If you aren't the fastest you will fail.

Above a quality threshold (lower than you think it is) the only thing that matters is speed.

The phenomenon is omnipresent but most visible across:

• Media
• Work
• Tech

The Failure that Could

It no longer matters if your product works, only that it is faster than the competition.

Uptime is the most important concept in modern tech. Uptime is defined as the percentage of time a service (website, app, streaming platform) is online and working. Mission critical services strive for 99.9999% uptime, colloquially "four nines" which means a service could be down up to 52 minutes and 34 seconds per year. That's seriously reliable.

The most popular services over the past six months have been LLMs, AI models, such as Anthropic's Claude and OpenAI's ChatGPT. What these services truly offer behind all of the marketing bullshit is speed. If an LLM can generate code or word documents or reports faster than a human at an acceptable level of accuracy, the technocracy deems the LLM superior, because it is faster. Claude in particular has the claim to fame here for providing more accurate answers than ChatGPT. If an LLM is able to arrive at a correct-enough answer with less prodding, then it wins the speed game.

You might think a service like this would aim for "four nines" reliability but you'd be incorrect. Claude's uptime for the past 90 days is 98.71%. It is the lowest uptime percentage I've heard about in my entire career in tech. It is so low that if I were on a team with anything less than 99.99% uptime I'd be dusting off my resume before the end of the week.

Even with this shitty reliability, Anthropic wins the speed game because their model arrives at a correct answer net faster than the competitors. If it's online and functioning, the technocrats will use their model, not the slower ones.

This is a glaring sign that speed has a flexible definition, but still rules.

And It Always Has

Tech and war are eternally intertwined. They are two sides of the same coin. People don't win wars, tech wins wars. SQLite was built to solve problems on a Navy ship. The internet was a DARPA project.

But it's not always circuits and bits and bytes.

The Nazi technique of blitzkrieg compressed decision making time and was one of the main reasons they were able to make so much progress early in the war. Their tactics existed on a timescale that was previously unimaginable to military strategists. Armies that had taken weeks to mobilize were outmaneuvered in days. The Allies didn't catch up until they matched the tempo.

The lesson kept repeating. Radar over biplanes. Code-breaking over manpower. Whoever moved information faster won.

Speed is power.

Living Under This Law

Every few weeks at least one tech employee emails me asking if I'll address the "enshittification" of the tech industry on my show. I can address it in one sentence: quality standards are not coming back.

The technocracy only has one speed: faster. It is a never-ending march towards more speed. First techniques: following a bread recipe is faster than deriving your own through multiple failed loaves. Then analog computers, then digital computers, until human-assisted speed is not enough. Eventually this will break humans which is why the tech oligarchs want to replace us with thinking machines in short order.

We are in a tough spot: we must work to provide income yet we will break ourselves if we move too fast.

Avoid the Band Aid Solution

The tech-owned media is pushing nicotine pouches very hard right now. Nicotine is a stimulant. It allows you to work a bit past your breaking point. Tucker Carlson extols the benefits of nicotine pouches on his show (never mind the fact that he profits from his own nicotine pouch company). Palantir is stocking their in-office vending machines with nicotine pouches (yes, really).

Virtually every excellent engineer I know in the Bay Area is coked out of their mind on stimulants. This is an open secret in the industry, and can you blame them? It does work to an extent. If you want the Staff Engineer job at Google or to become a "Member of Technical Staff" at OpenAI you gotta pay to play.

You could try and go without. But it'd be like competing for home runs versus Sammy Sosa and Mark McGwire in 1998. You can do it. You're only going to compete if you're juicing.

Band-aid solutions are always temporary, always have tradeoffs, and the pros are experts at hiding how "assisted" their efforts are.

Fast Where It Counts

We will not be able to keep up forever but there is one strategy that will help you stretch your career and your sanity a bit longer: just be fast where it counts. The easiest way to do that is AI. Here's a real simple process for identifying the bits you can speed up.

Get yourself a small notepad. Keep it by your desk while you're working. Whenever you get some downtime, jot down what tasks you've done, "wrote my weekly report, scheduled a meeting with so-and-so, etc." On Saturday, go over your list and pick out the biggest time-sink of your work and figure out how you can either automate the full thing with AI or automate some of the work with AI.

Then automate it.

An example. I write a weekly report to the person I report to at work. It outlines what I've done during the week, what I'm planning to do next week, and if anything needs immediate attention.

I do an insane amount of work in a week. It's impossible for me to remember what happened on Monday or Tuesday by the time it's Friday.

The valuable thing about that report isn't a checklist of what happened, it's what happened through my lens. That means I can automate the information retrieval but not the authorship, that is still mine entirely. So what I've done is set up AI to grab all of my Slack messages, emails, Notion docs, calendar events, and meeting notes from the entire week, then write me a one-page summary. I can read that, jog my memory, then jot down my report pretty quickly. It took about an hour to put the automation together and it saves me about 30 minutes each week that I would have spent digging through a bunch of boring data to remember everything that happened. I can take a walk outside with that extra time.

So don't break yourself by moving too fast. Look at where you can slot in AI for small menial tasks throughout your day. Eventually a critical mass of people will start breaking because the pace of everything has gotten too fast. Perhaps this is already what's happening in society. Regardless, this will buy us some more time.