Willing to bet my career that how we use LLMs in 2027 will look nothing like how we use them in 2026 because of harness churn. My take is: focus on providing value to your company with the tools available today that appear least likely to churn out of existence tomorrow. The more specific and bespoke your harness, the likelier it is it will become obsolete very soon (I.e. the next frontier model release).

Indeed.

It's promising technology, but the tools are far from mature yet.

And as they do mature, the ramp up will decrease and their won't be any particular benefit to being an early adopter. For reasonably bright people, there's essentially no penalty to "missing out" for a while.

As often, the FOMO-afflicted are churning on stuff that just won't matter. Which is fine if they enjoy it, but isn't something the rest of us need to fret over.

Most of the harness related work I've done has been writing better documentation in the repository, and importing existing external documents. This type of stuff is gonna be useful no matter what, and also helps human engineers!

I wish we could convince folks to write docs for human consumption, but docs are docs....