates and architects.

38
00:02:58.527 --> 00:03:08.715
Pirates are people like me who are just like, I just want to go really fast and I'm going to do stuff that's kind of be a little bit sloppy, but I'm going to try like a billion things in the space of a couple days.

39
00:03:09.996 --> 00:03:11.577
And I'm going to figure out what's useful.

40
00:03:12.763 --> 00:03:13.544
but it's gonna be a mess.

41
00:03:14.265 --> 00:03:18.592
Architects are people who are really good at being like, there's a whole big messy thing

42
00:03:20.585 --> 00:03:28.111
what would it look like to compress it a little bit and form it into a shape that actually works, that we can actually use?

43
00:03:28.131 --> 00:03:32.474
Because what we try to do on the Frontier team is we're experimenting with lots of stuff.

44
00:03:32.674 --> 00:03:33.875
Most of that stuff gets thrown out.

45
00:03:34.355 --> 00:03:39.339
And some of it ends up getting pushed into the rest of the team to see if the rest of the team likes it.

46
00:03:39.759 --> 00:03:45.964
And at that point, there's a point right before we do that that an architect becomes really useful to come in and say like,

47
00:03:46.604 --> 00:04:01.495
this is messy, we can't actually maintain this and it wouldn't work for these five different ways for the people internally at the company and helps turn it into a system that we can actually understand and build on instead of just being like a kind of one-off vibe-coded project.

48
00:04:02.176 --> 00:04:09.041
Dan, what do you think has been the most impactful thing you've seen about having a Frontier team at every?

49
00:04:10.021 --> 00:04:15.345
I think it lets us move incredibly quickly on what's new.

50
00:04:16.352 --> 00:04:17.973
without distracting the whole team.

51
00:04:18.133 --> 00:04:29.462
And that has a deeply, um, it's like confidence building for the team because you feel like, Hey, I'm going to know when things change, we're going to, we'll be able to handle it.

52
00:04:30.222 --> 00:04:33.084
And I'm not feeling crazy and frazzled and distracted all the time.

53
00:04:33.164 --> 00:04:41.230
And I think that is like the core important thing that affects the day-to-day experience of building and working in and running a company in this environment.

54
00:04:42.451 --> 00:04:42.971
Absolutely.

55
00:04:43.272 --> 00:04:43.812
Thanks, Dan.

56
00:04:44.573 --> 00:04:44.873
Thank you.

57
00:04:57.798 --> 00:05:00.360
So, Mike Taylor, welcome.

58
00:05:00.380 --> 00:05:04.842
We're so excited to learn about the work that you're doing on the Frontier team as the head of evals.

59
00:05:06.163 --> 00:05:09.445
So, can you tell us a little bit of what you're doing as part of the Frontier team?

60
00:05:10.253 --> 00:05:16.760
Yeah, so it feels like it's different every week or maybe every day, sometimes every hour.

61
00:05:17.080 --> 00:05:30.075
But the goal of the Frontier team is really just to kind of push the models to the limits and see what is now possible and what is a new way to work, given that the models are constantly getting better.

62
00:05:30.575 --> 00:05:32.157
And there's still a lot of...

63
00:05:32.697 --> 00:05:34.118
capability overhang.

64
00:05:34.618 --> 00:05:40.561
What I mean by that is the models are able to do much more than we have had time to try.

65
00:05:41.461 --> 00:05:44.023
And so we're trying to find those edges.

66
00:05:45.803 --> 00:05:48.945
And specifically, what is your title as part of the Frontier team?

67
00:05:49.333 --> 00:05:55.675
Yeah, so I made up my own job title, which is very me, but it's head of evals.

68
00:05:55.795 --> 00:05:59.216
So what that means, evals is short for evaluations.

69
00:05:59.576 --> 00:06:09.919
And so it's my job to kind of take the kind of vibe checks that we've been doing with models that are more like qualitative and then try and put some numbers behind them.

70
00:06:10.039 --> 00:06:14.140
So I've been working with a lot of the team to help them build their own personal benchmarks.

71
00:06:14.940 --> 00:06:15.640
including with you.

72
00:06:16.161 --> 00:06:25.664
And we're trying to roll that out as more of an internal product initially to get more people involved with testing the models.

73
00:06:26.004 --> 00:06:31.086
And I can talk a bit more about that, but also just running more experiments.

74
00:06:31.506 --> 00:06:35.327
When we say that AI is sycophantic, what do we mean by that?

75
00:06:35.367 --> 00:06:36.108
How do we define that?

76
00:06:36.168 --> 00:06:37.748
And can we test that?

77
00:06:37.768 --> 00:06:38.628
Those sorts of things.

78
00:06:39.389 --> 00:06:42.550
And then just also just pushing it a little bit further.

79
00:06:42.610 --> 00:06:42.810
So...

80
00:06:43.438 --> 00:06:45.254
And so what are you most excited about now?

81
00:06:46.370 --> 00:06:46.530
Yeah.

82
00:06:46.610 --> 00:06:52.031
One of the things I've been finding more is that Opus is like very good at video editing.

83
00:06:52.291 --> 00:07:00.733
And this is, uh, you know, maybe obvious to some people who have been testing it, but I've been testing AI on video editing for a while.

84
00:07:00.873 --> 00:07:08.335
I have a popular Udemy course, uh, and, uh, it's like the worst part of that side project is having to edit the videos afterwards.

85
00:07:08.915 --> 00:07:10.156
Um, so, uh, for me, I've

86
00:07:16.817 --> 00:07:19.979
like a lot that I can do that I couldn't previously have done.

87
00:07:20.599 --> 00:07:27.503
So one of the things I did recently was I made an explainer video for the Chex product for the personal benchmark product that we're making.

88
00:07:28.343 --> 00:07:38.068
I asked Opus to go back through all my Slack messages, go back through all my granola notes, and then pull the narrative of the video together with receipts, right?

89
00:07:38.088 --> 00:07:42.371
Like with references to the specific things that I'd said in Slack or in meetings.

90
00:07:42.691 --> 00:07:44.852
And then it like cut that whole video together

91
00:07:45.453 --> 00:07:47.656
Uh, and, and like, then I had an explainer video.

92
00:07:47.737 --> 00:07:54.648
So I feel like that that's opening up a lot for people who are like me, who may be too busy running experiments to ever like tell anyone what we're doing.

