No matter your training and role in knowledge work, most of us will now publicly acknowledge we are confronted with a choice: augment or be replaced. We have an identity around our training, our actions, and our title. Augmentation, to survive, is more than doing “your job the way you did it before, but faster”.
—
Instead of disclaimer, I offer you a preamble and a human prompt so you may cope my writing. The grammar will be sub optimal. The style and transitions will be jarring. It is artisan. Enjoy the roughly hewn lumber, the uneven amount of chocolate chips in a cookie, and the sub-genius black and white movie pace I am capable of. With sincere respect, I suggest you may paste this into an LLM if helpful. I will not use an LLM today. I love them. They have their place.
—
At an extreme, capitalism may only have VCs. Deploy the capital and autonomous companies are formed and make money for the holder of cash, for whatever reason they deserve the power. The CEO is automated. The mission statement, the nurturing of culture, the “day-to-day” are all manifested and motivated by the lifting of a finger, a subtle wisp of movement, from a VC of the future. Entrepreneurship is the only job. Otherwise, bake bread for fun on a shameless life on UBI.
Or.
Be a CEO.
Optimization for wealth creation is a bleak future without culture and hope. A CEO does more than capital creation, in my biased opinion; maybe creates a vision, with care for a team, accountable to profit but doesn’t matter if the market don’t want what you make. You can be attacked and undermined from within and without. You are accountable. You own a culture.
Like a CEO, you are the limit. If you have never managed a fleet of AI agents running on GPT 5 or Claude 4.5 or later, (and they aren’t plugged into a tool or generating code), then you may not have experienced this feeling. How much you empower your team, your ambition, your judgement, your ability. Like a CEO, I need to think for hours, carefully, before I *should* tell my team of agents what to do. They will be done in minutes. I take hours. I am the CEO. My slow, deliberate, human pace matters because I am human.
Teams of CEOs, aligning their “companies” is the company of the future. You can’t perfect your CSS or your C++, you already lost. You can’t spreadsheet better or out-math your teammates. You have a team of A players at your disposal. You aren’t an IC any more. However, you are not only a manager now, you are a CEO. I understand you may not want to be a frontend development CEO. I understand if the prospect of churning out 100x of what you already do might not be right for you or the company.
Meanwhile, climate change, biodiversity, disease, science, nationalism, health care and other meat-puppet problems await solving. We learn to ignore many things, like death, for our mental health, as long as we can. Don’t look away. Surely, there is something worth being a CEO of.
AI
Measuring the digital divide
How many calculations do I consume per day (or per second), compared to a person with less access to a smartphone or the Internet?
Computation per capita could be a scientific measure of the digital divide.
Computing may forever be bits and powered logic gates. We measure computer output with FLOPS. [1] We can average total computational power per person. Segmenting by country adds more insight, which would be similar to maps of Internet access.
We make decisions in some parts of the world without much computation. To improve inequality we provide greater access to computation.
Google and other open services are a force for equality. The cost of a search for any user in one part of the world is near equal to all others. The equality isn’t perfect. Google has more data about me and how I live, so there’s more computation done on my behalf. But even if that is true, less Google is likely better than no Google.
We could try to measure computation per “decision”.
Decisions have measurable amounts of computation. What is the average amount of computation that happens when you tap your phone (in the cloud and on the phone)? I don’t know if we could measure the impact of a tap.[2] But we could calculate the computational output of a click or tap.
- a single google search
- an automated stock purchase, over a network
- pricing an airline ticket
- a second of a [self]-driving a car (NVidia suggests 24 teraflops)
- buying a dress
- a single web page loading
How could you compare the list above to one hundred years ago, or one hundred years in the future?
Perhaps the closest analog to the Internet in the last century was the number of telegrams, pieces of mail, or minutes of phone calls;
- to design and build a locomotive
- buy a stock
- buy a ticket on a trans-Atlantic zeppelin
- decide when to plant corn
- decide to get married
One hundred years in the future I’m not sure what we’ll be doing, but I hope we will still make useful decisions. Some possible examples:
- 3D print your fourth house by tomorrow
- create a new species of flower for your children’s garden
- make a copy of your brain and have it decide if you should get married by interacting with a copy of your potential partner’s brain
- bring your cloned cat Waffles the XIV home from its vacation to Mars
What could we name “computations per decision”?
Since I’m an unqualified amateur, I’ll make up a term; decision density.
