Latest / AI Fire Daily / #675 Neil: 6 AI for Business Tools I Actually Use in My Daily Work

Transcript

#675 Neil: 6 AI for Business Tools I Actually Use in My Daily Work · AI Fire Daily

• 0:00
  I've tested over a hundred AI tools. Here's the
• 0:04
  funny part. Some are better at charging your
• 0:05
  credit card than doing actual work. Yeah. I mean,
• 0:08
  that is brutally honest and unfortunately it's
• 0:10
  incredibly accurate. Right. Too many platforms
• 0:12
  are optimized purely for billing. They aren't
• 0:15
  actually optimized for building. Exactly. So
• 0:18
  welcome to this deep dive. Today we are cutting
• 0:20
  through all that software bloat. Definitely.
• 0:23
  We're unpacking a builder's private notes today.
• 0:25
  They took over a hundred shiny AI tools and just
• 0:28
  aggressively whittled them down. down to just
• 0:31
  seven essentials. And these aren't toys. These
• 0:33
  tools automate massive daily business tasks.
• 0:36
  They help you build custom internal apps, and
• 0:39
  crucially, they slash those massive monthly subscription
• 0:42
  bills. Which is huge because everyone is quietly
• 0:45
  drowning in monthly software fees. It's like
• 0:48
  death by a thousand digital cuts. It really is.
• 0:51
  I'll be honest with you, I still wrestle with
• 0:53
  subscription fatigue myself. Managing a dozen
• 0:56
  $20 a month apps gets exhausting. It feels completely
• 0:59
  overwhelming sometimes. You are definitely not
• 1:02
  alone in that feeling. But what's fascinating
• 1:04
  here is the underlying strategy. It is all about
• 1:07
  strategic, ruthless consolidation. You can actually
• 1:11
  stop paying for single -use web apps. Think of
• 1:13
  it like stacking Lego blocks. You don't need
• 1:16
  a pre -built plastic castle. You just need the
• 1:19
  core foundational bricks. Once you have the right
• 1:21
  pieces, you build anything. I love that analogy.
• 1:25
  So let's look at the very first brick. Before
• 1:27
  you can automate a whole business, you need documents.
• 1:30
  You need local files, and you need logic. We
• 1:33
  are looking at Cloud Code and Codex. These are
• 1:36
  your primary builders. Exactly. Looking at these
• 1:39
  private notes, the strategy is super clear. They
• 1:42
  completely canceled several popular monthly subscriptions.
• 1:44
  Like what? Well, they dropped Gamma, which is
• 1:47
  used for making slides. They canceled Poppy,
• 1:50
  an AI tool for creating content. They even dropped
• 1:52
  Loom for screen recording. Wow. That is a massive
• 1:55
  amount of monthly savings. Right. They replaced
• 1:58
  all those fragmented web apps. They moved to
• 2:01
  local AI using Cloud Code and Codex. And we really
• 2:04
  need to explain why that matters. These aren't
• 2:07
  just websites you log into. These are essentially
• 2:09
  local AI assistants. They live directly on your
• 2:13
  actual machine. That is the critical difference
• 2:15
  right there. When an AI lives on your local machine,
• 2:18
  it has context. It can natively access your actual
• 2:22
  local files. It reads your project folders and
• 2:24
  your code base. Because if you use a standard
• 2:27
  web -based AI... You are constantly uploading
• 2:29
  files. Yeah, you have to copy and paste context
• 2:31
  back and forth. It is incredibly tedious. But
• 2:34
  Cloud Code actually indexes your workspace. It
• 2:37
  sits in your terminal and reads your environment.
• 2:39
  And you just communicate with it in plain English.
• 2:42
  You don't need complex programming languages
• 2:43
  at all. You just explain exactly what you need.
• 2:46
  It can read your specific business documents.
• 2:49
  It can write custom Python scripts for you. It
• 2:51
  can even create detailed spreadsheets instantly.
• 2:54
  The notes break down the actual daily usage,
• 2:57
  too. These two local tools handle 90 % of the
• 2:59
  workload. Cloud Code is organizing all their
• 3:02
  business knowledge. They keep raw meeting notes
• 3:04
  in specific local folders. And the AI automatically
• 3:07
  indexes those specific folders. Exactly. It uses
• 3:11
  that context for every future task. It essentially
• 3:13
  becomes an intern with a perfect memory. Cloud
• 3:16
  Code is also brilliant for structural brainstorming.