93
00:07:55.951 --> 00:07:56.532
So cool.

94
00:07:57.072 --> 00:07:58.154
Can you show us an example?

95
00:07:59.060 --> 00:08:01.341
So this is the one I was talking about.

96
00:08:02.362 --> 00:08:05.503
So this prompt goes back really, really long way.

97
00:08:06.544 --> 00:08:09.566
But I went back and forth with Opus on this.

98
00:08:10.306 --> 00:08:18.110
And initially it was in UltraCode, which if you don't know, is like one of the settings that is like the highest effort setting.

99
00:08:18.570 --> 00:08:21.412
And so it will use hundreds of agents

100
00:08:22.112 --> 00:08:24.974
basically that it's orchestrating to do the task.

101
00:08:25.074 --> 00:08:31.657
And the reason I chose UltraCode for this is I knew that had a lot of context that I wanted it to read.

102
00:08:32.077 --> 00:08:34.858
And like one model can't read all of that context.

103
00:08:34.978 --> 00:08:45.683
So what it does is it sent hundreds of agents, you know, one agent per meeting that I had, and then pulled everything back into this like library essentially.

104
00:08:47.400 --> 00:08:48.481
These are all the sources, right?

105
00:08:48.501 --> 00:08:53.604
So you can see all the different calls I had with different people, and it found all the ones that are relevant.

106
00:08:54.205 --> 00:08:55.386
But now we have it all.

107
00:08:55.426 --> 00:08:56.927
We have this big library of sources.

108
00:08:57.747 --> 00:08:59.609
But it didn't just do that, right?

109
00:09:00.089 --> 00:09:06.894
It also then created this story spine, which is like it tried to boil it down to what is the narrative?

110
00:09:06.934 --> 00:09:10.196
Like if someone doesn't know the Chex product, what is Chex?

111
00:09:10.616 --> 00:09:13.138
And it boiled it down to like these 12 sentences.

112
00:09:13.158 --> 00:09:16.140
So like personal benchmark is a small set of your own real work,

113
00:09:16.540 --> 00:09:18.941
saved so that am.org can run on it again.

114
00:09:19.021 --> 00:09:21.702
So that's the definition I just gave you earlier.

115
00:09:22.062 --> 00:09:23.482
But now we have the receipts, right?

116
00:09:23.522 --> 00:09:26.983
I can click in and see where did that sentence come from?

117
00:09:27.463 --> 00:09:31.664
Back to the insight card, and then from the insight card back to the specific things I said in meetings.

118
00:09:31.844 --> 00:09:34.325
So it's all auditable.

119
00:09:34.485 --> 00:09:39.627
And then it puts everything together, and then we have this explainer video, right?

120
00:09:40.087 --> 00:09:40.727
Yeah, so.

121
00:09:42.612 --> 00:09:46.093
So the interesting thing is, so like see all these screenshots?

122
00:09:46.713 --> 00:09:49.993
Like it went and found those screenshots from my Slack.

123
00:09:50.574 --> 00:09:54.914
These are things that I shared, you know, and then like it put everything together.

124
00:09:54.954 --> 00:09:56.395
It annotated the screens.

125
00:09:57.355 --> 00:09:58.715
It did the voiceover.

126
00:09:58.895 --> 00:10:00.015
I used Google Gemini.

127
00:10:00.155 --> 00:10:01.776
I gave it a Gemini API key.

128
00:10:02.296 --> 00:10:06.377
And so these are the 12 sentences, but like shown in video form.

129
00:10:06.837 --> 00:10:07.777
And that is

130
00:10:08.697 --> 00:10:10.259
Again, pretty amazing.

131
00:10:10.779 --> 00:10:13.322
And then I can share that with the rest of the company.

132
00:10:13.422 --> 00:10:17.046
And then now they understand, like, oh, yeah, a score is a count of yeses, right?

133
00:10:17.066 --> 00:10:19.628
So, like, now they understand the Chex product.

134
00:10:20.409 --> 00:10:26.815
So maybe last question for you would be, you know, as head of evals, you're evaluating a bunch of models.

135
00:10:26.835 --> 00:10:29.298
You're seeing what actually works, what the new capabilities are.

136
00:10:29.926 --> 00:10:32.750
What is the cheat sheet that you could give someone?

137
00:10:33.030 --> 00:10:37.837
I know it's late October 2026, so this might be outdated in a matter of days.

138
00:10:37.977 --> 00:10:43.644
But can you give us a cheat sheet for what we should know about using models today?

139
00:10:43.664 --> 00:10:43.965
Yeah.

140
00:10:44.462 --> 00:10:45.823
So what is my cheat sheet?

141
00:10:45.863 --> 00:10:53.831
So I think one thing that people are not doing enough of is just giving all their context over to AI.

142
00:10:53.851 --> 00:11:04.220
I think previously, if you've tried this in the past and you've asked AI to look through your call notes or maybe been hesitant to give it access to your Google Drive or whatever...

143
00:11:04.680 --> 00:11:06.781
And I was that person like earlier in the year.

144
00:11:06.841 --> 00:11:13.102
I didn't have it cooked up to my Gmail or my Slack and living within the every community, like being in the team.

145
00:11:13.462 --> 00:11:16.443
It's like now I'm like full on like AI has access to everything.

146
00:11:16.763 --> 00:11:24.265
And I would say like just make that leap because it's amazing how good it is at picking out good context once you give it access.

147
00:11:24.305 --> 00:11:27.705
So I'd say like that's gotten so much better and people need to do more of that.

148
00:11:28.366 --> 00:11:34.367
And you can trust it more to use its judgment in what it brings into that task.

149
00:11:35.007 --> 00:11:47.697
The second thing is that I feel like it's your job, like now that you're producing a lot more, hopefully if you're using AI more, it's your job now to protect other people's context.

150
00:11:47.757 --> 00:11:53.882
So I'm like regularly accused of saturating the company LinkedIn, like every channel I'm posting in.

151
00:11:54.242 --> 00:11:55.503
It's like Mike, Mike, Mike.

152
00:11:55.843 --> 00:11:57.444
And I'm really conscious of that.

153
00:11:57.484 --> 00:12:04.408
So creating that video, creating more videos like that, making the work that I'm doing more digestible, I think is now actually my job.

154
00:12:06.089 --> 00:12:07.489
I'm coming up with good ideas.

