June 28 2026
Ok, not exactly “Siri” but a local LLM that backs Siri will, or whatever local assistant is installed on the device you received or bought.
Here are my predictions on how we get there.
AI will reach a good enough threshold (if we have not already). You won’t really check the output and much of the hand holding we do today will be codified behind the scenes. It will get things right so often, that you stop checking on it just like you don’t necessarily nitpick trusted coworkers and sometimes do them a favor by “reviewing” there work quickly.
Timeline: We are almost here, if not already. Its not that the technology has advanced so far that we should stop checking, but that people already have. Lets face it, most people don’t care that much. If they did, you wouldn’t see major corporations leveraging LLMs for customer support, press releases, and product announcements. These are external facing outputs that will be tied directly back to your reputation! Imagine the slop they are producing internally!
Most people won’t be able to tell the difference between model outputs. I would bet money (ok not a lot but at least some) that today, if you had to match outputs to models for the same prompt, people wouldn’t be very good at it. There might be some key characteristics still today, but if you take away prose (“You’re absolutely right!”) then you end up with indistinguishable results.
Additionally, I bet you could start from Opus 4.5 all the way to Fable and everything in between, people would not be able to accurately relate outputs to specific models.
Timeline: This will happen in a year. We will probably find out giving a “personality” to any LLM will have impacts on benchmarks. To maximize benchmark scores for bragging right we will converge as a similar or the same personality for every LLM.
Because models will be “good enough” and completely interchangeable, most people will default to whats easiest. Because the model on your computer will be sufficient for almost every task, it will become the default for consumers. People won’t see the point to pay extra per month or just be too lazy to sign up. They might have opinions on which model they want, so they will use the dropdown in settings to select their favorite one.
But Thomas, most of the industry will be enterprise!
Sure but most of all enterprise gets away with good enough tools already. We are spoiled as engineers to have access to multi-thousand dollar MacBook Pros by working for slick tech companies, but outside of technology is this the case? My guess is most people get a Lenovo or something worse. You don’t need the bleeding edge to edit word documents or creates slides in powerpoint. Most of enterprise just won’t require the bleeding edge for day to day LLM usage, they will get their emails summarized by Siri.
But Thomas, I work in a super special field that needs the bleeding edge.
Companies large enough in super special fields will deploy an LLM just like they do PostgreSQL, or outsource it to someone like Amazon who will be a provider.
Timeline: With some much investment behind the current companies and their intertwining with the US government this one might take awhile. Consumers will cut over way before enterprise, we are just waiting for advances in the models to be small, good, and fast. Siri is already slated to roll out with Gemini backing it. Once an open source model small enough to fit on phones that achieves the good enough threshold you’ll see something different. Consumer will take 5 years, enterprise will take 10. The hype and psychosis is too gnarly at the moment. We need reality to set in a bit more for these executives to realize its not the second coming and they wont be able to fire their entire staff.