Log in
Subscribe
May 4, 2026
Speed Is Everything in Tech Right Now
Speed Is Everything in Tech Right Now
00:00
14:03
Transcript
0:00
It no longer matters if you make the worst product in the world. It only needs to be good enough and faster than the competition. And that last bit is key, faster than the competition.
0:11
This week's tech power law is speed.
0:14
We're gonna go through a couple of real-world examples, one modern day, one historical, on how this plays out, and then we're gonna get into how you can leverage this concept in your work to get ahead.
0:26
Let's start with a concrete and modern-day example. If you're unfamiliar with the concept of uptime, it is the most important concept in distributed systems.
0:37
If you're going to launch a service, whether that be YouTube, Instagram, a website, Uber, anything that runs on web architecture, web infrastructure, the most important metric to you is gonna be uptime.
0:50
And uptime, simply put, is what percentage of the time is your service online and running.
0:56
Uptime requirements vary from industry to industry, but in a lot of the industries I've worked in, mission-critical defense applications, banking, financial tech, you wanna shoot for something called four nines.
1:08
And four nines means that your service is online ninety-nine point nine nine nine nine percent of the time, which means it could be down for a maximum of fifty-two minutes and thirty-four seconds per year.
1:20
That's a very, very reliable service that could only be down for a little bit of time throughout the entire year.
1:25
So if we look at this ecosystem of web services and we ask ourselves, what are the most used services lately? What, what's everybody talking about? Why does this channel...
1:34
why did this channel even come into existence in the first place? AI. Everybody is using LLMs, whether those be from OpenAI or Anthropic.
1:42
And if we look by the model, which one is the most efficient in terms of getting you to a correct answer the quickest, it's gonna be Anthropic's Claude.
1:51
And so bringing this back to our concept of speed, speed in this situation would mean the LLM that gets you to a correct answer faster.
2:01
So although a model from OpenAI may eventually get you to the answer after some poking and prodding and move that button there, and that's not what I meant, if another model like Claude can get you there faster, then it's speed advantage.
2:15
Now, you might assume the uptime of a system like that should be really, really high, right? It should be four nines. It's mission critical. Everyone's using it. But that's not so. I have Claude's status page up here.
2:26
I'm recording this video on Saturday, so I wanna read you the latest figures as of Saturday. For the last ninety days, Claude's uptime is ninety-eight point seven one percent. Ninety-eight point seven one percent.
2:39
Listen, if I am an engineering manager or a software engineer on a service that has something like ninety-nine point nine nine percent uptime, I'm starting to get worried, and if it slips below ninety-nine percent,
2:53
I would be very, very surprised that I hadn't been fired long before that. A-and that's not, like, a high bar for myself. That's just the standard of operating in our web ecosystem.
3:03
Almost by definition, if your service is offline, you can't win the speed game.
3:07
Anthropic has so definitively made a better model that they have realized that even with that amount of downtime, people are still gonna use them.
3:15
They're gonna tolerate the outages because their model arrives at the correct answer faster than the competitor models, so they still have a speed advantage.
3:24
But overall, it doesn't matter if your product is crappy, it doesn't matter if the uptime is bad. As long as it provides speed, a speed advantage, you can win in this modern technocracy.
3:37
And this is always how things have worked. Technology and war are deeply intertwined.
3:42
They are inseparable, much to many people's chagrin or urging that we should not use this terminology over here or that termin- terminology over there.
3:52
They are essentially two sides of one coin, and I don't say that just because I've worked in defense tech. This is provably true.
4:00
The SQLite database was invented to solve issues on a Navy ship with routing plumbing around the ship. ARPANET is what became the Internet, and ARPA became DARPA, Defense Advanced Research Projects Agency.
4:15
These two worlds are entirely intertwined. They build each other up out of necessity over time. But speed doesn't always lie in bits and bytes in uptime.
4:26
If we take technology in its broader meaning, that is a repeatable technique, something that you can do over time, we could also look at military tactics, which are a kind of technology.
4:35
It's a repeatable process or platform that can be used over and over again to shortcut thinking and provide, hopefully, an advantage if you have the better technology. And the Nazis did exactly that early in the war.
4:48
They used a technique called blitzkrieg, where they could rapidly descend on a city. They were operating at a timescale that was totally unpredictable. It was not planned for. Other strategists could not keep up with it.
