1
00:00:00,000 --> 00:00:01,520
Welcome back to the Deep Dive.

2
00:00:01,520 --> 00:00:04,280
This is where we take the sources, cut through all the noise,

3
00:00:04,280 --> 00:00:06,360
and really pull out the key insights.

4
00:00:06,360 --> 00:00:08,720
You know, act as your shortcut to being well informed.

5
00:00:08,720 --> 00:00:10,920
And today, oh, we've got a fascinating one.

6
00:00:10,920 --> 00:00:12,220
We're diving into Suna.

7
00:00:12,220 --> 00:00:15,480
It's described as an open source generalist AI agent.

8
00:00:15,480 --> 00:00:17,760
Sounds complex, maybe, but we think

9
00:00:17,760 --> 00:00:20,120
it's going to be a huge part of our digital future.

10
00:00:20,120 --> 00:00:20,760
But hang on.

11
00:00:20,760 --> 00:00:22,520
Before we really unpack Suna, we need

12
00:00:22,520 --> 00:00:25,120
to give a big thank you to our supporter, Safe Server.

13
00:00:25,120 --> 00:00:27,360
Safe Server takes care of hosting this very software,

14
00:00:27,360 --> 00:00:30,120
and they're there to support you in your digital transformation.

15
00:00:30,120 --> 00:00:34,460
You can find out more at www.safeserver.de.

16
00:00:34,460 --> 00:00:36,080
OK, so let's set the scene a bit.

17
00:00:36,080 --> 00:00:37,740
We're all drowning in information, right?

18
00:00:37,740 --> 00:00:40,000
Too many tabs, tasks, piling up.

19
00:00:40,000 --> 00:00:40,800
It's a lot.

20
00:00:40,800 --> 00:00:42,500
What if you had a kind of digital partner, something

21
00:00:42,500 --> 00:00:44,960
that can actually do things for you, not just talk?

22
00:00:44,960 --> 00:00:46,120
That's the promise here.

23
00:00:46,120 --> 00:00:48,840
But maybe first, what exactly is an AI agent?

24
00:00:48,840 --> 00:00:49,480
Let's start there.

25
00:00:49,480 --> 00:00:50,200
Keep it simple.

26
00:00:50,200 --> 00:00:51,560
Why is Suna making waves?

27
00:00:51,560 --> 00:00:53,480
Right, it's a really important distinction.

28
00:00:53,480 --> 00:00:56,880
Think beyond just chat bots that answer questions.

29
00:00:56,880 --> 00:01:01,100
An AI agent is, well, it's an AI that can actually act.

30
00:01:01,100 --> 00:01:02,080
It takes initiative.

31
00:01:02,080 --> 00:01:05,160
You give it a goal, say, plan my vacation or research

32
00:01:05,160 --> 00:01:06,320
competitors.

33
00:01:06,320 --> 00:01:09,120
And it figures out the steps, uses the tools it needs,

34
00:01:09,120 --> 00:01:10,640
and gets it done.

35
00:01:10,640 --> 00:01:11,960
It's autonomous.

36
00:01:11,960 --> 00:01:13,200
Autonomous, it acts.

37
00:01:13,200 --> 00:01:13,880
I get that.

38
00:01:13,880 --> 00:01:15,720
And Suna adds that generalist layer.

39
00:01:15,720 --> 00:01:17,560
It's not just built for one single thing.

40
00:01:17,560 --> 00:01:19,980
It's designed to handle a whole range of tasks.

41
00:01:19,980 --> 00:01:21,060
That's why it's different.

42
00:01:21,060 --> 00:01:22,000
Right, generalist.

43
00:01:22,000 --> 00:01:25,280
So it's not just like a travel planning bot or a research bot.

44
00:01:25,280 --> 00:01:27,760
It can potentially do lots of things.

45
00:01:27,760 --> 00:01:30,600
How does that acting on your behalf work in practice?

46
00:01:30,600 --> 00:01:31,180
Exactly.

