I agree with the general idea with one important caveat. The cuts were needed because things were impossibly bloated during ZIRP/covid boom that were in dire need of correction. AI just became the final straw/a sort of spotlight vs the real underlying reason. He and other tech ceos have to do the same thing to recover from the overeating of the last decade plus.
I’ll disclose my disposition first so that it can be read with that tone: “I’ve been telling people for years.”
The bad thinking and resulting problems have existed before AI. I think you agree, given you’ve mentioned them too for awhile. All problems: functional organization, too few reports per manager, company too big, too many managers, wrongly skilled managers, silos/walls generally but specifically in the wrong places, no vision and no one coordinating a vision (every leader expects it from the leader above them and gets plausible deniability), there are many.
More blandly, there has now long been a default way to build companies that founders (and then the top and middle managers they hire) can lazily follow. None of it tailored to your skills, your market, your problem at hand. The best they can do is fund some functions more than others to follow the landscape of their problems, but it doesn’t solve the massive “cross functional” problems (which exist because you’ve separated functions) and the general sociopathy behind trying to get promoted, which is incentivized by having so many layers. If you’re focused on money or title, but importantly also if you’re focused on the product, you’re blocked by functional managers and the fact that you can’t build or accomplish anything with your singularly functioned team.
I say this to describe the current “skill set” of employees and leaders and identify the major obstacle for Block and the industry in making this move (and perhaps why the safer bet is that they’ll hedge because of the obstacle). All the employees think and have been trained to work this way. The higher you go, the more success they had in this system, the more they were picked for these unwanted skills. These guys can’t handle 8 or 20 reports in the way you want. They wouldn’t know what to do with prototypes because they’re not use to being a part of the vision in the first place. How would they know if this fits or doesn’t fit it? They’re too far from the specifics of the project on the ground and too far from the big picture. Maybe some of them would gather the skills back that made them a good IC, but many of them were inadvertently selected for their obtuseness and their ability to not ask questions and slow “progress.”
If the transition towards this new company structure/process works, it’s going to benefit from a lot of wiping the board clean. Whether that means hiring from scratch and somehow having the people who understand and are capable of hiring differently or somehow retraining everyone alongside the reorg, this is very difficult—but it’s specifically difficult for leaders who have their hands on the reins to do it now, but also had them on the reins leading the team here in the first place.
It was a smart bet for him to make. However, companies that have actual customers still have to care about quality, uptime, and architecture in a way that startups never have. That has always been a huge tax and will continue to be.
Startups always had young people grinding out code for 80 hours a week, because the execution path from 0 to 1 can be more straightforward, and without customers you can strategically cut corners. AI Startups will still eventually hit a wall in how they operate.
The bet is these LLMs will continue to get exponentially better, but I also think there's a chance they plateau and successful teams will always have to refine work and be strategic about scaling.
Basecamp (DHH and Jason Fried) and David Marquet are great to study. Basecamp would simply not have grown the way he did. They don't need prototypes, they don't need large teams, they eliminated that meeting. AI is not a factor in their low hierarchy. I think Basecamp represent the closest thing to Dorsey's dream. See Shape Up and Rework books. Maybe AI can help them scale even more, but I think the design is already sound.
Marquet would keep middle management, but would retrain them to be coaches and mentors. For him it's all about how you can push decisions down, instead of bringing decision up in the hierarchy. If you push authority down, you must obsessively train and certify your people. It requires deep human investment. I think Dorsey using AI to do that task is probably going to hurt him.
This is also by definition not leading to a wonderful business that an idiot could run.
look who has made his was to substack
Love this. Understanding your own and others' strengths and weaknesses is important context in every interaction.
that duality is what makes him so hard to evaluate. most people can only hold one of those jacks in their head at a time
This was a great read
I agree with the general idea with one important caveat. The cuts were needed because things were impossibly bloated during ZIRP/covid boom that were in dire need of correction. AI just became the final straw/a sort of spotlight vs the real underlying reason. He and other tech ceos have to do the same thing to recover from the overeating of the last decade plus.
I’ll disclose my disposition first so that it can be read with that tone: “I’ve been telling people for years.”
The bad thinking and resulting problems have existed before AI. I think you agree, given you’ve mentioned them too for awhile. All problems: functional organization, too few reports per manager, company too big, too many managers, wrongly skilled managers, silos/walls generally but specifically in the wrong places, no vision and no one coordinating a vision (every leader expects it from the leader above them and gets plausible deniability), there are many.
More blandly, there has now long been a default way to build companies that founders (and then the top and middle managers they hire) can lazily follow. None of it tailored to your skills, your market, your problem at hand. The best they can do is fund some functions more than others to follow the landscape of their problems, but it doesn’t solve the massive “cross functional” problems (which exist because you’ve separated functions) and the general sociopathy behind trying to get promoted, which is incentivized by having so many layers. If you’re focused on money or title, but importantly also if you’re focused on the product, you’re blocked by functional managers and the fact that you can’t build or accomplish anything with your singularly functioned team.
I say this to describe the current “skill set” of employees and leaders and identify the major obstacle for Block and the industry in making this move (and perhaps why the safer bet is that they’ll hedge because of the obstacle). All the employees think and have been trained to work this way. The higher you go, the more success they had in this system, the more they were picked for these unwanted skills. These guys can’t handle 8 or 20 reports in the way you want. They wouldn’t know what to do with prototypes because they’re not use to being a part of the vision in the first place. How would they know if this fits or doesn’t fit it? They’re too far from the specifics of the project on the ground and too far from the big picture. Maybe some of them would gather the skills back that made them a good IC, but many of them were inadvertently selected for their obtuseness and their ability to not ask questions and slow “progress.”
If the transition towards this new company structure/process works, it’s going to benefit from a lot of wiping the board clean. Whether that means hiring from scratch and somehow having the people who understand and are capable of hiring differently or somehow retraining everyone alongside the reorg, this is very difficult—but it’s specifically difficult for leaders who have their hands on the reins to do it now, but also had them on the reins leading the team here in the first place.
I always thought that "bring a demo or prototype" was underrated even pre-AI. Real 8th grade English class stuff: show don't tell
It was a smart bet for him to make. However, companies that have actual customers still have to care about quality, uptime, and architecture in a way that startups never have. That has always been a huge tax and will continue to be.
Startups always had young people grinding out code for 80 hours a week, because the execution path from 0 to 1 can be more straightforward, and without customers you can strategically cut corners. AI Startups will still eventually hit a wall in how they operate.
The bet is these LLMs will continue to get exponentially better, but I also think there's a chance they plateau and successful teams will always have to refine work and be strategic about scaling.
Great post, thanks!
Basecamp (DHH and Jason Fried) and David Marquet are great to study. Basecamp would simply not have grown the way he did. They don't need prototypes, they don't need large teams, they eliminated that meeting. AI is not a factor in their low hierarchy. I think Basecamp represent the closest thing to Dorsey's dream. See Shape Up and Rework books. Maybe AI can help them scale even more, but I think the design is already sound.
Marquet would keep middle management, but would retrain them to be coaches and mentors. For him it's all about how you can push decisions down, instead of bringing decision up in the hierarchy. If you push authority down, you must obsessively train and certify your people. It requires deep human investment. I think Dorsey using AI to do that task is probably going to hurt him.
This is also by definition not leading to a wonderful business that an idiot could run.
Underrated leader.