155
00:12:07.529 --> 00:12:08.570
I'm giving it to Claude.

156
00:12:08.770 --> 00:12:13.032
Claude's finding those ideas more automatically for me when I've said them.

157
00:12:13.152 --> 00:12:15.794
But now I'm flooding people with new ideas.

158
00:12:15.934 --> 00:12:17.875
So that's the other part that I'm working on.

159
00:12:17.895 --> 00:12:20.576
I think that that's where a lot of people's jobs are going to go.

160
00:12:21.537 --> 00:12:21.997
Super cool.

161
00:12:22.097 --> 00:12:23.378
Thanks so much for joining us, Mike Taylor.

162
00:12:23.897 --> 00:12:24.337
Yeah, yeah.

163
00:12:24.397 --> 00:12:26.479
Good to talk to you.

164
00:12:26.499 --> 00:12:27.840
So welcome, Arielle.

165
00:12:27.860 --> 00:12:28.861
So excited to have you.

166
00:12:29.261 --> 00:12:32.904
And can you tell us what your role is as part of the Frontier team at Every?

167
00:12:33.685 --> 00:12:50.959
Yes, I'm officially the Frontier Wrangler, which means that I help make sure that our experimentation is in some ways organized enough for us to then harvest and synthesize learnings and digest it for the outside world and direct our

168
00:12:52.060 --> 00:12:53.222
exploration internally.

169
00:12:53.922 --> 00:12:59.188
So I work on the systems and processes that allow us to do that internally.

170
00:12:59.208 --> 00:13:00.770
And I partner closely with Dan on that.

171
00:13:01.271 --> 00:13:04.695
And then I'm also running my own experimentation.

172
00:13:04.755 --> 00:13:10.461
So unlike other folks on the team, I'm less technical, but I'm an operator.

173
00:13:10.601 --> 00:13:11.342
And so there's

174
00:13:12.323 --> 00:13:24.620
A lot of interesting opportunity for practical systems that we can experiment with using some of these new tools and techniques on the inside that I'm playing around with pretty consistently.

175
00:13:25.902 --> 00:13:28.045
Why does the Frontier team need a Wrangler?

176
00:13:28.930 --> 00:13:36.977
I think one of the cool things about working with folks in the org is that everybody is a meta thinker.

177
00:13:37.057 --> 00:13:41.361
People are so creative and they're so excited to just sort of chase the next thing.

178
00:13:43.142 --> 00:13:43.743
And that's great.

179
00:13:43.903 --> 00:13:44.984
We need people like that.

180
00:13:45.144 --> 00:13:50.648
And then we also need a organizational layer to say, like, what are we learning from this?

181
00:13:50.788 --> 00:13:55.532
And how can we actually digest everything that's coming out of these experiments?

182
00:13:56.233 --> 00:13:57.475
for our audience.

183
00:13:58.076 --> 00:14:05.846
And so I think it's pretty important to have a layer that basically looks at everything that people are doing

184
00:14:07.053 --> 00:14:23.903
maps it to digestible themes and tries to find some of the meaning in all of this experimentation, and then creates a layer that allows us to turn it into content and products and even concepts that we're teaching in consulting.

185
00:14:23.963 --> 00:14:28.805
So it's a way of harvesting all of this and then translating it into other areas of the business.

186
00:14:29.566 --> 00:14:31.007
So what are you working on now?

187
00:14:31.752 --> 00:14:37.613
With all of these models coming out constantly, the frontier of AI capability keeps pushing forward.

188
00:14:37.733 --> 00:14:38.834
And that's really exciting.

189
00:14:39.594 --> 00:14:44.875
That also means that there's so much to play with and experiment with that we are constantly burning through tokens like Wildfire.

190
00:14:45.275 --> 00:14:49.316
And that's an important investment in our business and experimentation.

191
00:14:49.676 --> 00:14:54.557
But also understanding at a meta level, like what experimentation are we doing?

192
00:14:54.697 --> 00:14:56.418
How much money are we spending on that?

193
00:14:56.758 --> 00:15:01.219
And what are we getting out of it in return is pretty important.

194
00:15:01.859 --> 00:15:12.519
Can you show us an example of kind of like a recent like insight or maybe process that you've tried to apply lately in your job as a Wrangler?

195
00:15:13.413 --> 00:15:14.374
Yeah, definitely.

196
00:15:14.654 --> 00:15:17.715
So this is what the usage dashboard looks like.

197
00:15:17.916 --> 00:15:23.939
Some of this is going to be a retrospective because we don't have a full month of spend for October yet.

198
00:15:24.760 --> 00:15:29.943
But what this basically allows us to do is understand, like, what is the total dollar value that we've spent?

199
00:15:30.483 --> 00:15:31.824
What are the tokens?

200
00:15:32.164 --> 00:15:34.425
Like, what's the token count that we've spent it on?

201
00:15:34.985 --> 00:15:35.346
And then...

202
00:15:36.066 --> 00:15:37.768
There's actually a breakdown by person.

203
00:15:38.529 --> 00:15:47.638
So we created a skill that looks across all the different harnesses that you're using on your computer to harvest.

204
00:15:48.399 --> 00:15:52.583
These are the tokens that you used across each of these harnesses.

205
00:15:52.603 --> 00:15:54.485
You can also click in and see...

206
00:15:56.353 --> 00:15:59.477
There's, we're collecting data in real time.

207
00:15:59.537 --> 00:16:02.640
So it gives you the actual like model and effort level.

208
00:16:03.401 --> 00:16:15.674
And this dashboard also includes a workflow that basically looks across all of your sessions and categorizes at a high level what you want.

209
00:16:16.615 --> 00:16:34.961
spent those tokens on and then runs it by you for approval so that you can see, you know, both the percentage share of the work of the token spend that's gone to each different work bucket and then what the estimated dollar value of those tokens are.

210
00:16:35.021 --> 00:16:38.843
So you can make a sort of quick calculation of like, oh, like, was that worth it or not?

211
00:16:39.243 --> 00:16:39.503
All right.

212
00:16:39.603 --> 00:16:41.584
And Ariel, maybe last question for you.

213
00:16:41.644 --> 00:16:44.925
What are you most excited about next as the Frontier Team Wrangler?

214
00:16:46.332 --> 00:16:54.402
I think what I'm most excited about is as these models get smarter and smarter, my ability to actually build the capacity for that grows.