47
00:01:31,180 --> 00:01:34,760
That's the core idea behind Suna being this fully open source

48
00:01:34,760 --> 00:01:36,040
AI assistant.

49
00:01:36,040 --> 00:01:37,800
Generalist means it's adaptable.

50
00:01:37,800 --> 00:01:40,080
It could be doing market analysis one minute,

51
00:01:40,080 --> 00:01:43,160
then helping you draft an email, or manage some files the next,

52
00:01:43,160 --> 00:01:45,400
all within the same conversation.

53
00:01:45,400 --> 00:01:47,960
It acts on your behalf because you talk to it naturally.

54
00:01:47,960 --> 00:01:50,220
You say, hey, find me profiles on LinkedIn,

55
00:01:50,220 --> 00:01:53,640
or summarize these articles, and it understands the intent.

56
00:01:53,640 --> 00:01:57,560
Then it figures out the how, which tools to use,

57
00:01:57,560 --> 00:02:00,120
what steps to take, and just delivers a result.

58
00:02:00,120 --> 00:02:02,160
It's meant to be intuitive, conversational.

59
00:02:02,160 --> 00:02:03,760
You just tell it what you need.

60
00:02:03,760 --> 00:02:06,560
And the open source part, why is that important here?

61
00:02:06,560 --> 00:02:08,440
That's a huge piece, especially when

62
00:02:08,440 --> 00:02:10,740
you're talking about an AI acting for you.

63
00:02:10,740 --> 00:02:13,640
Open source means the code, the instructions Suna follows.

64
00:02:13,640 --> 00:02:14,720
It's all public.

65
00:02:14,720 --> 00:02:16,800
Anyone can look under the hood, see how it works,

66
00:02:16,800 --> 00:02:19,760
check for issues, even contribute to making it better.

67
00:02:19,760 --> 00:02:22,480
It's built on transparency and collaboration.

68
00:02:22,480 --> 00:02:25,020
Plus, it uses the Apache 2.0 license,

69
00:02:25,020 --> 00:02:27,200
which means it's free to use and modify.

70
00:02:27,200 --> 00:02:29,920
That really builds trust and speeds up innovation.

71
00:02:29,920 --> 00:02:31,640
Transparency, yeah, that makes sense when

72
00:02:31,640 --> 00:02:33,280
it's doing things autonomously.

73
00:02:33,280 --> 00:02:33,780
Right.

74
00:02:33,780 --> 00:02:35,880
So OK, how does it actually do these things,

75
00:02:35,880 --> 00:02:40,080
going from, hey, Suna, do this, to actually executing tasks?

76
00:02:40,080 --> 00:02:41,420
What kind of tools does it have?

77
00:02:41,420 --> 00:02:44,240
It's got quite a powerful toolkit, really.

78
00:02:44,240 --> 00:02:47,120
Think of it like a digital Swiss army knife combined

79
00:02:47,120 --> 00:02:48,840
with a really good researcher.

80
00:02:48,840 --> 00:02:51,680
First off, it has really strong browser automation and search

81
00:02:51,680 --> 00:02:52,520
skills.

82
00:02:52,520 --> 00:02:53,840
It doesn't just search Google.

83
00:02:53,840 --> 00:02:55,920
It can actually use websites.

84
00:02:55,920 --> 00:02:59,720
It can navigate complex sites, log in if needed securely,

85
00:02:59,720 --> 00:03:03,240
extract specific data points, crawl through multiple pages.

86
00:03:03,240 --> 00:03:06,640
It uses tools like Playwright, Taveli, Firecrawl

87
00:03:06,640 --> 00:03:09,280
to basically become a super-powered web user.

88
00:03:09,280 --> 00:03:10,520
OK, so it's not just searching.

89
00:03:10,520 --> 00:03:11,960
It's interacting with the web.

90
00:03:11,960 --> 00:03:13,040
Exactly.

91
00:03:13,040 --> 00:03:14,600
Then it has file management.