• 3:18
  You can talk through complex logical problems
• 3:21
  first. It helps you decide the architectural
• 3:23
  next steps. And then Codex fills a completely
• 3:27
  different gap. It is highly specialized for raw
• 3:30
  computer tasks. It handles automated browser
• 3:32
  interactions incredibly well. So you run these
• 3:35
  two engines side by side. But let me push back
• 3:38
  on this a bit. How is someone with zero coding
• 3:40
  experience supposed to start? Opening a developer
• 3:43
  terminal sounds really intimidating for beginners.
• 3:46
  It definitely sounds scary, but it's simpler
• 3:49
  than you think. You do not start by building
• 3:52
  a massive app. You just start with your simple
• 3:54
  daily routines. Okay, so like what? You set up
• 3:58
  a basic workspace folder on your desktop. You
• 4:00
  drop your company guidelines in there, then you
• 4:03
  simply ask the AI question. You ask it to summarize
• 4:06
  yesterday's meeting notes. Start small, like
• 4:09
  summarizing meeting notes before trying to build
• 4:11
  apps. Exactly. You review the output and fix
• 4:14
  any small errors. You save those good instructions
• 4:17
  for your next session. It builds your technical
• 4:20
  confidence gradually over time. Makes sense.
• 4:23
  So you have these brilliant foundational Lego
• 4:25
  bricks. You are building workflows and drafting
• 4:27
  scripts locally. But here's where the friction
• 4:30
  usually starts. Oh, for sure. Once you scale
• 4:33
  up, using multiple different models gets messy.
• 4:35
  You might want Claude for coding and OpenAI for
• 4:38
  writing. Managing all those separate accounts
• 4:41
  gets incredibly expensive, very fast. It completely
• 4:44
  breaks your workflow. You definitely don't want
• 4:46
  50 different billing logins. You need a centralized
• 4:49
  brain to manage the traffic. Right. That brings
• 4:52
  us to our second essential tool. It is called
• 4:54
  OpenRouter. It is essentially the ultimate AI
• 4:57
  cost manager. OpenRouter is a total game changer
• 5:00
  for builders. It provides instant access to over
• 5:03
  500 AI models. It aggregates more than 80 different
• 5:06
  AI providers. You get OpenAI, Anthropic, and
• 5:09
  Google all in one place. And they all run through
• 5:12
  one single API. Right, an API being a digital
• 5:15
  bridge letting two different apps talk to each
• 5:17
  other. That is the perfect way to visualize it.
• 5:20
  You are no longer managing a dozen API integrations.
• 5:23
  OpenRouter lets you switch active models instantly
• 5:26
  in your code. You don't have to rebuild your
• 5:28
  entire software setup. This single connection
• 5:30
  gives you massive control over operational costs.
• 5:33
  The actual routing strategy here is completely
• 5:35
  brilliant. You can dynamically route tasks based
• 5:38
  on complexity. Give me an example of that. Let's
• 5:41
  say you are processing incoming sales leads.
• 5:44
  Qualifying a basic lead does not require a genius
• 5:47
  model. You can send those easy tasks to much
• 5:50
  cheaper models. You might use something like
• 5:52
  Llama 3 for pennies. Right, because you saved
• 5:54
  the heavy lifters for later. You strictly use
• 5:57
  the stronger, expensive models for complex tasks.
• 6:00
  Exactly. Writing a highly personalized client
• 6:02
  email requires careful reasoning. You route that
• 6:05
  specific task to Claude 3 .5 Sonnet. You only
• 6:08
  pay for premium processing power when you actually
• 6:11
  need it. It's kind of like hiring different contractors
• 6:13
  for different jobs. You don't hire a master architect
• 6:16
  to paint a fence. Perfect analogy. OpenRouter
• 6:19
  also lets you configure automated backup models.
• 6:22
  Let's say your primary model suddenly goes down.
• 6:24
  It happens all the time with cloud providers.
• 6:27
  Oh, constantly. OpenRouter automatically tries
• 6:30
  the next model on your list. Your workflows keep
• 6:32
  running without requiring constant human intervention.