215
00:16:54.662 --> 00:17:01.130
It's not a power that I've had before, and it feels magical, and it feels really exciting.

216
00:17:01.410 --> 00:17:05.115
And being able to take some of the ideas that I have and –

217
00:17:05.415 --> 00:17:08.257
And will them into existence is actually a totally new thing.

218
00:17:08.297 --> 00:17:22.649
So in my previous life as an operator, I would have waited had to wait for different model companies to add this kind of tracking capability that I'm looking for, or I would have had to partner with an engineer to do any of this.

219
00:17:23.690 --> 00:17:26.752
And now I actually have the ability to create some of it myself.

220
00:17:26.832 --> 00:17:38.058
So what I'm looking forward to is as these models get better and better, my capacity to actually build systems to do this work for us gets better and better.

221
00:17:38.399 --> 00:17:40.360
And I'm really excited about diving into that.

222
00:17:41.561 --> 00:17:42.661
I'm excited to see you, Arielle.

223
00:17:42.821 --> 00:17:43.962
Thank you for joining us, Arielle.

224
00:17:44.442 --> 00:17:44.642
Cool.

225
00:17:44.822 --> 00:17:45.323
Thanks so much.

226
00:17:46.023 --> 00:17:50.466
One of the great things about working with AI is you can just talk to it now.

227
00:17:51.640 --> 00:17:58.628
If you're interested in groundbreaking voice experiences, you need to check out Gemini Audio.

228
00:17:59.309 --> 00:18:02.513
Gemini Audio is the new voice models from Google DeepMind.

229
00:18:03.093 --> 00:18:07.398
Gemini 3.5 Transcribe turns speech into clean formatted text.

230
00:18:08.547 --> 00:18:16.109
If you say, let's meet Tuesday, no Wednesday, it'll detect the correction and handle it automatically and remove filler words.

231
00:18:16.669 --> 00:18:20.370
There's less cleanup between a rough thought and something you can actually use.

232
00:18:21.570 --> 00:18:27.631
And now with Gemini 3.8 live extended thinking, you can talk through more complex work.

233
00:18:28.231 --> 00:18:35.453
It can reason and speak simultaneously, which keeps conversations going while it's working in the background on your behalf.

234
00:18:36.889 --> 00:18:40.231
You can even ask follow-up questions and give feedback as you go on together.

235
00:18:40.931 --> 00:18:46.734
It makes voice a useful way to develop an idea with room to change your mind along the way.

236
00:18:47.514 --> 00:18:56.438
You should check out the link in the description and learn more about Gemini 3.5 Transcribe and Gemini 3.8 Live Extended Thinking.

237
00:18:57.098 --> 00:18:58.459
And now, back to the episode.

238
00:18:58.819 --> 00:19:01.702
Yeah, so I'm Jack Chang.

239
00:19:01.982 --> 00:19:04.264
I'm a senior editor at Every.

240
00:19:05.625 --> 00:19:07.727
And yeah, part of the Frontier team.

241
00:19:07.747 --> 00:19:13.433
I think my role on the Frontier team...

242
00:19:14.709 --> 00:19:19.951
I think we're all figuring out what our roles are on the Frontier team.

243
00:19:19.971 --> 00:19:26.034
And I've definitely been testing more of the early access models that we've been getting.

244
00:19:26.294 --> 00:19:38.299
And something I've really started doing more of there is basically just creating these interactive experiences and games and things like that, and to just really

245
00:19:39.720 --> 00:19:45.225
try to push at what is even capable now with each new model release.

246
00:19:45.325 --> 00:19:50.171
What can it do that it maybe wasn't good enough?

247
00:19:50.271 --> 00:19:55.296
It wasn't good enough that I would even think to try with a model before.

248
00:19:55.316 --> 00:19:55.416
Yeah.

249
00:19:56.598 --> 00:20:02.685
And so lately that's kind of meant I'm just like creating video games and things like that.

250
00:20:03.686 --> 00:20:10.114
And then the other piece of it is just actually like making my own applications.

251
00:20:10.134 --> 00:20:10.674
So, you know,

252
00:20:12.043 --> 00:20:20.849
I was finding that I was really getting overwhelmed by my browser's tabs and being constantly forced to contact switch.

253
00:20:21.569 --> 00:20:28.013
And I was like, well, wouldn't it be nice if I had a browser that didn't have tabs and that you could turn off the tabs?

254
00:20:29.995 --> 00:20:35.738
And I think before, I would have maybe built a Chrome extension or something like that.

255
00:20:36.499 --> 00:20:40.782
But now the models are so good that I was like, no, I can just build my own.

256
00:20:41.462 --> 00:20:45.326
So here is my browser, basically.

257
00:20:46.207 --> 00:20:48.509
Yeah, like I mentioned, it's a tabless browser.

258
00:20:49.289 --> 00:21:00.180
And that was the original intent, is to have this browser where I'm not tempted to switch tabs and open things in the background and never get to them.

259
00:21:01.340 --> 00:21:14.044
But once I had this going, I was like, well, in order for me to use this day-to-day, I have my whole editing workflow that I normally use ChatGPT in the browser for.

260
00:21:14.104 --> 00:21:16.265
So basically, I added this agent pane.

261
00:21:17.546 --> 00:21:32.264
That could let me pick any model provider and use any model for editing, which I couldn't really even do with Claude and the Gemini models because their plugins for Chrome weren't as good.

262
00:21:33.325 --> 00:21:41.390
but there are things like, you know, like just in sort of daily use, I'm like, oh, there are times when I actually do need to like peek at something.

263
00:21:42.050 --> 00:21:53.838
And so the latest thing I added was sort of this kind of like peak mode where I could like very quickly temporarily open up a page and then like, you know, like look up something and then go back to my document.

264
00:21:55.979 --> 00:22:01.023
And I think like what I'm starting to learn with this is just the way that like

265
00:22:03.099 --> 00:22:16.498
I think we had this concept of plugins and extensions for a while, but those things never really let you get close to the actual application and changing how the application worked.

266
00:22:17.079 --> 00:22:35.110
And now with sort of the latest generation of like frontier models, it's really like possible to build this like, you know, fully like personalized to you or personalized to your team and their workflows application that you couldn't do before.

267
00:22:35.130 --> 00:22:36.230
All right.

268
00:22:36.310 --> 00:22:43.775
I think maybe a good closing question might be, what are you most excited about building next?

