> I wonder if the overall trend in tech is that engineers are experiencing less bottom-up autonomy and more top-down controlled environments
I've thought a little about this.
First: I don't believe that you can have bottom-up autonomy, without taking clear responsibility over outcomes.
Previously the main bottleneck in engineering orgs has been figuring out scaling and reliability. So you could own the outcome of "99.99%" and everyone was happy.
But I think the bottlenecks in the orgs have changed. Partly due to increased productivity, and lower barriers to solve previously hard problems (thanks to good old AI), and partly due to shift in focus from maintenance of a product, to feeling forced to think about "which AI-native company/product will eat us next quarter".
Some of this shift is towards product. Like... what does a great product in market x look like given the condition that we now have cheap and good AI models? Should it be different?
These aren't engineering questions... unless you want to dip your toes in the business/management/product side of things (which I 100% think engineers should do!).
But it requires you to go from owning the 99.99% outcome to owning "how do we create a product with y user retention curve and k number of users".
I've thought a little about this. First: I don't believe that you can have bottom-up autonomy, without taking clear responsibility over outcomes.
Previously the main bottleneck in engineering orgs has been figuring out scaling and reliability. So you could own the outcome of "99.99%" and everyone was happy. But I think the bottlenecks in the orgs have changed. Partly due to increased productivity, and lower barriers to solve previously hard problems (thanks to good old AI), and partly due to shift in focus from maintenance of a product, to feeling forced to think about "which AI-native company/product will eat us next quarter". Some of this shift is towards product. Like... what does a great product in market x look like given the condition that we now have cheap and good AI models? Should it be different? These aren't engineering questions... unless you want to dip your toes in the business/management/product side of things (which I 100% think engineers should do!). But it requires you to go from owning the 99.99% outcome to owning "how do we create a product with y user retention curve and k number of users".