92
00:03:14,600 --> 00:03:18,720
It can create Word docs, Excel sheets, PDFs, edit them,

93
00:03:18,720 --> 00:03:20,640
organize them within its secure space.

94
00:03:20,640 --> 00:03:22,160
It can take info from the web and put it

95
00:03:22,160 --> 00:03:23,640
straight into a file for you.

96
00:03:23,640 --> 00:03:26,800
Yeah, and it can perform certain system tasks, too,

97
00:03:26,800 --> 00:03:30,720
basic command line operations, executed safely, of course.

98
00:03:30,720 --> 00:03:32,680
Plus, it has integrations.

99
00:03:32,680 --> 00:03:34,980
It can connect to other software through APIs,

100
00:03:34,980 --> 00:03:37,760
pulling data from one place, sending it to another,

101
00:03:37,760 --> 00:03:39,760
triggering actions in different apps.

102
00:03:39,760 --> 00:03:42,060
And the key thing is, all these tools

103
00:03:42,060 --> 00:03:44,240
work together harmoniously.

104
00:03:44,240 --> 00:03:47,360
Suna figures out which ones to use, in what order,

105
00:03:47,360 --> 00:03:49,940
to solve your problem based on that simple conversation

106
00:03:49,940 --> 00:03:50,600
you had.

107
00:03:50,600 --> 00:03:53,000
It orchestrates the whole process.

108
00:03:53,000 --> 00:03:56,040
That orchestration sounds like the magic ingredient.

109
00:03:56,040 --> 00:03:57,400
But let's make it concrete.

110
00:03:57,400 --> 00:03:59,360
For someone listening, how could Suna actually

111
00:03:59,360 --> 00:04:01,120
change their day-to-day?

112
00:04:01,120 --> 00:04:02,680
Give us some real-world examples.

113
00:04:02,680 --> 00:04:04,480
OK, yeah, let's bring it to life.

114
00:04:04,480 --> 00:04:06,300
Think about business intelligence.

115
00:04:06,300 --> 00:04:08,680
Maybe you need to understand a new market takes ages,

116
00:04:08,680 --> 00:04:09,920
normally, right?

117
00:04:09,920 --> 00:04:11,720
You could tell Suna, analyze the market

118
00:04:11,720 --> 00:04:13,680
for my next company in the health care industry,

119
00:04:13,680 --> 00:04:15,200
located in the UK.

120
00:04:15,200 --> 00:04:17,880
Give me the major players, their market size, strengths,

121
00:04:17,880 --> 00:04:19,800
weaknesses, and their websites.

122
00:04:19,800 --> 00:04:21,600
Then generate a PDF report.

123
00:04:21,600 --> 00:04:23,640
Wow, so it does the digging and the reporting.

124
00:04:23,640 --> 00:04:24,160
Exactly.

125
00:04:24,160 --> 00:04:26,240
It goes out, finds the info, synthesizes it,

126
00:04:26,240 --> 00:04:27,640
and boom, PDF report.

127
00:04:27,640 --> 00:04:29,480
Saves potentially days of work.

128
00:04:29,480 --> 00:04:31,120
OK, I can see the value there.

129
00:04:31,120 --> 00:04:31,920
What else?

130
00:04:31,920 --> 00:04:34,560
How about recruitment or HR?

131
00:04:34,560 --> 00:04:36,800
Finding candidates is often a grind.

132
00:04:36,800 --> 00:04:39,560
You could prompt Suna, go on LinkedIn,

133
00:04:39,560 --> 00:04:43,240
find me 10 profiles available, people not working right now,

134
00:04:43,240 --> 00:04:45,720
for a junior software engineer in Munich.

135
00:04:45,720 --> 00:04:49,080
They need a relevant bachelor's degree and at least one year

136
00:04:49,080 --> 00:04:50,160
of experience.

137
00:04:50,160 --> 00:04:52,640
And it filters through LinkedIn based on that.