269
00:22:45.065 --> 00:22:56.293
I think I'm most excited about building more things that help basically protect my attention.

270
00:22:58.154 --> 00:23:07.181
And I think with all this, first of all, I have ADHD, have a diagnosis of ADHD, and

271
00:23:09.823 --> 00:23:13.226
And that's always just been a huge challenge throughout my life.

272
00:23:13.266 --> 00:23:18.851
And I've kind of had to come up with like these like adaptive strategies throughout the years in order to manage it.

273
00:23:19.431 --> 00:23:26.738
And, and I think like now, you know, it's even more kind of like tempting to kind of just like try to do a million things.

274
00:23:27.239 --> 00:23:34.305
And, and, and yeah, and I do feel like I am like doing much so much more now, but at the same time.

275
00:23:35.066 --> 00:23:54.417
I've never felt such strain on my ability to focus and ability to like keep my attention, um, on something for long enough that, you know, that, um, uh, it's, it's sort of, yeah, for, for long enough that I think it, it makes a meaningful difference, uh, in my life.

276
00:23:55.157 --> 00:23:58.980
Um, and, you know, experimenting and experimentation is definitely like, um,

277
00:23:59.780 --> 00:24:00.501
important to me.

278
00:24:01.081 --> 00:24:10.549
But I think there's also this other side of me that's like, yeah, you know, I want to build things that like, help me write more help me like create more beautiful things in the world.

279
00:24:12.231 --> 00:24:18.296
I want to build things that, you know, connect me closer to like my local community.

280
00:24:19.176 --> 00:24:21.819
And so really, like thinking about it, not as just like,

281
00:24:23.169 --> 00:24:33.673
building something for creativity sake, but like doing it in a way that like also helps, um, you know, reinforce and serve and deepen my values.

282
00:24:35.273 --> 00:24:35.694
Wonderful.

283
00:24:35.834 --> 00:24:37.094
Thank you so much for joining us, Jack.

284
00:24:37.934 --> 00:24:39.115
Welcome Katie.

285
00:24:39.615 --> 00:24:46.157
Uh, so excited to be back on the pod with you and to be talking about your role as part of the frontier team.

286
00:24:46.337 --> 00:24:47.658
What do you do in the frontier team?

287
00:24:48.621 --> 00:24:51.763
Yeah, so there are a couple different things that I do on the Frontier team.

288
00:24:52.323 --> 00:25:06.313
Number one, I'm on the testing team for the new Frontier models that come out from OpenAI, Anthropic, Gemini, Grok.

289
00:25:06.493 --> 00:25:07.733
We are on the test.

290
00:25:08.114 --> 00:25:09.695
We get early access to these models.

291
00:25:10.095 --> 00:25:14.538
We're putting them through our workflows and just seeing, okay, what is it like to use this model properly?

292
00:25:15.078 --> 00:25:18.641
So we can turn around and pass that information along to the people who read our vibe checks.

293
00:25:18.721 --> 00:25:20.542
So that's kind of one of my lanes.

294
00:25:21.042 --> 00:25:28.947
The other lane is writing the vibe checks themselves and in other ways, kind of like sharing out what we're learning on the Frontier team.

295
00:25:28.987 --> 00:25:29.788
So I'm kind of like...

296
00:25:30.308 --> 00:25:37.289
the resident writer embedded inside the Frontier team to write about what the Frontier team is learning.

297
00:25:37.529 --> 00:25:39.290
And then I'm building stuff myself.

298
00:25:39.330 --> 00:25:42.730
And that's where stuff like my plugin Compound Writing comes in.

299
00:25:43.370 --> 00:25:49.651
That's where using new models that are fine tuned for writing comes in.

300
00:25:50.212 --> 00:25:58.053
And that also includes crash test driving, being a crash test dummy for some of the things that other people on the Frontier team are building, like,

301
00:25:58.593 --> 00:26:01.096
Kieran and Mike as well.

302
00:26:02.558 --> 00:26:05.161
And can you give us an example of something that you're working on now?

303
00:26:06.182 --> 00:26:06.462
Yeah.

304
00:26:06.522 --> 00:26:10.667
So right now I am, I'm continuing to build out compound writing.

305
00:26:11.768 --> 00:26:18.395
And so for folks who don't know, that is my plugin that sort of takes you through the writing process from ideation through.

306
00:26:19.183 --> 00:26:22.888
planning and outlining through drafting to finalization.

307
00:26:22.908 --> 00:26:27.274
And the whole thing is up to this point has been built around two core documents.

308
00:26:27.394 --> 00:26:30.539
The voice guide that captures how your voice sounds on like

309
00:26:31.157 --> 00:26:42.966
a sentence level, and like the word choice that you use, and then style.md, which covers what you write about, how you like to talk about your subject matter.

310
00:26:43.707 --> 00:26:53.155
And right now what I'm doing is I'm adding a third document, which actually in hindsight feels very obvious to the writing process, and that is audience.md.

311
00:26:53.875 --> 00:26:58.660
Um, so that, so who is the reader that is ultimately going to be receiving this content?

312
00:26:58.740 --> 00:27:04.867
And so I'm adding that to my plugin and saying, okay, how does this change the, the experience of using this plugin?

313
00:27:04.887 --> 00:27:06.889
Doesn't make it more, more useful.

314
00:27:06.949 --> 00:27:08.190
So I can share my screen.

315
00:27:09.767 --> 00:27:25.460
And so for anyone who's not familiar with Katy Parrott's compound writing plugin, you can find it on our GitHub and you can just point your AI of choice to the page and you can download it and it'll take you to this, yes?

316
00:27:25.960 --> 00:27:27.201
And so what are we seeing here?

317
00:27:28.002 --> 00:27:36.930
So this is me setting up my audience.md for my column, Working Over Time, which is where I write about my misadventures and regular adventures.

318
00:27:38.070 --> 00:27:40.752
you know, applying AI in my own working life.

319
00:27:40.772 --> 00:27:47.296
So I wanted to set up an audience.md folder for that work that's specific to this column.

320
00:27:48.197 --> 00:27:53.020
And it gives me this readout of who the working overtime audience is.

321
00:27:53.080 --> 00:28:01.445
So what the territory it covers, you know, the main reader being a combination of like the early seeker who's kind of AI curious.

322
00:28:01.925 --> 00:28:06.368
And then that operator who's looking for like practical applications to their day-to-day work.