The number of bits used per a unit of time to make a perceived decision.
Note “perceived”. By definition, we must be able to observe when we’ve made a decision. We do so based on the support of many calculations, but most computation is hidden [3] and increasingly ubiquitous.
If we can measure decision density per human, we would have another tool to measure the digital divide.
I look forward to mathy, articulate, and design oriented humans to elevate my pre–post-human writing. Otherwise, I’ve got this post until I can ask Her to fix it.
Notes
Elon Musk mentions a “recursive y axis” in the video below as a way to measure progress, over time, of computation.
[1] – Factoring in energy you have performance per watt and Koomy’s law. This doesn’t tell us if the decisions were correct, or efficient (see big O notation or examples of bad code). Nor does this measure tell us how much pre-calculation went into a decision.
[2] – Maybe impact could be measured by how much money flows as a result. Or using HDI.
[3] – A calculation is contained in a function like f(x)=y, and the decision is the output. We could also say a calculation is made by a computer, and a decision is made by a human. (That line is already blurry).
Also – Check out the posts on Quora, the conversation on Hacker News
Sneak peek at early decisions at HAI
Recently launched a site with some basic vision behind HAI.
But the tech stack is where the rubber meets the road. I’ve been coding about two months now. At the very beginning I went through a fair amount of thinking and ended up selecting a language for the backend based on a number of factors. From languages I knew, C++, Go, PHP, Python, Java/Scala, and Node.js were on the table. Python and Java were the two top contenders, but I ended up going with Python.
So far I’ve been really happy with Python for both flexibility of the language, the available libraries for both web and machine learning, and the developer community. Ruby / Rails has an amazing community and great web stack, but given my own lack of familiarity and less work being done in machine learning, it didn’t make my list.
Then I started evaluating open source projects that would be the platform. There are 132 on the list below (looked at least 4x that many). It’s been amazing getting up to speed on the projects that are open source. Although Google, IBM, Amazon and others are clearly going to lead in the machine learning space for the foreseeable future, the open source community is catching up.
Open source is a moving target, and there’s no one size fits all when you are piecing together something new. So, I’ve been using the awesome ZeroMQ library to connect services between libraries, languages.
Finally, thanks to everyone who has provided feedback so far. Can’t wait to get what I’m working on out into the world.
More about HAI
For a few weeks I’ve been having meetings with advisors and colleagues. For those I’ve not been meeting, I’ve fallen behind in communicating what I’m working on.
Not a stealth startup, but there’s also a lot that’s yet to be determined. I’d prefer to be open, but there are some specifications that I’ll keep under wraps for a variety of reasons…. When building a stealth aircraft, at the least you can tell people that you are building an aircraft. Skunkworks doesn’t make sandwiches.
Company vision and culture will be in large part determined by cofounders. Here’s where there’s some definition:
- HAI means AAI (Artificial Artificial Intelligence) – humans intelligence built into a process that’s usable by computer intelligence
- Ethical prime directives. See Friendly AI
- a sustainable business model early on
- Company culture of sustainable innovation modeled after Google’s large revenue generating platform supporting R&D.
- large, very talented, diverse founding team. Diversity is a no-brainer. Large is about five people; I’d rather create value from equity by distributing to founders than funders.
- Boston still has untapped talent and potential. Even if developers can find jobs easily, what kind of job would an engineer want for the rest of their life?
Oh. HAI.
HAI is the new company I’m working on. Human Assisted Intelligence. HAI will help computers learn about people. I’m so excited to be starting a new venture.
The future is happening right now. How far in the future will a product be relevant if you start developing it today?
Computer software / hardware outperforms humans in many specialized tasks today, and will likely surpass humans in categories reserved for our most revered public figures (scientists, politicians, performers) within 10-40 years. After the Singularity, quoting the WikiPedia main article -“Since the capabilities of such intelligence would be difficult for an unaided human mind to comprehend, the technological singularity is seen as an occurrence beyond which events cannot be predicted.” Where will humans fit into this future? A question for science fiction, perhaps.
Before the Singularity, what will help us engage effectively with a world increasing its complexity, knowledge, and economic dynamics exponentially? If super computer intelligent systems are used only by the most powerful institutions, what kind of intelligent service represents the individual?
I think these are the most interesting, important challenges around AI. Today, I’m building a team and product prototypes. If you’re interested in collaborating, feel free reach out.
Jonathan Hendler 