138
00:04:52,640 --> 00:04:55,800
Yep, it navigates LinkedIn, applies those filters,

139
00:04:55,800 --> 00:04:58,080
and gives you a list of potential candidates matching

140
00:04:58,080 --> 00:04:59,800
your specific criteria.

141
00:04:59,800 --> 00:05:01,040
Huge time saver.

142
00:05:01,040 --> 00:05:01,680
Definitely.

143
00:05:01,680 --> 00:05:03,320
OK, maybe something more personal.

144
00:05:03,320 --> 00:05:05,160
Sure, personal trip planning.

145
00:05:05,160 --> 00:05:06,920
We all know how tedious that can be.

146
00:05:06,920 --> 00:05:08,840
Juggling flights, hotels, activities.

147
00:05:08,840 --> 00:05:11,480
Imagine saying, generate a personal trip to London,

148
00:05:11,480 --> 00:05:13,920
leaving Bangkok May 1 for 10 days.

149
00:05:13,920 --> 00:05:17,760
Find central accommodation with Google reviews over 4.5 stars.

150
00:05:17,760 --> 00:05:19,960
Suggest interesting outdoor activities.

151
00:05:19,960 --> 00:05:21,800
Create a detailed itinerary.

152
00:05:21,800 --> 00:05:23,400
So it becomes your personal travel agent.

153
00:05:23,400 --> 00:05:24,160
Pretty much.

154
00:05:24,160 --> 00:05:27,360
It researches options, checks criteria like review scores,

155
00:05:27,360 --> 00:05:29,040
and puts together a whole plan for you.

156
00:05:29,040 --> 00:05:30,120
That's impressive.

157
00:05:30,120 --> 00:05:32,000
One more, maybe data related.

158
00:05:32,000 --> 00:05:34,720
Yeah, data analysis, or even just data gathering.

159
00:05:34,720 --> 00:05:36,840
Let's say you need public info compiled.

160
00:05:36,840 --> 00:05:39,320
My company needs an Excel sheet with info

161
00:05:39,320 --> 00:05:40,880
on Italian lottery games.

162
00:05:40,880 --> 00:05:44,360
Generate a spreadsheet with all the basic public information.

163
00:05:44,360 --> 00:05:47,920
Zuna can go find that public data, structure it,

164
00:05:47,920 --> 00:05:51,000
and deliver it to you in a ready to use spreadsheet.

165
00:05:51,000 --> 00:05:52,600
These are really practical examples.

166
00:05:52,600 --> 00:05:53,760
It's not just theory.

167
00:05:53,760 --> 00:05:58,120
It's automating real, often tedious tasks.

168
00:05:58,120 --> 00:06:01,880
So this agent doing all this work,

169
00:06:01,880 --> 00:06:04,120
it must have some kind of structure behind it, right?

170
00:06:04,120 --> 00:06:05,840
How is it built to handle all this?

171
00:06:05,840 --> 00:06:08,800
And what does 100% open source mean for its architecture?

172
00:06:08,800 --> 00:06:11,280
Absolutely, it's built on a solid modular foundation.

173
00:06:11,280 --> 00:06:13,760
There are basically four main parts working together.

174
00:06:13,760 --> 00:06:16,400
First, the back end API, that's like the central nervous

175
00:06:16,400 --> 00:06:17,480
system, the brain.

176
00:06:17,480 --> 00:06:20,200
It handles your requests, manages the different tasks

177
00:06:20,200 --> 00:06:22,600
Suna might be working on simultaneously.

178
00:06:22,600 --> 00:06:25,360
And crucially, it integrates with the big large language

179
00:06:25,360 --> 00:06:28,280
models, like Anthropix models, that give Suna its understanding

180
00:06:28,280 --> 00:06:29,160
and reasoning power.

181
00:06:29,160 --> 00:06:30,480
OK, the brain, what else?

182
00:06:30,480 --> 00:06:31,480
Then you've got the front end.

