myx1d ago
I'm really appreciating the general discourse around the cognitive exhaustion of managing high-throughput AI workflows. I've been thumping the table for years about the value that you bring as a "knowledge worker" is all about your judgement calls, not your data entry skills, and that the ideal work-day should converge on an hour or two of tough (agonizing, even) and intensive judgement calls which are more meaningful and impactful than the typical week of old-style work. Maybe we're getting there.
Show threadEleanor Saitta
@myx
One of the things I've learned over the past eight years of fractional CSO work is that the judgment call work is useless or impossible without enough time maintaining your working context of the system. The judgement time doesn't provide nearly enough context to perform well without supplemental time, and that knowledge goes stale surprisingly quickly. The real product of a software team is a group of people who understand the system deeply.
Apr 6 at 9:17amWeb
Show threadarclight1d ago
@dymaxion @myx A concrete example: VS Code has moved from a monthly to weekly release schedule. For a codebase as large as it is, how are developers expected to develop and retain an accurate mental model of the system with that high frequency of change? How are users expected to do the same while trying to keep the same sort of mental model of their own work? What compensates for this high frequency of change to keep the system dependable and able to be reasoned about? Who does high change frequency benefit: users, developers, or careerists trying to game their KPIs?
Show thread-dsr- (not a dog)1d ago

@dymaxion @myx

The real product of a software team is a group of people who understand the system deeply.

Seems deeply connected to:

The project isn't the deliverable. The project is the vehicle. The deliverable is the scientist that comes out the other end

from https://ergosphere.blog/posts/the-machines-are-fine/

Sometimes the same ideas appear to everyone at once, because they are the natural result of the shared situations.

The machines are fine. I'm worried about us.

On AI agents, grunt work, and the part of science that isn't replaceable.

Show threadEleanor Saitta1d ago
@dashdsrdash
Very much so, yes
@myx
Show threadHoldMyType1d ago

@dashdsrdash @dymaxion @myx
> We have built an entire evaluation system around counting things that can be counted, and it turns out that what actually matters is the one thing that can't be.

Yes tell that to the world , which hopefully won't change in our lifetimes !