• 6:35
  Developers are integrating this with specialized
• 6:37
  agents like Hermes, right? They access every
• 6:40
  single AI brain through one connection. You can
• 6:43
  rapidly test different models for your specific
• 6:45
  tasks. You can compare response quality and costs
• 6:48
  instantly, and you do it without breaking your
• 6:50
  core integrations. Oh, wait. I have to ask about
• 6:53
  tracking all this. Tracking costs across 500
• 6:56
  different models sounds completely chaotic. I
• 6:59
  can barely manage my own AWS billing dashboard.
• 7:02
  Doesn't this become an absolute accounting nightmare?
• 7:05
  It absolutely would be if you used one master
• 7:08
  key. But the smart strategy here is compartmentalization.
• 7:12
  You just generate separate dedicated API keys.
• 7:15
  Ah, I see. You create a unique key for every
• 7:18
  single internal workflow. That completely isolates
• 7:21
  the billing for that specific task. If your marketing
• 7:24
  automation suddenly spikes in cost, you know
• 7:26
  exactly why. Just create a separate API key for
• 7:29
  every specific workflow you build. Exactly. It
• 7:32
  keeps your overall budget completely transparent
• 7:34
  and controllable. Okay, so let's look at the
• 7:36
  Lego stack so far. You have your local models
• 7:39
  drafted and properly routed. Your cost -effective
• 7:42
  API logic is completely dialed in. But now we
• 7:45
  hit a massive physical limitation. The hardware
• 7:48
  issue. Right. What happens when you physically
• 7:50
  close your laptop? The second that screen goes
• 7:52
  black, your automations die. Which entirely defeats
• 7:55
  the core purpose of business automation. A true
• 7:58
  digital employee needs an engine that never sleeps.
• 8:01
  You obviously cannot leave your MacBook open
• 8:03
  on your desk forever. This crucial gap brings
• 8:06
  us to hosting our VPS. It acts as the always
• 8:09
  -on engine of this entire setup. Think of it
• 8:11
  as the baseplate for your Lego bricks. Basically,
• 8:15
  a VPS is a cloud computer that stays awake so
• 8:17
  yours doesn't have to. That is exactly the right
• 8:20
  mental model. Looking at the notes, this is where
• 8:22
  the real scale happens. They use this virtual
• 8:25
  private server for continuous background tasks.
• 8:28
  Like what, specifically? Specifically, they run
• 8:30
  an automation tool called N8N. They manage scheduled
• 8:34
  background tasks with autonomous AI agents. They
• 8:37
  are even generating and posting Instagram content
• 8:39
  automatically. All of this background processing
• 8:42
  happens while their laptop is shut. The specific
• 8:45
  N8n workflow example in the notes is really striking.
• 8:48
  Let's dig into how that actually works under
• 8:50
  the hood. Imagine your clients are emailing you
• 8:53
  business receipts daily. They send signed contracts
• 8:56
  and random PDF document attachments. Oh, downloading
• 9:00
  and organizing those files manually gets old
• 9:02
  incredibly fast. It is the ultimate soul -crushing
• 9:06
  administrative chore. But N8n on a VPS completely
• 9:09
  eliminates that friction. It acts like a digital
• 9:12
  nervous system for your business. It essentially
• 9:14
  listens to a webhook connected to your Gmail.
• 9:17
  The moment an email arrives with specific financial
• 9:19
  keywords, it triggers. It extracts the attached
• 9:22
  PDF instantly in the cloud. Then it authenticates
• 9:26
  with your Google Drive API automatically. It
• 9:29
  sorts the document into the correct client folder
• 9:31
  seamlessly. And it does this 24 hours a day.
• 9:34
  It never takes a coffee break or asks for vacation.
• 9:37
  It just runs silently in the background of your
• 9:39
  business. But let me ask the obvious, slightly
• 9:42
  skeptical question here. Dealing with cloud servers
• 9:45
  sounds highly technical and intimidating. Do
• 9:47
  you need a degree in server management to keep
• 9:50
  this safe? You really don't. Not with modern
• 9:53
  hosting solutions today. In the past, provisioning
• 9:56
  of VPS was a massive headache. But platforms
• 10:00
  like Hostinger offer ready -made installation
• 10:02
  templates now. So it's basically plug and play.
• 10:05
  Yeah, you do not have to build the server architecture
• 10:07
  from scratch. You literally just click a template,
• 10:10
  and install the automation apps. Not at all.
