Shaking President Obama's hand with "touch feedback" in 2016: https://www.youtube.com/watch?v=itkgmMLi7l4

Eating a taco in 2018: https://www.youtube.com/watch?v=fUjfA78FuZM

Robot arm in 2018: https://www.youtube.com/watch?v=MjFr0rnbT24

Playing Final Fantasy 14 with a BCI in 2019: https://www.youtube.com/watch?v=WjNHkRH0Dus

Non-invasive robot arm control in 2011: https://www.youtube.com/watch?v=8eOSlzDdOpg

Non-invasive robot arm control in 2020: https://www.youtube.com/watch?v=asDwupMbE2I

Speech/voice generation in 2024: https://www.youtube.com/watch?v=v8frSsvwPp4

The technology to do these sorts of things as proof-of-concepts is fairly old. You do not see widespread deployment because brain surgery betas are not a very good idea. There is insufficient evidence the technology is mature or safe enough to support full-scale deployment. A common class of problem being brain scarring on the invasive insertions that reduce efficacy of the implant requiring further damaging brain surgery to remove the implant in a few years.

When you have insufficiently mature technology for deployment you optimize for research. For that, you only need enough to saturate your researchers with data and well-designed tests which is usually achieved with only a small number of units. This is similar to the reason why you only need a few prototype cars even when you are going to make millions of them. If you are not deploying, then you do not need a lot to saturate your design/development process and making a bunch of each half-baked version prior to the final release candidate is a waste of time.

When the technology is minimally adequate, then you scale up. In contrast, deploying middling quantities of proof-of-concept versions as if that "tests" anything is a recipe for a slow-burning disaster. Nobody else is "trying to compete" on who can deploy more because competing on who can deploy more half-baked brain implants would be unethical.