Musk counted robots, not output

Have you ever looked at a campaign, or an internal deck, or a sales pitch, and thought, "We are talking to ourselves here"? Written from our perspective rather than from what the customer actually needs?

I have seen it at a lot of companies and it has nothing to do with talent. When you work at a big company like Nike or Philips you learn the internal language fast, you get good at it, and everyone around you nods along. Then it leaves the building and lands with people who have never heard any of it. Same thing consulting for SaaS companies. Everyone in the room understands the words, so you stop asking whether anyone outside it does.

Back when content took time to create, that time caught it. You would be halfway through a content piece and stop and say, this is not it. The slowness was not the enemy. The slowness was the review. Nobody put that conversation in a calendar. It happened because writing was slow enough that you had to keep looking at the thing while you made it.

On 1 September Elon Musk addressed the G20 Innovation Ministerial by video and made a forecast he called conservative. Well over a billion humanoid robots inside ten years, each about five times as productive as a human, and together more productive than all of humanity combined. He said he would bet serious money on it. Once robots start making robots, he argued, growth turns recursive: slow at first, then explosive.

What struck me is what is being counted. Output per unit, multiplied. Nothing at all about whether the output is worth anything.

The common read on high output and flat numbers is that the prompting was weak or the model was wrong. Better instructions, better tool, better results. I think it is closer to the opposite. The tool is working exactly as advertised. It produces fairly good. That is the problem.

Most of us do not question "fairly good". A bad draft starts an argument, because bad is loud. Decent is quiet. It goes round the team, nobody has a strong objection, and it ships. The baseline moves down a notch without anyone deciding to downgrade.

Then you scale that. Forty of those. Forty newsletters, forty posts, because more content is the strategy everyone got told to run. Those forty are not compared against your best work. They are compared against the new benchmark. That is the quality your AI is now working against, and it will hold that line perfectly.

None of this is the AI's fault. AI will confidently produce horrible content and call it great. That is the job description. What changed is that somebody used to say this is not it, and stopped saying it. Usually it is not laziness. Killing a draft that arrived in seconds looking finished feels like a bigger call than killing three paragraphs you were still typing.

Everybody says human in the loop. In the loop means you are the last checkbox before publish. A human in the lead means you set the bar before the AI starts and you hold it after. That is judgment and taste, and it is the one input that does not get cheaper.

My goal with this newsletter is simple. Something you can read fast that is worth three of your day. You might be looking at this issue of LEVERAGE thinking, “This is fairly good”. And I would be quite pleased with that to be honest.

The question I keep coming back to is this. If you took the last three things your team published and ask these four things, how many would pass? Does this actually make our customers' lives better? Does it sound like us? Does it fit what we have been saying? Is it true, and did we really do the thing we said we did?

Whatever number comes back is closer to your real benchmark than the number in the content calendar. I am figuring this out alongside everyone else navigating the constantly evolving world of AI in marketing and sales.