• 10:12
  Ready -made templates handle the complex setup
• 10:14
  for you. It dramatically lowers the technical
• 10:16
  barrier to entry for builders. OK, so let's recap
• 10:19
  where we are. Your server is awake continuously
• 10:22
  in the cloud. Your AI models are optimally routed
• 10:25
  and cost -effective. But what specific annoying
• 10:29
  chores should you actually offload? What exact
• 10:32
  jobs are we assigning to these always -on machines?
• 10:35
  That question introduces our fourth essential
• 10:37
  tool, Grokbot. The builder describes this tool
• 10:40
  perfectly in their private notes. It functions
• 10:42
  as a highly competent human reminder app. It
• 10:46
  handles all those repetitive daily tasks that
• 10:49
  drain your energy. These are the small, easily
• 10:52
  forgettable administrative jobs. They are deeply
• 10:54
  annoying, but they absolutely must get done.
• 10:56
  Exactly. Grokbot can autonomously access web
• 11:00
  apps and browser interfaces. It lives on your
• 11:03
  cloud computer and executes tasks perfectly.
• 11:06
  And the real power is the asynchronous management,
• 11:08
  right? Yeah, you can actually assign and monitor
• 11:11
  tasks right from your phone. You just check the
• 11:14
  execution results while you're on the go. You
• 11:16
  do not have to sit at your desk babysitting the
• 11:19
  bot. The notes give some fantastic real -world
• 11:22
  application examples. Think about running a live
• 11:24
  community Q &A session. Afterward, there is always
• 11:28
  tedious formatting work to do. Oh, absolutely.
• 11:31
  Grokbot can autonomously collect the raw session
• 11:34
  transcript. It cleans it up and posts it directly
• 11:36
  into the community platform. It saves a highly
• 11:39
  readable formatted copy for your members. It
• 11:42
  also handles tedious content distribution automatically.
• 11:45
  When a new YouTube video goes live, the bot activates.
• 11:48
  It automatically shares the video link to connected
• 11:50
  community channels. That's amazing! The builder
• 11:53
  even constructed a specialized daily news bot.
• 11:57
  It autonomously scrapes the web for important
• 11:59
  AI news updates. It summarizes the findings and
• 12:02
  sends those updates directly to them. It's kind
• 12:04
  of like having a dedicated research assistant.
• 12:07
  But I have to ask about the potential fail states.
• 12:09
  What happens when these automations inevitably
• 12:11
  break down? AI can definitely hallucinate and
• 12:14
  make weird, unpredictable mistakes. How do you
• 12:17
  troubleshoot if a bot goes completely off the
• 12:19
  rails? If we connect this back to software architecture,
• 12:23
  isolation is key. You never build one giant bot
• 12:26
  that handles everything. The builder specifically
• 12:28
  creates separate isolated bots for different
• 12:30
  jobs. So they don't overlap. Right. The transcript
• 12:33
  bot has absolutely nothing to do with the news
• 12:35
  bot. Each micro bot receives its own incredibly
• 12:38
  strict system instructions. If the news bot hallucinates,
• 12:42
  you know exactly which logic block failed. Give
• 12:45
  each specific job its own bot and distinct instructions.
• 12:48
  Exactly. It keeps troubleshooting incredibly
• 12:51
  simple and contained. So let's step back and
• 12:53
  look at the human element here. You are orchestrating
• 12:56
  all these different autonomous bots. You are
• 12:58
  managing complex workflows across always -on
• 13:01
  cloud servers, which ultimately means you are
• 13:05
  typing a massive amount of prompts. Yes, it requires
• 13:08
  a tremendous amount of manual human input. And
• 13:11
  that input phase is physically exhausting for
• 13:13
  your hands. You are typing emails and engineering
• 13:16
  complex prompts all day long. Often your brain
• 13:19
  knows exactly what logic it wants to build, but
• 13:21
  your fingers simply cannot type the code fast
• 13:24
  enough. The human body becomes a primary bottleneck
• 13:26
  in the workflow. So we urgently need to speed
• 13:29
  up the human input phase. This brings us to essential
• 13:32
  tool number five. It is called Glado. It is a
• 13:36
  highly specialized voice AI dictation tool. It
• 13:40
  instantly turns your spoken speech into structured
• 13:42
  text. And the brilliant part is how it integrates
• 13:45
  with your system. It works globally inside literally
• 13:49
  any app you are already using. You don't have
• 13:51
  to open a separate dictation window. You just
• 13:53
  hold down a designated shortcut key on your keyboard.