323
00:28:06.849 --> 00:28:15.500
Something I kind of baked into how the audience.md is structured is the idea of a hero's journey, which I really like.

324
00:28:15.581 --> 00:28:19.987
Every piece of reading should take the reader on their own hero's journey in some form or other.

325
00:28:20.007 --> 00:28:20.808
You're looking to

326
00:28:21.620 --> 00:28:23.581
enact some kind of transformation in them.

327
00:28:23.621 --> 00:28:26.542
And so I wanted to include that in my audience MD.

328
00:28:27.382 --> 00:28:43.889
And then what that unlocks for us in compound writing, I went ahead and did a pre-bake where I just asked it to review my most recent working overtime draft for audience fit, which was my piece about how I accidentally made all of my context worse by trying to make it better.

329
00:28:43.909 --> 00:28:50.492
And it's using the audience MD that I updated today to then give like,

330
00:28:51.232 --> 00:28:53.454
an overview of what already works.

331
00:28:53.554 --> 00:28:57.678
So it says it's clear, context gets an explanation.

332
00:28:58.419 --> 00:29:05.345
The fix is personal, not universal, which is something that I've said that I don't like to prescribe things.

333
00:29:05.385 --> 00:29:06.626
I like to describe things.

334
00:29:07.787 --> 00:29:13.393
And then here is where it loses the secret side.

335
00:29:13.413 --> 00:29:14.774
This is really good feedback.

336
00:29:15.194 --> 00:29:16.656
It's like, where am I not...

337
00:29:17.196 --> 00:29:21.997
being accessible to people on the earlier side of that spectrum that I might want to be reaching.

338
00:29:22.677 --> 00:29:24.718
So this is really good stuff that I'd want to engage with.

339
00:29:24.758 --> 00:29:28.239
And then some smaller operator notes, like what was the damage?

340
00:29:28.739 --> 00:29:33.500
Like what was the cost to me when like I wasn't getting the results I wanted out of AI?

341
00:29:33.540 --> 00:29:46.423
So this is the kind of like higher detail feedback and like more value that I think we can bring to tools like compound writing to help empower people like make better

342
00:29:47.967 --> 00:29:50.034
Do you have a sense of what your next experiment might be?

343
00:29:51.458 --> 00:30:03.622
My next experiment, so, well, I'll be working with Kieran on every type, which is his bid to get us all to stop using Google Docs.

344
00:30:04.062 --> 00:30:09.904
So I'll be sort of like test driving that and being very annoying in my feedback and my particular preferences as a writer.

345
00:30:10.364 --> 00:30:13.865
I'm also kind of like building my own word processor.

346
00:30:13.945 --> 00:30:15.766
I always, I joke it's like a rite of passage.

347
00:30:15.786 --> 00:30:19.407
Like everyone at a certain point decides to vibe code their own word processor and,

348
00:30:19.847 --> 00:30:30.296
So I'm like building my own word processor word better on the side just to – because it's easier for me to articulate what I want in a tool just to build – at this point to just build it myself.

349
00:30:30.776 --> 00:30:33.919
And then I can show it to Kieran and say, here's what makes more sense to me.

350
00:30:34.820 --> 00:30:38.123
So that is kind of what I'm doing now.

351
00:30:39.043 --> 00:30:43.727
Like I'm just on that part of my journey where I feel like I can build a better word processor than Kieran.

352
00:30:43.867 --> 00:30:44.148
Yeah.

353
00:30:44.893 --> 00:30:45.614
Fantastic.

354
00:30:45.735 --> 00:30:46.256
I love it.

355
00:30:46.316 --> 00:30:47.758
Well, thank you for compound writing.

356
00:30:47.899 --> 00:30:49.722
I've personally really enjoyed using it.

357
00:30:49.963 --> 00:30:50.864
And thank you for joining us.

358
00:30:51.806 --> 00:30:52.808
So welcome, Yannick.

359
00:30:52.848 --> 00:30:53.930
So excited to have you.

360
00:30:54.391 --> 00:30:57.337
And can you tell us what is your role as part of the Frontier team?

361
00:30:58.617 --> 00:30:59.378
Yeah, sure.

362
00:31:00.218 --> 00:31:04.000
I'm mostly working as an architect on some of the projects.

363
00:31:05.981 --> 00:31:09.063
I took over a lot of projects from then.

364
00:31:09.843 --> 00:31:17.147
I think my first project here was working on proof, getting it production ready, getting it back to life, I would say.

365
00:31:17.827 --> 00:31:21.089
Nowadays, I'm mostly focusing on compounding.

366
00:31:22.510 --> 00:31:33.382
basically working on Cakebench, having like a skill which tries to replicate the judgment and taste of our editor-in-chief to reduce some of our work basically.

367
00:31:34.161 --> 00:31:44.152
And maybe tell us a little bit about that, because I think a lot of people, you know, have tried vibe coding something, building a product, and then it can break in all kinds of ways.

368
00:31:44.272 --> 00:31:50.278
You're the person that gets brought in to kind of clean up the mess and see what's really there.

369
00:31:50.438 --> 00:31:51.800
So what's that experience like?

370
00:31:52.582 --> 00:32:09.862
I think AI is really good at like bringing something to life, just making assumptions, looking up what to use and just build something that looks good and potentially also works for that demo case or initial implementation.

371
00:32:11.263 --> 00:32:24.672
But I think there's still need for the architect role who has that background knowledge into knowing how to build stuff that's reliable, that's maintainable over time.

372
00:32:26.092 --> 00:32:31.216
How you should structure code, what libraries to use, how to deploy it.

373
00:32:33.237 --> 00:32:38.440
Know what's important on the software side to make it ready for the future, I would say.

374
00:32:39.096 --> 00:32:41.736
Now, can you give us an example of something you're working on now?

375
00:32:43.037 --> 00:32:43.237
Sure.

376
00:32:43.357 --> 00:32:46.757
I think like, I'm currently working on like two work streams.

377
00:32:47.337 --> 00:32:49.358
One is compounding with like Cakebench I mentioned earlier.

378
00:32:50.058 --> 00:32:57.139
And the other one is more explorative where I see how should we work with AI.

379
00:32:57.519 --> 00:32:57.839
All right.

380
00:32:59.960 --> 00:33:02.360
So this is my orchestrator.

381
00:33:02.620 --> 00:33:03.260
It's called Minga.

