Should You Start Slow and Build Speed?

You know, there's this piece of advice that I feel like has been drilled into us since we were kids. It applies to everything, really. Investing, buying a new car, and especially technology. It's always, don't be the guinea pig. Wait for the bugs to be worked out. Right. Let someone else crash first. And it sounds so sensible. It feels like the responsible, you know, the adult way to handle this whole AI revolution. Just wait. Wait until the dust settles. Wait until the manuals are written, then jump in. Exactly. But we have a source on the desk today that it takes that conventional wisdom, crumbles it up, and basically sets it on fire. It really does. We're doing a deep dive on this fascinating article titled, Start a Lungsumped Fur at Sedan Uka, which, you know, translates to start slow to then increase. And the answer it gives is just a resounding no. A hard no. A hard no. It argues that in the current landscape, caution is actually the most dangerous risk you can possibly take. That is a stressful. The idea that playing it safe is actually playing to lose. So our mission today is to figure out why this source claims that speed gives balance, why the traditional career ladders we've all been climbing are, well, dissolving. And we're going to spend a lot of time on this concept of the orchestrator. And honestly, if you're listening to this and you're thinking, I'll just wait until next year to figure out this whole agentic workflow stuff. This deep dive might make you sweat a little. Yeah. The source argues that. Waiting isn't just pausing. It's rapid depreciation of your value. Okay. So before we get to all the scary stuff, let's look at why our intuition is so wrong here. Yeah. Because historically, being the second mover was smart. We're not crazy for thinking that. No, you're not crazy at all. It was the smart play. I mean, if you look at the last 30 years of tech Excel, Photoshop, even the early Internet caution paid off. Right. You didn't install Windows 95 on day one. No way. You waited for the service pack. You waited for the best practices book to be written. You let the early adopters deal with all the blue screens of death. Precisely. And more importantly, once you learn those systems, that knowledge was a fixed asset. If you became an Excel wizard in 2005, knowing all the pivot tables and macros, you were still an Excel wizard in 2015. Your expertise had a long shelf life. A very long shelf life. The source calls this the old model. But the source says the math has completely changed. Why doesn't that apply to AI? Why can't I just wait for the... You know, the AI service pack? Because two fundamental assumptions have just broken. First, the assumption that your expertise lasts longer than the technology changes. That's gone. The half-life of a specific technical skill right now is just incredibly short. And the second one? The assumption that career paths are stable silos. To explain this new reality, the source uses a metaphor that I think is really, really sticky. It compares adopting AI to riding a bicycle. It's a perfect analogy for the physics of this transition. I mean, think about watching a kid learn to ride a bike. When they go really slowly because they're scared, what happens? They wobble. The handlebars go everywhere. They overcorrect. And then, you know, they usually tip over. Exactly. In physics, stability comes from angular momentum. It comes from speed. And the source argues that AI competence is exactly the same. So if you try to learn it slowly and safely... You're wobbling. You're constantly braking. You're just dipping a toe in once a week. Because by the time you learn the safe version of the tool... The tool has already changed. Right. You're studying a snapshot of a river that has already flowed downstream. Stability doesn't come from standing still or moving cautiously. It comes from moving fast enough for the system to actually carry you. You have to generate enough speed so that momentum keeps you upright. So speed is security. That's the headline here. Now, if we're all pedaling this fast, where are we actually going? Because the source talks about a massive merger of roles. It suggests the way we define our jobs is about to get very, very messy. This is the part that I think will resonate with anyone who feels like their job description is getting blurry. Historically, corporate structures were built on silos. You were a developer or a marketer or a designer. You stayed in your lane. The marketer didn't write code and the coder hopefully didn't write the ad copy. Exactly. But the source argues these roles are melting into a single universal competence. The orchestrator. The orchestrator. Okay, it sounds impressive. But let's ground that. What does an orchestrator actually do on a Tuesday morning? An orchestrator doesn't focus on doing the task. They focus on getting the past done by managing resources. And in this case, the resources are digital agents. The source breaks this down into maturity levels, and most people right now are at level zero or maybe level one. Let me guess. Level zero is using ChatGPT like a glorified Google search. That's it. Tuna, write me an email. What's the capital of France? That's level zero. It's useful, but it's not orchestrated. The goal is level five, full orchestration. That's where you aren't just asking questions. You're managing a team of digital agents to execute complex workflows. Okay, I want to pause on that word agent. We hear it a lot. It's becoming a buzzword. How is an agent different from just chatting with a bot like we've been doing for two years? That is the crucial distinction. A chat bot talks. An agent does. Think of a standard large language model in LLM as a brain in a jar. It can think it can write poetry, but it can't touch anything. An agent is that brain given hands and eyes. Hands and eyes. Metaphorically. Metaphorically. An agent is an AI model that's been given access to tools, your calendar, your email, your code repository, your CRM, and the permission to use them to achieve a goal. So level zero is asking the bot to write an email draft for