When Seth Godin first came on Invisible Machines, generative AI was still the shiny object and The Song of Significance was the handbook. He said then that you cannot talk people out of fear. They will move when the thing is cheaper, more convenient, or more powerful. He also said we were entering an era as different as electricity—and that competing with the steam shovel was a bad career plan.
He is back three years later, on the day The Knot: Problems Can Be Solved comes out, and the compression is tighter. Josh Tyson names the arc: from Purple Cow through Song of Significance to The Knot, Godin has been compressing how people make decisions toward something that still works when you hydrate it with new data. Bad compression dissolves. A demo that only works on one dataset is bad compression. Godin calls the read a brilliant insight, then gets stricter with himself: if it is a special case, he has not added any value.
The trickiest word, he says, is purpose. He does not start there. He starts with two questions that sound smaller and are not: what's this work for, and who is it for? Fake mission statements are a known technology. A company can engrave patients, doctors, nurses, employees, then shareholders in granite ten feet tall and still get caught serving the bonus pool. They are afraid to put the real purpose in giant letters. Get clear about what you actually signed up for, and a lot of AI decisions get easy. If you are here to serve these people, of course you use the tool. If you are here to fully employ artisans, maybe not.
Robb Wilson's corollary is the one that will travel. People already know when you are lying. They hear "we're here for patients" and translate it to stock price. AI just believes you. Feed it the granite and it will optimize for the granite. Then you will ask what happened to the stock price. Godin: exactly.
Entanglement in a business suit
That is entanglement in a business suit. Godin borrows the physics metaphor on purpose. Einstein was wrong about spooky action at a distance; tiny particles are entangled. So are we—with feelings, fears, obligations, and goals we do not want to say out loud. When change shows up, decision-making goes haywire. We get trauma instead of resilience because we refuse to deal with the invisible commitments holding the knot in place.
The posture that cuts it is in the subtitle. Problems can be solved. AI is a problem because it is a change agent: it changes what competitors do, what customers expect, things outside your control. If problems are bad, AI is bad. If problems are opportunities, this is the change agent you have been waiting for. Fear, in the 2023 conversation, was already a situation. You do not sword-cut a situation. You build the thing that makes the fear lose its structural job.
The demo was perfect
History is the other half of the patience argument. Godin believes AI is the biggest change since electricity. Edison fought AC by electrocuting elephants in public. Companies promised to adopt last. There were no outlets; you unscrewed the light bulb. Washing machines marched around the room and killed people. You did not see that at the demo. The demo was perfect. It took decades to get good, safe, and productive, and the things we do with electricity now were not on Edison's top hundred. A CEO who says the tool must be perfect or she will ignore it is making a mistake. A CEO who says it will do this, right now, forever, is making the other one.
Josh's warning sits on that seam: revealing what's possible is powerful, and a bad demo jams another pin into the center of the knot. Godin has been saying a version of this since the first visit. The hospital-crib factory only took the CNC machine because the identity was "make cribs for sick kids," not "bend steel." Name the real commitment and the knot can fall apart without a fight.
You are the circumstances
The split the season keeps finding is here too. At the micro level, of course your employees will use AI to save time. High school students went first; now it is everyone. At the macro level, most companies are not making the right thing. The furniture company that kept making furniture after container ships made imports a quarter of the price struggled and died. The ones that said we design and sell furniture did fine. This is that times a thousand. The people who will not choose will blame inertia. They are Kodak. They will say that under the circumstances they did their best. Godin: you are the circumstances. You are not allowed to say that.
Screwdrivers, not solutions
Why isn't it happening at a corporate level yet? No one is in charge of AI—it is not a single product. And the biggest early movers are toolmakers, not solution providers. FedEx sold a whole product: when you absolutely positively have to get it there overnight for $12, the envelope appears. Anthropic and OpenAI handed over screwdrivers, wrenches, and saws. Figure it out. That produces trauma, because features are not a solution.
The successful users, Godin says, will erect constraints. The industry would like you to believe there are none: it is a miracle, it is free, it will always be both. He has not had meat in 45 years, mostly because he has a rule and does not have to re-litigate lunch. He publishes a blog post every day so he does not have to debate whether this is the best post. Constraints are a foundation. You cannot think outside the box. It is very dark out there. At the edge of the box is where you can make a miracle happen. For LLMs, Josh notes, the hallucination problem is a constraint problem. Godin: correct. An unconstrained system wanders. A constrained one can decide.
False proxies
False proxies are how stuck organizations fake the work. You should not hire a programmer by words per minute. You should not manage AI by how many tokens someone is burning. It gives the clueless boss a ranking. IBM learned this with the 360: lines of code per day is a terrible way to build a mainframe. Robb has been saying the same thing as token maxing. Godin names the pattern: we measure what is easy, then act surprised when the purpose does not show up in the metric.
Proximal empathy
The human work does not go away. Josh brings Marina Nitze's call center that was not a call center: you have to walk there. Godin would call that proximal empathy. You cannot be a tourist. San Francisco is more AI-friendly than Topeka not because the people are smarter. They have scaffolding and peer support. Culture is "people like us do things like this." The Costco hot dog is a signal. Howard Schultz closing every Starbucks was a signal. The foundation holds the thing up, not a magic wand from the top.
Compromise splits the same way
Most of the compromises we make are with people who do not want us to succeed, who are not our customers. The steakhouse puts a mediocre vegan dish on the menu to make the argument go away, then wonders why vegans do not come. Refuse to compromise with non-constituents. The other kind is empathy: I lacked the imagination to see the world the way you see it, and if my purpose is to serve you, I will change the building. Josh brings Mary Parker Follett's electrician—half a house on fire and half unlivable is the compromise of average; a third plan is the point. Godin: beautifully said. A great architect is not a prima donna.
The volvelle
He has a two-sided volvelle of questions nobody wants to answer. When it is important enough, they do. Great captains do not get in the boat and accidentally arrive. They bring oranges. The sailor who wants a coconut is told: this boat does not go there. Josh closes with Vonnegut: we are who we pretend to be, so be careful. If you feed the machine the person you pretend to be, that is who it will project. Godin: brilliant.
AI together
The opportunity he is most impatient for is still unbuilt. Everything on the internet is networked except AI. AI is solo—the first successful thing on the internet that was. He built "AI together" inside Discourse so a chatbot sits with eight people, talks once an hour or once a day, and does not suck the air out of the room. The people who will make AI a network thing have not built it yet. He can think of 500 places to put it tomorrow.
The pin is usually an identity, a borrowed scorecard, or a purpose you will not say in giant letters. Pull it, or keep hydrating a story that only works on the demo data.
Companion essay for the You Are the Circumstances ideation cluster—Seth Godin’s S7E18 return to Invisible Machines, read against S1E14 (2023). Related: Five Obvious (But Not So Obvious) Truths from Seth Godin, Five ONO Truths About Service Design, The Pyramid Was Always Wrong (S7E14).