What does the future hold?
This is a question that plagued humanity for thousands of years. We all wish we had a crystal ball that was accurate and concise.
In spite of this absence, it does not stop us from dusting off the old relic this time of year and seeing what it tells us. That is what I will try to do in this article.
What will AI do in 2026?
This is, naturally, a very loaded question. Things are moving fast to precise forecasts are impossible. There is really no way to tell what utility AI will provide in another 12 months.
We can see the network effects starting to take place. This is something we will focus upon.
What is happening?
If we look at the research papers, we see a lot of stuff being worked upon. There are many paths that researchers are going down. Ultimately, I believe none of this matters on an individual scale.
For example, Deepseek brought out a paper that potentially is another breakthrough. This company is known for achieving success using smaller parameters compared to the larger model counterparts.
This is a reason to give some credence to what Deepseek is doing. Of course, they are far from the only ones innovating. Each week there are dozens of papers detailing what could be the "next path" AI takes.
So why do I say that none of this matters at the individual level?
My answer lies in the fact that the next big thing is really a totality of smaller breakthroughs. A large portion of the different innovations end up in models such as Claude or Gemini. Eventually, they pull in the hundreds of innovations creating the next advanced model.
Therefore, whether it comes from Deepseek, Anthropic, or some college, the flow is into these different models. There is a lot of copycat going on.
Here is where we see the network effects.
All of this increases utility. Over time, as more is offered, people find use cases for it.
This is why the automation curve is so dangerous.
People still contest the impact this will have on jobs. There is a large sector of knowledgeable people who deny jobs will be obliterated. To me, this is countered by simply pointing to network effects.
This can be summed up as slowly, slowly, quickly.
Network effect in this context deal with utility. It isn't the number of users per se as the utility the models off. As this grows, more is generated, providing the feedback loop.
When it comes to automation, nothing happens until it does. Close is not good enough. However, as things get closer and closer, eventually there is the impact of the network effect. With enough utility, things get automated.
In the economy, this ultimately means jobs.
Will this happen in 2026? I am not going to make that claim. We still could be a couple years away from a major impact. It appears we are in the early stages, with some jobs being lost. That said, we are seeing the proverbial jumping of the gun. Many who were laid off due to AI were rehired.
The technology is simply not there yet.
That might not be the case by the end of 2026. It is safe to say we will be closer than we are now. Breakthroughs will happen advancing things forward. Grok, Gemini, Claude, and Llama all have major upgrades due. The next generation of these models should propel things ahead.
What is the utility? That is still the major question. Models are being upgraded by the applications incorporating the technology still lag. 2026 could be the year we see that change. Grok is being integrated into Tesla. Google is incorporating Gemini in everything it has. Anthropic and OpenAI are partnering with different companies. Llama is open source, available to anyone.
The capability of what people have access to will see a marked improvement by the end of the year.
