Lot of engineering is debugging and parsing logs and stack traces. Some investigations require bringing a lot of information into the context window.
Because itβs programmatic, engineers might even kick off a bunch of work in parallel, in the background where they potentially spend just a minute looking at the final output.
Same way that cost for metrics can blow up with cardinality
the 80/20 of the AI capex boom right now is coding agents. they're the real driver of spend, revenue, valuations.
but it's fearful how much people are treating this revenue similar to a diversified & durable revenue base (MSFT)
engineers will flip on a dime for the faster, cheaper option. the only reason this revenue (and revenue growth) maintain is that we stay compute constrained for 5+ years, which is such a hard bet to make given the astronomical valuations and spending that's resulting from this revenue
the 80/20 of the AI capex boom right now is coding agents. they're the real driver of spend, revenue, valuations.
but it's fearful how much people are treating this revenue similar to a diversified & durable revenue base (MSFT)
engineers will flip on a dime for the faster, cheaper option. the only reason this revenue (and revenue growth) maintain is that we stay compute constrained for 5+ years, which is such a hard bet to make given the astronomical valuations and spending that's resulting from this revenue
the problem with tech week and tech events now is that tech is too broad and all encompassing.
folks selling b2b saas to the logistics industry should really only be mingling amongst the logistics industry. they have no business mingling with consumer social founders building the next restaurant reservation app
but tech events can be useful across different industries if they're functional. e.g. your engineers go to a meetup to discuss agentic observability
this probably comes from a desire more to mingle amongst your social peers, rather than potential business partners --> more status seeking than value creation
anyone here try out factory? I assume sequoia convinced all portco's to throw them a contract
if that's you, would love to get a review. are you also trying competitors? or just finding general claude code / codex works well enough?
@BeingPractical I used to think that but I'm really coming around to appreciating the need for getting views/clicks on personal posts on LI/twitter
it's a ~free distribution channel. how else do you get so many views from your target market at such a low cost
@kevinakwok that's a huge stretch -- claiming that PG left for the same reasons that thiel did.
feel like PG just went to raise his kids away from the noise and try to send them to eton
@bernhardsson the problem is that measuring output is generally hard. as soon as you gain consensus on a metric, it will be gamed
the group of peers closest to them will only be able to tell how much value they're adding relative to everyone else
toronto has the talent but simply does not have the same grind culture as NY/SF. the grind culture only comes from a deep belief of the potential of massively leveling up from their baseline economic prospects.
ask the american employees at faire what they're seeing from their majority-canada-based engineering team
someone should test out paying SF level total compensation for toronto talent and firing fast if their output is lower than the median of the SF counterparts
@pitdesi@typesfaster your blood sugar is tanking which causes adrenaline to spike and wakes you up.
have a spoon full of almond butter before going to bed