382
00:33:04.400 --> 00:33:07.661
And what I can do here is, uh,

383
00:33:08.928 --> 00:33:09.728
create a project.

384
00:33:09.968 --> 00:33:12.590
For example, I created a new project called Replyguy.

385
00:33:14.170 --> 00:33:28.256
The idea is that we give a colleague, Becky, a tool which she can use to explore what's trending on Twitter and basically reply to it with one of our new vibe checks, for example.

386
00:33:29.036 --> 00:33:30.417
The main goal for this tool was

387
00:33:31.459 --> 00:33:34.762
I want to take the work to a higher level.

388
00:33:34.782 --> 00:33:45.230
I want to work on the level of goals and a plan, but not focus too much on the individual coding sessions itself.

389
00:33:49.462 --> 00:33:51.783
We have an orchestrator agent here.

390
00:33:51.983 --> 00:33:53.504
That's like my main interface.

391
00:33:53.524 --> 00:34:01.068
So I can talk to that agent, discuss what we're trying to build, tell it like the requirements, what tech stack I wanna use.

392
00:34:02.949 --> 00:34:09.292
And it can take all of that and break it down into the individual tasks of a plan basically.

393
00:34:10.513 --> 00:34:14.239
And this plan has separate steps.

394
00:34:14.439 --> 00:34:17.585
It shows the dependencies, what needs to happen in which order.

395
00:34:18.566 --> 00:34:23.014
And it has some autonomy to kick off work in the background.

396
00:34:24.060 --> 00:34:32.165
Um, so it's like implementing it and whenever it needs me, it can ping me for like a decision, um, where it needs my input.

397
00:34:32.905 --> 00:34:37.728
It can show me like, um, drafts or mock-ups of some UIs where I can give my input on.

398
00:34:38.688 --> 00:34:42.490
And I just give it my feedback and it continues working basically by itself.

399
00:34:43.351 --> 00:34:43.871
Super cool.

400
00:34:44.712 --> 00:34:46.173
Um, thanks for sharing.

401
00:34:46.553 --> 00:34:48.394
Uh, what do you think your next experiment might be?

402
00:34:49.474 --> 00:34:53.337
For this week, I'm probably going to focus on this orchestration idea.

403
00:34:53.957 --> 00:34:57.360
There are some questions I want to explore.

404
00:34:58.421 --> 00:35:01.883
For example, the idea is that this could be multiplayer as well.

405
00:35:02.564 --> 00:35:10.470
So you could invite other people from your team, or potentially even the whole company, and have a shared context of what you're working on.

406
00:35:11.370 --> 00:35:12.971
And everybody can bring their strength.

407
00:35:13.131 --> 00:35:19.774
So like the pirate and architect ideas, like the pirate can bring his ideas and what he wants to experiment on.

408
00:35:20.455 --> 00:35:24.897
And the architect can bring like the technical knowledge, but like at the same time.

409
00:35:25.417 --> 00:35:36.342
So we don't have to like hand off the work after one is done, but we can, yeah, like build the same project at the same time with a shared context and everybody brings in their expertise.

410
00:35:37.263 --> 00:35:37.503
Great.

411
00:35:37.823 --> 00:35:38.784
Thank you so much, Janik.

412
00:35:38.924 --> 00:35:39.584
Really good to see you.

413
00:35:42.092 --> 00:35:43.814
So welcome, Kieran.

414
00:35:44.054 --> 00:35:48.419
So excited to be talking about the work that you're doing as part of the Frontier team.

415
00:35:48.599 --> 00:35:50.502
What are you doing on the Frontier team?

416
00:35:51.765 --> 00:35:53.066
Yeah, it's really cool.

417
00:35:53.266 --> 00:36:01.530
I love the Frontier team because I basically, when I started with every, I started building Quora and experimenting with stuff.

418
00:36:01.690 --> 00:36:05.211
And out of that came compound engineering and so many more things.

419
00:36:05.432 --> 00:36:12.235
And I just felt this team was already there, but we never gave it a name.

420
00:36:12.635 --> 00:36:15.296
And now saying, hey, we're the Frontier team.

421
00:36:15.657 --> 00:36:17.517
We're allowed to do experiments to

422
00:36:19.619 --> 00:36:22.242
figure out what might work and what might not work.

423
00:36:22.622 --> 00:36:23.943
Yeah, it's great.

424
00:36:24.124 --> 00:36:29.589
So what I'm doing now is I'm just trying out

425
00:36:30.935 --> 00:36:33.396
what we can build with AI today.

426
00:36:33.616 --> 00:36:36.056
And that changes every time a new model comes out.

427
00:36:36.236 --> 00:36:43.398
I'm running an experiment every week, at least one, most of the time, multiple, and I learned something.

428
00:36:43.598 --> 00:36:58.281
And the big part is then to share what I learned with the team, with every subscribers, to make sure they can adopt or they can direct whatever they do in the direction that I learned from.

429
00:36:58.341 --> 00:36:59.562
So that's the biggest part.

430
00:37:01.583 --> 00:37:24.379
to do bigger things to be more ambitious and it's a little bit less scary if someone already did it and said yeah it's actually working this is a good pattern or someone says this is not a good pattern you should go this other direction so that's really my job to yeah go wide but also go deep it's both

431
00:37:25.188 --> 00:37:49.763
And I'd be curious, Kieran, when you started playing with the models, kind of like really testing what was possible, could you give us maybe just a, could we back up and could you share an example of how the capabilities have evolved in terms of what you were testing, say GP3-4, GPT-5-4, Opus 484, versus what you are now doing with the models that you have access to?

432
00:37:50.960 --> 00:37:52.881
Yeah, it's every time something new.

433
00:37:53.441 --> 00:37:56.702
And the whole point is to see it fail.

434
00:37:56.862 --> 00:38:00.663
It's like, how far can you push it until it fails?

435
00:38:00.683 --> 00:38:03.724
A lot of models are very good, and they do a lot of things.

436
00:38:03.764 --> 00:38:09.246
So the hard part is, how do we test even whether it's good or not?

437
00:38:09.446 --> 00:38:11.027
And how far can it go?

438
00:38:11.867 --> 00:38:14.728
And in the beginning, it was like, can it write code?

439
00:38:14.848 --> 00:38:15.888
Wow, it can write code.