me. Level five is telling an agent, look at my calendar, find a slot, book the meeting with the client and send them the agenda. And the agent actually does it. Precisely. And that requires a completely different mindset. You're not a creator anymore. You're a manager. You're managing a workforce of digital entities. This shifts the power dynamic significantly. And the source actually gives examples of who is going to win in this new world. And it wasn't who I expected. I kind of expected it to be the hardcore coders, the tech bros. It's not. The source highlights three avatars, three types of people who are perfectly positioned for this. And they're very different. The first one is Anton. The digital native. He's a young lawyer, right? Maybe a junior associate. Right. And Anton's superpower isn't that he knows how to code Python. It's that he doesn't have baggage. He doesn't have to unlearn things. He doesn't know how things used to be done. Exactly. He doesn't have to unlearn 20 years of this is how we draft a contract manually. He can skip directly to orchestrating agents because he doesn't have these ingrained habits fighting him. He treats the AI workflow as his native language. I feel that. Sometimes the hardest part of learning a new tool is stopping yourself from trying to use it like the old tool. It's like trying to use a tablet with a mouse. You're just fighting the interface. It's exactly like that. But then look at the second example, which I found really encouraging. Ellen, the experienced manager. Right. Ellen has been managing teams for 10 years. She's not technical in the traditional sense. So why is she winning? Because she knows how to delegate. Her advantage is that she already thinks in terms of responsibilities and outcomes. She doesn't think I need to type this document. She thinks, I need this project delivered by Friday and I need these three people to contribute. She's used to delegating to humans. Now she's just swapping human direct reports for digital ones. Bingo. The skill of saying, here's the goal. Here are the constraints. Go do it and report back. That is exactly what prompting an agent requires. Her management skills are her technical edge. That's a great reframe. Management is the new coding. And then we have the third one, Thomas, the senior generalist. Thomas is the veteran. He's had many roles. Many industries. He's seen trends come and go. His critical asset is something the source calls taste. Taste. That sounds so vague. How do you put taste on a resume? It sounds vague, but it's actually very concrete in an A.I. world. I mean, think about it. A.I. can generate infinite variations of a marketing campaign or a code base or a design in seconds. Generating content is now cheap and infinite. So the bottleneck isn't creating the stuff, it's knowing if the stuff is any good. Exactly. The ability to look at 10 A.I. generated options and instantly know that one is garbage, that one is generic, but this one, this one will work. That is pure gold. That requires deep experience and intuition. So judgment is the barrier to entry now. Yes. If you can't discern quality, you can't orchestrate effectively. You'll just generate high speed mediocrity. You'll be peddling very, very fast towards a cliff. OK, so we know who can succeed. But let's talk about how. The source gets into the mechanics of this managerial mindset is we have to stop thinking like workers and start thinking like CEOs of our own little digital corporations. It goes back to that shift from doer to orchestrator. If you're treating A.I. like a tool you hold in your hand, like a hammer, you're doing it wrong. You need to treat it like an employee you are training. And the source gets surprisingly specific here. It says to be effective, you have to understand agent reality. You can't just yell at the cloud. You need to understand three specific things about the agent you're working with. The toolbox, the memory and the workflow. Let's break these down because this feels like the how to manual. Right. Let's start with the toolbox. This is the most common failure point for beginners. You ask an agent, get me the client data for the last quarter. Sounds reasonable. It does. But does the agent actually have access to your CRM? Can it read your Excel files or is it just a touch generator trapped in a browser tab? It's like yelling at a carpenter to saw a word when you haven't given him a saw. Exactly. You have to know what systems the agent can touch. Grounding is the technical term for it. If the agent isn't grounded in your data, it's just going to hallucinate. It will make up a reasonable sounding sales report because it wants to please you. You have to bridge that gap. OK, second one, memory. This one trips me up sometimes. I'll be in a long thread and suddenly the bot forgets what we were talking about five minutes ago. It trips everyone up. You have to understand the context window. Does the agent remember what you talked about yesterday? If you say update that draft we were working on and the agent starts a fresh session every time, it has no idea what that draft is. So you have to manage the agent's memory. You have to be the keeper of the context. You have to constantly reminded of the constraints. Remember, we are writing for this specific audience or use the data from the file I just uploaded. You cannot assume it remembers the backstory unless you explicitly manage that. And the third one, workflow. This sounds like standard project management. It is. But it's applied to non-human logic. This is about defining the result, not the method. A micromanager tells a human exactly which buttons to click. A leader just says, here is the goal. The source gives a great example. Don't say increase sales. That's too vague. It's a wish. So what's the better command? Deliver a prioritized list of leads from our database that match Profile X and draft an outreach email for each one. That's a huge difference. One is a vague desire. The other is a deliverable work product. And that shift learning. To speak in clear outcome based deliverables is the hardest skill to learn. It's not coding. It's clarity. If