• 13:57
  You speak your complex thoughts out loud naturally.
• 14:00
  Then you release the key and Glado types it instantly.
• 14:04
  The builder relies on this for all their heavy
• 14:06
  typing tasks. They dictate massive complex AI
• 14:10
  prompts completely out loud. They reply to nuanced
• 14:13
  daily business emails this exact same way. They
• 14:16
  capture fleeting architectural ideas instantly
• 14:18
  using just their voice. They even draft their
• 14:21
  social media content verbally while pacing the
• 14:23
  room. What's truly fascinating here is the advanced
• 14:26
  coding application. The notes specifically mention
• 14:29
  a developer, Dave Ebelar, from Data Lumina. He
• 14:32
  literally uses voice instructions to write complex
• 14:35
  software code. He builds intricate applications
• 14:38
  much faster just by speaking the logic. Just
• 14:41
  imagine combining that capability with Claude
• 14:44
  code. You open your developer terminal workspace
• 14:46
  on your screen. You hold your Glado dictation
• 14:49
  shortcut key down firmly. You literally just
• 14:51
  explain the software feature you want to build.
• 14:54
  And Glado captures the nuance and turns it into
• 14:56
  a perfectly structured prompt. The AI translates
• 14:59
  your natural speech into executable logic. Then
• 15:03
  you just quickly review the text before hitting
• 15:05
  enter. It is a massive transformative time saver
• 15:08
  for heavy builders. But does voice dictation
• 15:11
  mess things up structurally? If I'm coding, I
• 15:14
  need my syntax to be absolutely perfect. Does
• 15:17
  it ruin the formatting of highly structured technical
• 15:19
  prompts? Sometimes you need very specific JSON
• 15:22
  layouts or markdown structures. That is exactly
• 15:25
  the beauty of this specific AI tool. You aren't
• 15:28
  just getting raw, unformatted transcribed text.
• 15:31
  You can completely customize the backend output
• 15:34
  formatting instructions. So you train it on your
• 15:37
  style? You explicitly tell Gladys exactly how
• 15:40
  to structure your specific outputs. It matches
• 15:42
  your strict programming or workflow needs perfectly
• 15:45
  every time. You can customize how Gladys formats
• 15:49
  text to match your exact needs. It keeps you
• 15:51
  in complete architectural control always. Alright,
• 15:55
  so text generation and code logic are now fully
• 15:57
  optimized, but modern business inherently requires
• 16:00
  compelling visual assets too. We need marketing
• 16:02
  images and we need promotional videos. Subscribing
• 16:06
  to every single new AI image generator adds up
• 16:09
  incredibly fast. Oh, it is another massive hidden
• 16:12
  drain on the startup wallet. Every visual platform
• 16:15
  has its own separate monthly billing cycle. They
• 16:18
  all feature different underlying models and wildly
• 16:20
  different pricing tiers. Managing mid -journey,
• 16:23
  runway, and others is a logistical nightmare.
• 16:26
  This operational headache leads us to our final
• 16:28
  essential tool. It is called kia .ai. It functions
• 16:33
  basically as a massive visual model aggregator.
• 16:36
  It gives you instant API access to over 100 visual
• 16:39
  models. It includes access to the absolute heavy
• 16:42
  hitters. You get high -end video generators like
• 16:45
  Veo and Kling. It even includes highly specialized
• 16:48
  niche models like Nano Banana. You access all
• 16:51
  of these powerful visual engines through one
• 16:54
  single platform. The cost control angle here
• 16:57
  is absolutely crucial for lean builders. Unlike
• 17:00
  standard subscriptions, it strictly charges you
• 17:03
  based on actual compute usage. You do not pay
• 17:06
  flat monthly subscription fees just to retain
• 17:09
  access. This completely changes how you approach
• 17:12
  creating visual marketing content. Because you
• 17:15
  pay per generation, you can do proper A -B testing.
• 17:19
  Let's say you desperately need a new YouTube
• 17:20
  thumbnail, right? Right. You generate one stylized
• 17:23
  version using the first model. Then you generate
• 17:25
  a totally different version using a second model.
• 17:29
  You can compare the visual results side by side
• 17:31
  instantly. You find the absolute best aesthetic
• 17:34
  fit for your specific brand. And crucially, You
• 17:37
  didn't pay for two separate monthly SaaS subscriptions.
