the inference costs of the frontier models is too high for consumers to pay and I’m still unsure how the companies survive long enough to bring that cost low enough to make any profit. Currently those costs scale with usage which compounds the issue.

so this is a universal problem? our 19yo has refused to pick out a coat and is “fine” since he has hoodies 😒

In hindsight I'm amazed at the quality of those classed. CS 1 was QBasic on 16Mhz 386 machines in 1999, CS 2 was Pascal. He made it a point to teach concepts and the language was just the tool.

After I graduated they finally got upgraded machines and switched to Java before the classes ended.

This reminds me how lucky I was to get 3 years of CS in HS because it was a pet class from one of my Math teachers.

When he left my HS they ended CS classes. I just happened to catch a window there, that completely put me on the career path I stayed on.

Great conversation and it reminded be of a similar analogy I’ve had in my head for awhile, great players rarely make great coaches. Similarly the best software engineers rarely make good managers. Above average athletes and engineers usually can bridge that gap in a way the best struggle with.

"They'll keep pushing this as they're replacing workers with AI!" no they're not! That isn't happening. They're laying people off because they sunsetted Section 174 of the tax code that let tech companies immediately deduct 100% of R&D if it "helped improve products."

qz.com/tech-layoffs...

Its been over a decade since I have had a company ask for references or been asked to be a reference.

Many places don’t even give references out of a fear of being sued. They’ll only say you worked there and for how long.

One was a hospital where after insurance paid they owe us a small refund. Apparently you have to catch that and explicitly request it yourself.

😒 thanks American healthcare system