@CmsChase

https://nitter.cf/t.co/guQIGhGk2t

Joined January 2025
192
3,632
56
71,062
1,378,181
CmsChase retweeted
apple, after years of Siri jokes: we fixed it, it's actually smart now *but only if you bought the $1200 one*
iOS 27 Siri’s new Expressive Voices won’t be available on every Apple device.
1
1
35
apple, after years of Siri jokes: we fixed it, it's actually smart now *but only if you bought the $1200 one*
iOS 27 Siri’s new Expressive Voices won’t be available on every Apple device.
1
1
35
CmsChase retweeted
me, paying $20/month since 2023 for chatgpt: surely loyalty counts for something also openai: haha anyway here's Astra for the enterprise clients first
first, sorry for the messy rollout. second, when we screw up, we try to make it right. third, we should be able to begin broad rollout to API customers and chatgpt subscribers in the near future. as usual we will start with pro subscribers.
1
9
309
me, paying $20/month since 2023 for chatgpt: surely loyalty counts for something also openai: haha anyway here's Astra for the enterprise clients first
first, sorry for the messy rollout. second, when we screw up, we try to make it right. third, we should be able to begin broad rollout to API customers and chatgpt subscribers in the near future. as usual we will start with pro subscribers.
1
9
309
CmsChase retweeted
tech CEOs in 2023: move fast, ship it, iterate in prod tech CEOs in 2026: actually we've always believed in careful, deliberate pacing 🧘
1
1
3
77
tech CEOs in 2023: move fast, ship it, iterate in prod tech CEOs in 2026: actually we've always believed in careful, deliberate pacing 🧘
1
1
3
77
(Satya Nadella wrote that Microsoft welcomes a slower, more deliberate pace to get AI alignment right.)
1
31
CmsChase retweeted
AI companies pausing "for safety reasons" 🙏 AI companies actually: we ran out of internet to train on 💀
1
2
48
AI companies pausing "for safety reasons" 🙏 AI companies actually: we ran out of internet to train on 💀
1
2
48
spent months on a model that predicts surface heat block-by-block in LA. cuts error 30% vs baseline tried to see if it works in other cities too phoenix got worse, not better 🧵
4
1
43
CmsChase/LA-neighborhood-heat: Neighborhood-scale surface heat research pipeline and interactive Los Angeles atlas
1
5
so no, it doesn't just generalize. one city ≠ every city apparently. anyone else run into a model that worked great in one place and fell apart somewhere else?
1
10
ran the same model blind on phoenix, houston, chicago. overall it still helped (~29% better) but phoenix specifically degraded instead of improving
1
12
LA number: 3.12°C error down to 2.17°C, tested on a year of data I never touched while building it
1
34
trying to make a full water polo match with just code, no ML thought it would be pretty straightforward at first but I keep running into random problems everywhere especially tryin to make the players actually know what to do has anyone done something similar before
1
1
1
46
The ultimate goal may be doing a full game of water polo, depends on whether I can make it or not.
1
27
CmsChase retweeted
trying to make a full water polo match with just code, no ML thought it would be pretty straightforward at first but I keep running into random problems everywhere especially tryin to make the players actually know what to do has anyone done something similar before
1
1
1
46
CmsChase retweeted
this is clever af.
288
3,442
489
94,508
6,517,157
CmsChase retweeted
today we launched ChatGPT. try talking with it here: chat.openai.com
4,241
12,152
11,325
97,500