• 17:40
  You just paid a fraction of a cent for the compute.
• 17:43
  The notes detail using this exact workflow for
• 17:46
  social ads and carousels. They search models
• 17:48
  by name or filter them by specific visual task.
• 17:52
  But they always check the specific pricing section
• 17:55
  before generating anything. That actually brings
• 17:58
  up a really important friction point. How do
• 18:00
  you avoid accidentally wasting your credits?
• 18:03
  Experimenting wildly with visual AI can get expensive
• 18:07
  incredibly quickly. It seems painfully easy to
• 18:10
  accidentally drain your account balance on bad
• 18:12
  prompts. This raises a really important question
• 18:15
  about personal workflow discipline. You must
• 18:18
  cultivate a habit of starting small and verifying
• 18:21
  first. The builder always tests single, isolated
• 18:24
  prompts on one specific model. They critically
• 18:27
  ensure the prompt structure works before generating
• 18:30
  large batches. Always check pricing sections
• 18:33
  and test single prompts before batch generating.
• 18:35
  Exactly. It protects your operational budget
• 18:37
  while you creatively experiment. And real quick,
• 18:40
  a word from our sponsor. All right, let's take
• 18:42
  a moment to synthesize this entire deep dive.
• 18:45
  We have covered a tremendous amount of technical
• 18:47
  ground today. Looking at these notes, the real
• 18:50
  power isn't in any single tool. The actual magic
• 18:54
  happens when you meticulously start stacking
• 18:56
  them together. It really is like stacking complex
• 18:59
  Lego blocks of data. Each specific tool handles
• 19:03
  a very narrow, defined business function. But
• 19:06
  when you lock them together, the synergy is incredible.
• 19:09
  You combine these basic bricks to build an entirely
• 19:11
  automated enterprise. Let's quickly walk through
• 19:14
  the ultimate automated workflow stack. You use
• 19:17
  local cloud code to build the actual application
• 19:20
  logic. OpenRouter elegantly manages those logic
• 19:23
  drafts across the most cost -effective models.
• 19:26
  You plug in key .ai to generate all the compelling
• 19:29
  visual assets dynamically. And finally, your
• 19:32
  Hostinger VPS runs this entire interconnected
• 19:35
  system on a constant schedule. The whole integrated
• 19:38
  system just hums along beautifully in the background.
• 19:41
  It is a highly optimized, completely autonomous
• 19:44
  digital machine. Whoa. Imagine an entire agency
• 19:48
  running effortlessly while your laptop is shut.
• 19:51
  It represents a profound structural shift in
• 19:55
  how we actually work. You are no longer just
• 19:58
  an exhausted employee doing administrative chores.
• 20:01
  You have fundamentally become the manager of
• 20:04
  an autonomous digital workforce. The core conclusion
• 20:07
  from these private builder notes is really sharp.
• 20:09
  You do not need a massive chaotic collection
• 20:11
  of expensive web apps. Two highly versatile tools
• 20:15
  you master and use every single day are incredibly
• 20:18
  valuable. They are functionally far better than
• 20:21
  10 expensive subscriptions just collecting digital
• 20:23
  dust. Your monthly credit card statement will
• 20:25
  definitely thank you later. This level of strategic
• 20:28
  consolidation just makes logical and financial
• 20:30
  sense. We want to challenge you directly right
• 20:32
  now. Look at your own workflow. Pick just one
• 20:36
  deeply repetitive task you want to automate today.
• 20:39
  Start incredibly small, test the logic thoroughly,
• 20:42
  and build outwards from there. Do not try to
• 20:44
  boil the ocean and automate your entire life
• 20:46
  immediately. Just aggressively fix one annoying
• 20:49
  daily friction point first. Build that first
• 20:52
  successful Lego block. But we want to leave you
• 20:54
  with a slightly provocative thought. We are currently
• 20:58
  setting up AI tools specifically to manage our
• 21:00
  costs. We are literally using AI to route and
• 21:03
  hire other AI tools based on efficiency. How
• 21:08
  long until these systems start optimizing our
• 21:10
  lives entirely? Oh, wow. How long until they
• 21:13
  analyze our usage and start canceling our own
• 21:15
  software subscriptions without even telling us?
• 21:18
  Now that is a truly fascinating and slightly
• 21:22
  terrifying concept to ponder.