183
00:06:31,480 --> 00:06:33,360
That's simply what you interact with the chat interface,

184
00:06:33,360 --> 00:06:35,280
the dashboard where you see tasks progressing.

185
00:06:35,280 --> 00:06:36,680
It's the face of Suna.

186
00:06:36,680 --> 00:06:37,400
Makes sense.

187
00:06:37,400 --> 00:06:41,560
Then, really important, is the agent docker or environment.

188
00:06:41,560 --> 00:06:44,720
Think of this as a secure sandbox.

189
00:06:44,720 --> 00:06:47,960
When Suna needs to browse the web or run some code,

190
00:06:47,960 --> 00:06:50,360
it happens inside this isolated container.

191
00:06:50,360 --> 00:06:52,400
It keeps everything safe and controlled.

192
00:06:52,400 --> 00:06:53,520
Right, security is key.

193
00:06:53,520 --> 00:06:54,640
Definitely.

194
00:06:54,640 --> 00:06:57,080
And finally, there's the SuperBase database.

195
00:06:57,080 --> 00:06:58,120
That's Suna's memory.

196
00:06:58,120 --> 00:07:01,360
It stores your conversations, the files Suna generates,

197
00:07:01,360 --> 00:07:04,400
the current status of all the tasks, user data,

198
00:07:04,400 --> 00:07:07,680
everything needed for continuity and record keeping.

199
00:07:07,680 --> 00:07:09,720
So those four parts work together.

200
00:07:09,720 --> 00:07:12,580
And the open source aspect means you can potentially

201
00:07:12,580 --> 00:07:14,920
run this whole setup yourself, that self-hosting thing.

202
00:07:14,920 --> 00:07:17,840
Exactly, because all those components are open source,

203
00:07:17,840 --> 00:07:20,440
you aren't locked into using their hosted version.

204
00:07:20,440 --> 00:07:21,880
You can actually download the code

205
00:07:21,880 --> 00:07:24,000
and set up Suna entirely on your own servers,

206
00:07:24,000 --> 00:07:25,640
your own cloud infrastructure.

207
00:07:25,640 --> 00:07:26,600
What does that involve?

208
00:07:26,600 --> 00:07:27,600
Is it complicated?

209
00:07:27,600 --> 00:07:29,680
Well, it's a comprehensive process, yeah.

210
00:07:29,680 --> 00:07:31,720
There's a setup wizard to help, but you'd

211
00:07:31,720 --> 00:07:35,120
be configuring the database, the secure agent environment,

212
00:07:35,120 --> 00:07:37,520
connecting it to your chosen LLM provider,

213
00:07:37,520 --> 00:07:39,200
setting up the web tools.

214
00:07:39,200 --> 00:07:41,920
It gives you ultimate control over data privacy, security,

215
00:07:41,920 --> 00:07:43,440
and customization.

216
00:07:43,440 --> 00:07:45,640
It's powerful for businesses or individuals

217
00:07:45,640 --> 00:07:48,320
who really want to tailor it or keep everything in-house.

218
00:07:48,320 --> 00:07:51,360
OK, that's a great option for more technical users.

219
00:07:51,360 --> 00:07:54,080
So bringing it back to our listener,

220
00:07:54,080 --> 00:07:58,080
after hearing all this, what's the bottom line?

221
00:07:58,080 --> 00:08:00,920
How does Suna actually help you deal with that information

222
00:08:00,920 --> 00:08:04,720
flood or those everyday tasks that just eat up time?

223
00:08:04,720 --> 00:08:08,600
I think it comes down to efficiency and focus.

224
00:08:08,600 --> 00:08:10,720
Suna acts as that shortcut we talked about.

225
00:08:10,720 --> 00:08:13,600
Instead of you spending hours researching or compiling data,

226
00:08:13,600 --> 00:08:14,400
Suna does it.

227
00:08:14,400 --> 00:08:15,520
It automates the lead work.

228
00:08:15,520 --> 00:08:17,660
So it helps you get informed faster, yes.