440
00:38:15.928 --> 00:38:18.769
GPT-4 suddenly could write pretty good code.

441
00:38:18.889 --> 00:38:20.250
And that was the first unlock.

442
00:38:21.030 --> 00:38:33.959
And then we had Sonnet 3.5 that could write code across multiple files pretty well, like multiple pieces of code together and UI, like you could create UI with it.

443
00:38:37.150 --> 00:38:42.634
we went a little bit further and started to see agentic work.

444
00:38:42.674 --> 00:38:46.357
Like models could pass work along.

445
00:38:46.457 --> 00:38:51.361
Like before it was just like a prompt in and then something came out and then you iterate on it.

446
00:38:51.461 --> 00:39:00.568
And the agentic work with 3.7, Opus 3.7 with 4.4.1, like suddenly the agents could do work.

447
00:39:01.689 --> 00:39:04.651
And with Opus 4.5, it shifted again to...

448
00:39:05.952 --> 00:39:08.213
doing more complicated work.

449
00:39:08.513 --> 00:39:24.340
Like we could suddenly create like text editors with CRDT and like real time editing and all of that, like which is very complicated, but suddenly an AI could just do that in a week.

450
00:39:24.560 --> 00:39:26.721
And you saw Dan make proof.

451
00:39:26.961 --> 00:39:29.362
This was like the proof era, like GPT-5.5.

452
00:39:31.323 --> 00:39:34.064
So suddenly we're like, oh, we can be more ambitious.

453
00:39:34.164 --> 00:39:38.685
We can create bigger things without hand-holding too much.

454
00:39:39.806 --> 00:39:50.049
And especially with Fable now, with Fable 5.1, is that you have a model that can run even longer, even more deep.

455
00:39:50.969 --> 00:39:52.649
You can trust even more.

456
00:39:52.849 --> 00:39:54.310
Like, I'm a musician, so I...

457
00:39:55.591 --> 00:39:57.672
I always try to build a DAW.

458
00:39:57.712 --> 00:39:59.414
So I started maybe a year ago.

459
00:39:59.474 --> 00:40:04.076
And every time a new model comes out, I just try it again and say, hey, can we work more on it?

460
00:40:04.637 --> 00:40:08.279
And we're now at a point where I can just do that, which is insane.

461
00:40:08.619 --> 00:40:13.222
So maybe I should do a video editor because that's what we need.

462
00:40:14.003 --> 00:40:17.165
Could you tell us a few examples of things that you've built recently?

463
00:40:17.185 --> 00:40:20.427
So you are doing a lot of video stuff, for example.

464
00:40:20.467 --> 00:40:22.328
Yeah, my favorite is the video stuff.

465
00:40:22.348 --> 00:40:23.109
So I...

466
00:40:23.979 --> 00:40:26.241
do a video editor, but it's not just a video editor.

467
00:40:26.261 --> 00:40:40.171
It's also like a live streaming, like meeting recorder, like the whole thing that we have so many people at Avery that say and have such interesting ideas, but like they get lost in meetings.

468
00:40:40.832 --> 00:40:50.179
So this is also a way to record all of that, but also give that into a format that editors know, like, but also agent native, obviously.

469
00:40:50.219 --> 00:40:51.379
So everything agent native,

470
00:40:52.260 --> 00:40:57.187
And the other is an editor for writing.

471
00:40:57.388 --> 00:41:02.716
So this is also similar to like Proof, but we're like, I'm just experimenting.

472
00:41:02.736 --> 00:41:04.638
Can we make the next version of Proof?

473
00:41:04.759 --> 00:41:05.540
Can we make it better?

474
00:41:05.580 --> 00:41:06.842
Can we do even more?

475
00:41:07.848 --> 00:41:13.431
For anyone else who is on a frontier team, maybe at their company, right?

476
00:41:13.471 --> 00:41:14.351
Maybe it's a startup.

477
00:41:14.371 --> 00:41:16.332
Maybe it's a large enterprise team.

478
00:41:17.053 --> 00:41:26.618
What is one tip that you would have for them about how to succeed as someone who's operating at the frontier embedded in a larger organization?

479
00:41:27.524 --> 00:41:33.712
Yeah, I think the most important thing of a Frontier team is that you inspire your team, but also you listen to your team.

480
00:41:34.293 --> 00:41:41.363
If you do a hundred experiments and don't let it land and don't integrate it, it doesn't work.

481
00:41:42.064 --> 00:41:43.786
You need to listen to your team.

482
00:41:44.607 --> 00:41:50.812
get inspired by your team, and then do stuff, and then make sure that some things will be adopted.

483
00:41:50.852 --> 00:41:52.794
So it's really about integration.

484
00:41:53.194 --> 00:42:02.642
Yes, you should be free and have the freedom to do everything, but in the end, it is about integration and actually bringing it back to the actual organization.

485
00:42:02.682 --> 00:42:11.590
So I would urge everyone, make sure people get what you do and don't get overwhelmed, because it's very easy to overwhelm the entire organization by just...

486
00:42:11.990 --> 00:42:15.114
dumping 100 experiments in a channel.

487
00:42:15.515 --> 00:42:16.716
Make sure some stick.

488
00:42:16.977 --> 00:42:19.460
Make sure people understand what you do.

489
00:42:19.520 --> 00:42:21.202
That's your job as a frontier team.

490
00:42:22.023 --> 00:42:22.464
Amazing.

491
00:42:22.524 --> 00:42:23.205
Thanks so much, Kieran.

492
00:42:23.738 --> 00:42:26.561
If you liked this video, you should check out every.to.

493
00:42:26.681 --> 00:42:31.245
We have a two week free trial for anyone who's watching or listening at the link in the description.

494
00:42:31.986 --> 00:42:34.749
Every is the only subscription you need to stay at the edge of AI.

495
00:42:34.829 --> 00:42:41.736
It is the one subscription that if you're super AI-pilled and you wanna learn how to use it to do the best work of your life, it's got everything you need.

496
00:42:41.756 --> 00:42:44.519
It has education and equipment on the education side,

497
00:42:44.999 --> 00:42:48.022
We review every new model that drops and tell you what's good.

498
00:42:48.442 --> 00:42:56.630
On the equipment side, we have software tools specifically made for you, like an agent that helps you AI pill your whole company, all bundled together under one subscription.

499
00:42:57.130 --> 00:42:59.493
So get two weeks free now at the link in the description.