5:00
It was operating on a totally foreign timescale because they were moving so quickly, and it allowed them to gain a quick advantage early in the war. And that lesson kept repeating. Radar beat biplanes.
5:12
Code breaking beat manpower. Whoever moved information faster won, and this is how things will continue to go, although the definition of speed is shifting slightly, as we saw with Anthropic.
5:23
In the technocracy, speed is power. So here's where this gets really relevant to you.
5:28
I have at least one or two tech employees every month or so email the show saying, "W-what do we do about enshittification, and when is craft coming back to our discipline?"
5:38
Craft is not coming back to software engineering. Craft is dead. Craft is a relic of the past, and no amount of whining and complaining is going to bring a high level of craft back to software engineering.
5:49
It is dead and gone, and the faster you can get that through your head, the quicker you'll be happier.
5:55
The technocracy only has one speed, and that speed is fasterBut the logical stopping point of that faster, and a lot of you are noticing this, you're writing me about this, about your experiences with this.
6:08
I'm seeing it for the companies that I'm advising on a day-to-day basis.
6:12
Just as in our blitzkrieg example where other generals could not comprehend the timescale that the Nazi armies were moving at, we're gonna get to the point as humans where we cannot comprehend the timescale that the machines want us to move at.
6:26
We simply will reach a point where we cannot move faster. This is what's going on now if you look around American society.
6:33
People are breaking because that breaking point is different for everybody, but we're entering a range where for numbers of people, large numbers of people, they are past their breaking point. They cannot move faster.
6:43
They are biologically not equipped to move faster, and they are breaking because they cannot keep up with the rate of technological change. So what do you do about this?
6:53
I'm gonna give you the example of what a lot of people are doing now that is not working, although it's a seductive thing to do, and I'm gonna tell you why we should not engage in that.
7:01
And then I'm gonna tell you the real solution for this that'll buy you at least a little more time. So the Band-Aid solution that's going on now is to enhance yourself, right?
7:11
To use stimulants, and we're seeing this everywhere. How many podcasts, YouTube channels, sports teams do you know that are pushing nicotine pouches?
7:22
I saw just the other day, uh, Tucker Carlson is plugging these things as like saying nicotine is safe and it has no, virtually no downsides and you should just use it.
7:30
You know, don't pay attention to the fact that he owns a nicotine pouch company and he's also plugging these things. I'm seeing-- I watched a Formula One race last weekend.
7:39
I saw the Ferrari car has Zyn on top of the, uh, rearview mirrors of the car, the side view mirrors.
7:46
So this stuff is wedging itself into culture, and the message is you need to move faster and you need chemical enhancement to do that.
7:51
The number of friends I have that are addicted to nicotine pouches that could not quit tomorrow is surprising.
7:58
I mean, these are thinking, intelligent people that are not seeing the consequences of what they're doing of using these things every day and becoming addicted. It's everywhere in tech right now.
8:07
Palantir is actually stocking vending machines in their offices with nicotine pouches, so even the employers are encouraging you to use these Band-Aid solutions to go faster, faster, faster.
8:16
Now, I'd be remiss if I did not mention that it is an open secret in my industry of software engineering that most of the engineers at top-tier companies, at OpenAI, at Anthropic, at Google, they're using stimulants to keep up.
8:31
They're using Adderall, Vyvanse, what have you, these legally prescribed methamphetamine stimulants to keep up with the technology. And who could blame them?
8:40
I mean, if you're gonna go hard, if you're gonna swing for the fences, if you are trying to out-compete somebody for a member of technical staff job at OpenAI without using Adderall, I mean, that's like trying to beat Mark McGwire and Sammy Sosa for home runs in nineteen ninety-eight without shooting up with a bunch of testosterone and Deca.
8:59
Like, it's just not gonna happen for you. You can try, but you're gonna be humiliated. Band-Aid solutions are never the answer. They are always temporary. They always wear off. They always have downsides.
9:11
And anyone selling you a Band-Aid solution, anyone selling you a Band-Aid solution is a pro at hiding their own use of these solutions and a master at deceiving you from what the downsides are.
9:25
They do not tell you that the nicotine pouch wears off and they're grumpy when they go home with their wife and kids and it's souring their relationship.
9:32
They do not tell you that their Adderall addiction is something they cannot go without or else they'll go into severe withdrawals and they'll forget how to write code.
9:41
The Band-Aid solution is never the answer, and so I want to offer you something that is gonna help you out. It's not a permanent solve because there isn't a permanent solve at this point. There isn't one.