you're a messy thinker, AI will amplify your mess. AI amplifies your mess. That should be in a T-shirt. I want to circle back to why we need to do this now, because the source throws some numbers at us that are frankly kind of terrifying regarding the speed of improvement. They are. The section on the urgency of now is a wake up call. The source mentions SWE Bench. That's a benchmark for software. Engineering, right? Correct. It tests how well an AI can solve real world coding problems. Issues pulled from actual GitHub repositories. It's not a multiple choice test. It's fix this broken code. Two years ago, the success rate on that benchmark was four percent. So essentially useless for real work. Yeah, it got ninety six out of one hundred questions wrong. Right. And if you looked at that two years ago, you have been justified in saying this is a toy, I'll wait. And now ninety five percent. Whoa. From four percent to ninety five percent. In two years. In two years. So imagine you decided to wait and see two years ago. You blink and now the technology is operating at a near expert level and you have zero experience working with it. You are now competing with people or agents that are exponentially more capable and you haven't built the muscle memory to handle them. That is the danger of the wait for it to mature strategy. It matured while you were sleeping. And the money following this is just absurd. Over two trillion dollars. Projected in infrastructure investment from big tech. This isn't some crypto fad that might blow over. The physical infrastructure of the planet is being rewired for this. The source also mentions the depreciation of knowledge, which really hit hard. What you learned last year is half obsolete. It's the half life of skills. In the past, you went to university, got a degree, and that degree carried you for a decade. Now what you learn in January about a specific software framework might be the old way by October. So if the ground is moving that fast, it still is basically moving backwards at high speed. Exactly. And that brings us to the action plan. The source is very blunt. Stop waiting. It mentions specific tools, right? Like Claude Code or lovable. It does. And those are great tools for coding. But the specific tool matters less than the behavior. The advice is to accept the messiness. The early tools are messy. They hallucinate, they break, they get stuck in loops. And that's usually where people quit. They try it once, it fails and they go, see? It's not ready. I knew it. But the source says volume over perfection. The goal isn't necessarily to succeed at the project today. The goal is to build the feeling of how AI logic works. It's muscle memory. It is. You need to run enough prompts, hit enough walls and see enough successes that you stop looking at AI as magic and you start seeing it as logic. You start to intuitively know, oh, it's going to struggle with this part because of the context window. So I need to break this into two steps. You can't learn that from a book. You only learn it by doing. So the advice is, yeah, pick a problem today, not some huge strategic initiative, just something small and annoying. A broken process in your daily life. I hate sorting these receipts. I hate summarizing these meeting notes. Put an agent on it. Watch it fail. Fix the prompt, ground it in the right data. Try again. That iterative loop is how you become an orchestrator. I love that it takes the pressure off. You're not trying to transform your company overnight. You're just trying to train your brain to work with a new species of employee. And that leads to the final really powerful image the source leaves us with the Ferrari metaphor. Oh, this is good. So the source warns that if you take the slow approach, if you just use AI to do what you're already doing a little bit faster, it's like buying a Ferrari and then just setting the cruise control to 30 miles per hour to drive to the grocery store, you're technically using the car, you look cool in the driveway, but you're not using the capability. Exactly. The people who take it slow will just be slightly faster versions of their old selves, but the orchestrators, the ones who really lean into the speed, they won't just do things faster, they will do completely different things. Can you give me an example of that? What's a Ferrari move versus a grocery store move? Grocery store move is write this client email for me. It saves you five minutes. A Ferrari move is analyze the last five years of our customer support tickets. Identify the top three recurring complaints. Cross-reference. And then you can do that with our product roadmap and draft a strategic proposal for how to fix the root causes. Wow. Okay. That's not just saving time. That is generating new value that a human probably wouldn't have had the time to do at all. Exactly. A task that would have required a team of 20 people in six months is now a Tuesday afternoon for one skilled orchestrator. That is the difference. That is why you can't just wait and see. So as we wrap up this deep dive, the question really shifts. It's not, can I keep up? It's deeper than that. The question is, when you stop being a doer, when you stop taking the emails, writing the code, formatting the slides and you start orchestrating, what are you going to do with that freedom? What is your unique value when the execution becomes free? That's the provocative thought. If you aren't buried in the how, you have to get really, really good at the what and the why. You have to have a vision. And that requires that judgment we talked about. Thomas's superpower. Precisely. The machines can build the road, but they can't decide where it should go. So here is our challenge. To you, the listener. Don't wait for the perfect time. Don't wait for the safe version. Go find one small, annoying problem in your workflow today, not tomorrow, today, and throw an agent at it, let it wobble, let it crash, then get back on the bike and pedal faster because speed is the only thing that's going to keep you upright. Thanks for listening to this deep dive. We'll see you out there on the track. Keep pedaling.