Chico Liu🛸

95 posts

Chico Liu🛸 banner
Chico Liu🛸

Chico Liu🛸

@ChicoLiu

product designer, generalist

San Francisco, CA Entrou em Kasım 2020
309 Seguindo62 Seguidores
Bee 🐝
Bee 🐝@bee_human_·
the codex app is out was annoyed at the extra taps and flicks it took to get to codex, so made the shortcut my “app” for codex and chatgpt chats
Bee 🐝 tweet media
English
12
2
118
61.9K
Yanqing
Yanqing@YanqingCheng·
wow, Opus 4.8 is very... argument-happy? it picked a fight with me about my usage of the word "ontology", and when we eventually got back on the same page philosophically, told me to go to bed it's past 11:30 (it's 8:30). and when I told it "hey you actually have a clock?" it started erroring out aggressively. very Sydney Bing, I honestly approve
Yanqing tweet mediaYanqing tweet media
English
109
19
787
127.3K
Chico Liu🛸
Chico Liu🛸@ChicoLiu·
@GregHBurnham the creativity problem is a trap, it’s the same problem as thinking ai aren’t reasoning. don’t be anthropocentric, be galileo
English
0
0
7
2.3K
Greg Burnham
Greg Burnham@GregHBurnham·
If I had one sentence to summarize the state of AI math capabilities: “AI has resolved one problem of major interest to mathematicians, and a growing number of more minor problems.” If I had a second sentence: “The pace of improvement is rapid.” But if I had a third sentence…
English
27
41
887
234.4K
Chico Liu🛸
Chico Liu🛸@ChicoLiu·
@danywander @bchesky this will age like fine milk…almost all the problems you described also apply to plain search algorithms, as long as you’re curating a result you’re always biased. but ai gives vastly better understanding of intent and much more naturally, you need to be more imaginative with ui
English
0
0
1
239
dany
dany@danywander·
if you've ever questioned your presence on x, i'd send you this screenshot. but i agree with @bchesky it sounds easy to say "need a nice cabin within 2h drive" but that's super naive. in marketplaces data mismatch have a price. let me elaborate. a user says: "i need a nice cabin within 2h drive" okay. - 2 hours from where? current location? home? by car or train? - what does "nice" mean? luxury? cozy? cheap but charming? - does cabin mean an actual wooden cabin in the woods or just a house in the countryside? - when? this weekend? flexible dates? 2 adults? family? dog? budget? wifi? instant booking only? as a human we compress intent. but it doesn't work for booking systems. it needs exact constraints. chat makes the problem feel solved because the conversation feels natural. but finding a place to book is not JUST a conversation. the listings themselves are messy. - one host says "cozy" and means small. another says "cozy" and means dark basement. - photos make places look bigger. - listings are incomplete. - locations are hidden before you book. - descriptions are written like marketing copy. in natural way you could say "find me a quiet cabin with sunset views". then the question is - where is "quiet" and "sunset view" stored in database? so the ai guesses from whatever metadata it can find. sometimes it works. often it doesn't. and when wrong answer in chatgpt costs you nothing, on marketplaces mistakes cost money. - wrong cancellation rules. - pets actually not allowed. - listing unavailable. - distance wrong. - hidden fees. - bad check-in assumptions. travel is a transaction and accuracy matters way more than just add an entry on my calendar. travel is visual. people scan photos, prices, maps, ratings, amenities all at once. even if its look complex ui etc. but in chat ui it becomes just a queue: "here's option one." "here's option two." slower than a grid. way slower. and ranking gets weird. if the ai picks 5 listings why those 5? best match? paid placement? hidden bias? safety call? people still not ready to give up on personal control of the outcome. they want to be sure that they've done everything to find that exact place to stay. speed matters too. a good conversational booking needs intent parsing, availability checks, price lookup, policy fetch, ranking, maps, personalization. and many more. nobody likes waiting 10 seconds, over and over, for "thinking…" and this is just an exploration phase. then we have "book it" part. - which one? - what dates? - which card? - who's traveling? - did you accept house rules? chat feels nice until you have to sign or pay. so chat probably doesn't replace forms. it just makes discovery better. i'd bet on chat for intent to start with → filters for custom work and filtering → cards and maps for comparison → normal boring checkout. you can just make a chat as ui when the moat is still the boring stuff - trust, clean inventory and control.
dany tweet media
ben hylak@benhylak

so instead of saying “funky cabins within 2 hour drive” i will have to keep filling out your patient intake form

English
31
15
267
87.5K
Yuu💖
Yuu💖@QuantumTransf·
这是不是典型的程序员会做出来的 UI(( (被骂了,已经在改了
Yuu💖 tweet media
中文
42
0
77
29.9K
Yash Bhardwaj
Yash Bhardwaj@ybhrdwj·
I turned sam altman's texts to mira murati into 2011 style emo teenage heathrob anthem🫶
English
113
199
2.9K
278.8K
Chico Liu🛸
Chico Liu🛸@ChicoLiu·
@royxy 可最后是炼金术士诞生了现代化学…
中文
0
0
0
144
骆逸
骆逸@royxy·
下午和一个程序员聊了一下,觉得当今程序员有热衷制造自己根本不理解的黑箱的趋势。 凡事扔给大模型,找到什么资料都往里扔,逻辑让它自己找,跑出来什么自己也不懂。这根本就是人工智能时代的token炼金术士…
中文
37
5
192
33.7K
Chico Liu🛸
Chico Liu🛸@ChicoLiu·
everytime I refresh the timeline a new open source sota model has dropped
English
0
0
0
10