We're already seeing a dramatic slowdown in relative improvements.
The gap between Sonnet 3.5 (June 2024) and Opus 4.6 (Feb 2026) is large, but it's not 1,000x. Not even 100x.
LLMs struggle with novel technical issues.
They still require review and agents work best when they're hand-in-hand with an experienced human.
Let me ask you something: think of the average VP or Director you've interacted with. Do you think this person is capable of directing 100,000 autonomous coding agents? Do you think they have the verbosity in prompt and the skill to know when it's not making a subtle error?
The problem begins people see this as a know everything magic orb and trust it blindly for everything. It's still a pattern matching model, it's not sentient. It should remain a tool rather than one you should be asking for decisions.
Also I've seen people waste massive amount of tokens to add two tabs to a line of code to fix indents. Said they don't want to click the damn line. Bro you just typed a larger prompt, sent the complete file, instead of two keystrokes. And guess what it took multiple attempts. It's like watching someone type google into google 3 times before typing what they want.
Not all software are simple crud from your standard consulting business which makes more money the quicker something is finished. Some software runs critical life threatening infra everywhere. We need people who have the skills to build these, and they're discouraged from the school level not to thanks to AI bros.
Like when Trump tells Canada (paraphrasing) "You think we're going to let China eat you first and just have the scraps?"
Everything is a zero-sum game to them. They are philosophically void.
Then the models will put you out of work. Nobody will need you.
We'll have a world full of largely useless humans.
We're almost there.
Forgive me if I was off by a few years.
Writing code by hand is not going to be the default mode going forward. You either do the majority of your work controlling autonomous agents and reviewing their work or you get surpassed by all of your colleagues.
Are you going to be the farmer who refuses to buy a plow?
I also do not have sympathy for those who refuse to adapt. These people hold back organizations by appealing to tradition and resisting any form of change.