9:51
Eventually, everyone will break on the technological wheel as things speed up and get too fast. It shows no signs of slowing. But here's what's gonna buy you a little bit more time. Get yourself a small notepad,
10:05
something like this size. Get yourself a small notepad. Keep it by your desk. Keep a pen with you.
10:11
And for a whole week, whenever you have a little bit of downtime, a break, you're going to lunch, just jot down a couple of notes about what you did in the previous period of work, what tasks you accomplished.
10:21
At the end of the week, go get yourself some coffee on Saturday.
10:24
Go sit down at a coffee shop with your little notepad and look through what you did throughout the week and try to think, "What are the highest value things that I could start to automate there with AI?
10:34
What are the highest value things?" I'll give you an example if you're stuck, if you need your mind, uh, jogged here with a couple of ideas.
10:42
I write a report for the person I report to every Friday at work, and that report covers what did I do this past week, what are my plans for the next week, and then do I need immediate attention.
10:53
Is anything like going wrong in some area of my remit on the report? And I like writing these reports, but the valuable thing about this report is not a litany of everything that went on during the week.
11:05
That's silly, and who's gonna read that? And it's gonna be long too. I do an insane amount of work every week.
11:11
The valuable thing is for me to reflect on what I've done during the week, to plan what I'm doing next week, and to present that through my lens as a human.
11:22
What's not valuable is the time it takes, about thirty minutes every Friday, to go dig through what meetings did I attend, what was I doing Monday? Because I, like I said, I do a lot during the week.
11:31
There's no way I'm gonna remember what I was doing Monday or Tuesday just off the top of my head. I'm gonna miss stuff. So I have to go back into my calendar. I have to go through Notion.
11:38
I have to go through Slack, and I have to remember, what did I do? What initiatives did I drive forward early in the week that I'm not remembering, that I'm gonna miss out on on these reports?
11:46
That takes about a half hour to get all the info together before I can start writing my report. So instead, what I do is I've set up an AI task. I can kick it off, and that AI task goes through Slack, Notion.
11:56
It goes through my meeting notes. It goes through my calendar. It goes through my emails.summarizes everything I did per day on a per day basis down to about one page.
12:06
So I have a quick list of everything I accomplished in the week.
12:09
It jogs my memory and I'm able to write that report pretty easily, which is the valuable part of that is my subjective take on what happened in the week, how I would present the information, what I found meaningful and useful and what I found not so useful and maybe a waste of my time.
12:23
That is extremely helpful. Saves me 30 minutes every week. It took me about an hour to set up. It's a repeating game. With that extra time, maybe you could have a walk outside, get some sunshine, take a breather
12:34
because that I tell you people is really the important thing. As things speed up, you're going to need to take more breaks. You're going to need to take more breaks. A lot of people don't understand this yet.
12:43
You have wackos like Sam Altman, who is on performance enhancing mental drugs, by the way, and he's confessed to this in numerous interviews. He is now doing like a polyphasic sleep cycle.
12:53
So he sleeps, you know, an hour and a half a few times a day as opposed to a big eight hour chunk because there's simply too much to do. He says with the LLMs, the LLMs are so good. He just has to be awake and working.
13:04
Seeing this all across my industry right now, people are breaking themselves. They don't realize it. They're working 16 hours a day, just cranking information to the LLM. And you can do that.
13:15
Then you can even do that without breaking for a decent amount of time.
13:19
But your output that you're going to get from that is it's falling off a cliff and you have AI psychosis that you're deluding yourself into thinking that you're creating value.
13:28
So the real superpower here is going to be to be able to automate some of these lower level tasks that'll get you some breathing room. That'll slow down the clock for you a little bit.
13:37
And then instead of filling those gaps with, okay, I can do 10 more tasks now, don't do that.
13:41
Go outside, get yourself a coffee, get yourself some fresh air during the day, take a little break, meditate, compose yourself before getting back to work.
13:49
And I can promise you, I can promise you, you may not spend as many tokens as the next person, but you're going to come back refreshed and you're going to work on those higher value ideas.
13:58
But it looks like the clouds are beginning to gather outside and it's going to rain.
The Dr. J Show
Recent episodes
The Flock Surveillance Problem Nobody's Talking About
May 2, 2026
SoftBank's $10 Billion Bet on ChatGPT Just Exposed the AI Bubble
May 1, 2026
