The Commoditisation Of Intelligence

Demand for energy and compute will continue to soar for the near future, but there will be a point where it levels off.

It’s going to be a very challenging time for knowledge workers.

The main concern is that while the rate of technological shift used to be measured in blocks of 5–10 years, the turns have been getting shorter and shorter ever since AI became mainstream. If we track the recent major developments from LLM chatbots such as ChatGPT, open source frontier models such as DeepSeek, and agentic open source software such as OpenClaw, the rate at which paradigm-shifting technology is entering the market is unlike anything we’ve encountered before. It’s already a given that education systems have been unable to keep pace with the rate of development, resulting in fresh graduates who mostly lack the key skills that the market demands, but the new challenge is that businesses themselves are going to have a hard time adapting to this too.

Of course, the nature of business and capitalism is such that there will be a natural, and painful, adjustment, where a large number of businesses will adapt to address the market, and there will be a new wave of businesses entering. However, one of the key losers in this will be knowledge workers, who will suddenly find that they’re generally not in demand anymore. For knowledge workers, if they don’t pivot quickly and learn how to work with agentic AI systems, they’re going to have a hard time justifying their cost in an organisation.

Whatever principles you may have in terms of employee support, the market itself will demand lower prices and faster turnaround times. Right now it’s obvious that with AI, the cost of producing work will go down, and the turnaround time will shorten. What is unclear is where the cost and time savings will be captured. But that’s for businesses to wrangle over. The pressure will be on employees to justify why they need to be paid such salaries relative to the cost of AI. In terms of productivity, with a few more turns, a single OpenClaw hooked up with a decent AI model could produce the output of a whole group of average employees at a fraction of the price.

The cost of tokens is dropping so quickly, it’s hard to ignore. If we take tokens as units of intelligence, we’re realistically only going to need so much intelligence, and the cost of it is going to become increasingly affordable too. Humans just won’t be able to compete with that on price, speed, or scale. The rate of humans acquiring and generating intelligence pales in comparison to the rate of token generation by even simple AI models.

I’ve been stressing to everybody I meet that AI is quickly eating up employment opportunities vertically from the bottom up, starting from interns, juniors, and then upward. The challenge is that it’s eating up these roles faster than any human can hope to progress up the ranks.

In terms of response time, responding with deliverables in a few days will no longer be acceptable once the market gets used to agents working at multiples of human speed and not having to sleep.

Perhaps there will be certain roles that will truly never be replaceable by AI, although I doubt there will be many. But even if we assume that is the case, how many of those roles will actually exist in the market, not even taking into account whether you’re trained for them, and not forgetting the fact that the labour pool will be far into surplus relative to the opportunities in the market.

At least until physical robots become mainstream. And if it sounds like they’re still very far away, let’s not forget that specialised robots and humanoid robots are being developed at a frightening pace. We’re already on a pretty clear trajectory with the development of the “brain” for these humanoid robots.

If you need a visual on how important it will be to be equipped like this, just imagine how likely you’d be to hire someone right now who is unable to operate a smartphone and is unwilling to do so.

What can you do at this point? Honestly, aside from being invested in energy infrastructure and AI, I have no idea. But what I am very sure of is that competence with agentic AI bots, whether it’s OpenClaw or any similar technology, will be table stakes. The bare minimum for being employed in the near future will be coming to the table with a team of agentic AI bots to call upon.