229
00:08:17,660 --> 00:08:19,600
But it also delivers those aha moments

230
00:08:19,600 --> 00:08:23,120
because it presents structured results from complex requests.

231
00:08:23,120 --> 00:08:25,320
You get the insights without drowning in the process.

232
00:08:25,320 --> 00:08:26,560
Right, it cuts them to clutter.

233
00:08:26,560 --> 00:08:27,560
Exactly.

234
00:08:27,560 --> 00:08:29,600
It helps manage that info overload

235
00:08:29,600 --> 00:08:31,760
by distilling things down and even acting

236
00:08:31,760 --> 00:08:32,940
on the information.

237
00:08:32,940 --> 00:08:34,820
It finds the signal and the noise.

238
00:08:34,820 --> 00:08:37,000
And importantly, it's designed to be accessible.

239
00:08:37,000 --> 00:08:38,120
They have different plans.

240
00:08:38,120 --> 00:08:40,580
But the key one to start with is the free plan.

241
00:08:40,580 --> 00:08:41,080
Free.

242
00:08:41,080 --> 00:08:42,240
What do you get with that?

243
00:08:42,240 --> 00:08:45,520
You get 60 minutes of agent runtime per month,

244
00:08:45,520 --> 00:08:47,800
which is enough to try out quite a few tasks.

245
00:08:47,800 --> 00:08:49,480
You can work on public projects.

246
00:08:49,480 --> 00:08:52,040
And it uses a solid foundational AI model.

247
00:08:52,040 --> 00:08:54,640
It's a perfect way to just dip your toes in

248
00:08:54,640 --> 00:08:56,960
and see how it feels to have an agent working for you.

249
00:08:56,960 --> 00:08:58,640
60 free minutes a month.

250
00:08:58,640 --> 00:09:00,880
That's definitely a good starting point to experiment.

251
00:09:00,880 --> 00:09:01,960
For sure.

252
00:09:01,960 --> 00:09:03,600
And then if you need more, there are

253
00:09:03,600 --> 00:09:06,860
pro and custom plans with more features, more runtime,

254
00:09:06,860 --> 00:09:08,860
and support for private projects.

255
00:09:08,860 --> 00:09:11,800
That free tier sounds like the way to go for anyone curious.

256
00:09:11,800 --> 00:09:14,240
It really lets you experience it firsthand.

257
00:09:14,240 --> 00:09:18,280
So thinking ahead, as these AI agents get even smarter, more

258
00:09:18,280 --> 00:09:21,040
capable, how might a digital companion like Suna

259
00:09:21,040 --> 00:09:22,240
really change things?

260
00:09:22,240 --> 00:09:24,920
How could it fundamentally alter the way you listening right now

261
00:09:24,920 --> 00:09:28,280
work, or learn, or even just plan your week?

262
00:09:28,280 --> 00:09:30,280
It's kind of exciting and maybe a little daunting

263
00:09:30,280 --> 00:09:33,140
to think about reclaiming all that time and mental energy.

264
00:09:33,140 --> 00:09:34,920
Definitely something to mull over.

265
00:09:34,920 --> 00:09:37,000
We really encourage you to check out Suna, see if it clicks

266
00:09:37,000 --> 00:09:37,880
for your needs.

267
00:09:37,880 --> 00:09:39,640
And before we wrap up, one more big thank you

268
00:09:39,640 --> 00:09:41,280
to our supporter, SafeServer.

269
00:09:41,280 --> 00:09:43,520
Remember, they provide the hosting for the software

270
00:09:43,520 --> 00:09:46,480
and are fantastic partners for digital transformation.

271
00:09:46,480 --> 00:09:51,180
Find out more at www.safeserver.de.

272
00:09:51,180 --> 00:09:52,960
Thanks for joining us on this deep dive.

273
00:09:52,960 --> 00:09:54,880
We'll catch you next time for another exploration

274
00:09:54,880 --> 00:09:58,000
into the tech and ideas shaping our world.