pretty interesting, agent was doing unexpected stuff, turns out next.js setup injects this into agents.md to force up-to-date docs. not sure if genius or whether i should be mad for using my context window without my permission
I'll be surprised if they end up with a good UX around this, more likely they will abandon the system after the year because it wasn't sufficiently used. Just imagine the confusion of any non-German speaker to pick between "Vollpreis" and "Ermรครigt", and who else would use it?
I was so (positively) shocked to see how simple the ATM-ticket was when I first experienced it in the Brussels public transport system a few years ago. They even set this up so you can never spend more than the daily ticket if you swipe multiple times (like most nowadays).
This obviously doesn't have to be the case - you could allow paying on the first screen directly, and if you don't want discounts, there is a second button that says "show ticket options". Personally I think in that case you can just go to one of the higher-complexity terminals.
I think most of what this comes down to is Austria's obsession with discounts, it's incredibly inefficient. Making the 90% path super efficient and easy, and offering still-good solutions for others would be seen as discriminatory.
Most of Europe has a good UX for tap and pay. Vienna once again shows that they donโt understand how to do good ticketing UX. Before tapping you need to select the ticket type -.-
If not in the 90% bucket, the interaction frequency should be much lower, e.g. through monthly or yearly tickets, or a slightly higher-complexity terminal. Their intuition probably: if we don't offer discount tickets, people are going to swamp us with support tickets.
That's why I like self-checkout registers here: they work without me waiting for someone with 4 discount stickers (some of which don't apply, so they haggle), a bonus card for collecting points (why?), and paying with 25 separate 2โฌ meal vouchers with individual scanning
Jane Street just showed the inside of their AI training data center in Texas.
4,032 GPUs. 56 racks. 8,000 km of fiber. liquid cooling running through every server because air cooling can't handle the heat anymore.
but the part that got me was the origin story.
Ron Minsky, who co-heads their technology group. said their first compute cluster was literally six Dell boxes stacked on top of each other at the end of a desk row. they called it "the hive."
the trading systems sat out in the room with the traders because they wanted to be able to unplug them if something went wrong.
at one point, someone vacuuming the office unplugged a live trading system in the middle of the day.
from six Dell boxes and a vacuum cleaner incident to a liquid-cooled GPU data center processing trades in under 100 nanoseconds.
that's a 20-year arc.
have been using cursor for a while again, back to claude code i am quite frustrated when it "spins" on me for like 10 minutes spending 0 tokens, and then i ask what it's doing it says: "oh i haven't started yet, let me confirm something first" - basically happens every prompt
Happy 18th birthday to Sentry! ๐
I never imagined what the project would become- from a dinky Django project to surpassing 200k teams on our cloud this year. ๐
Thanks everyone who has contributed over the years, as a customer, a team member, or a collaborator. โค๏ธ
@mitsuhiko something about these numbers seems off to me, the largest existing data center google can tell me about is in china at 230 acres, surely we donโt just jump from 230 to 40.000?
full agree with the sentiment, but more importantly what an absolute banger move to add these typos into a text about LLM generated writing, 10/10 would get confused again
I analyzed my coding sessions and on the text interactions some words stand out. And well, they also show up on Google Trends as spiking. Oh and so much slop in my Twitter mentions and on GitHub. Thus here are some updated thoughts on all of this. https://t.co/vejDNcpv8g