
Leila Clark
@leilavclark
Followers
1K
Following
4K
Media
127
Statuses
963
expert data plumber, occasional data predictor longer thoughts here: https://t.co/v85nsumnGp
Joined January 2020
Been really enjoying my recent workday structure: .- calls & interrupts in the morning.- focused coding at night. there's a lot of wisdom in this @paulg quote
2
0
34
@ruth_hook_ I think itās like lighting money on fire the same way that eating out is lighting money on fire? And where homeownership = buying groceries and cooking. You always have to pay to live somewhere ā homeowners do so in maintenance + taxes, renters have it folded up into ārentā.
4
15
543
new theory: @AnthropicAI names models after the coffee beans at the Postscript near the office
9
3
480
@dissproportion I think she's fairly mainstream! She was a huge hit in Japan before the US! My Japanese mother bought her books and loves the philosophy. And I've seen copies on display in bookstores in Japan.
1
1
346
@dieworkwear maybe if youāre a sleepy introvert itās basically a very important work dinner?.
1
1
183
@Joel_P_Atkinson These arenāt the reasons my American friends cite. The most common reason I hear is āIām afraid to go to a country that isnāt a free democracy.ā. An unspoken reason is lack of cachet - people donāt aspire to a Chinese way of life the way they do French, Japanese, even Thai.
22
4
161
@otis_reid someone wrote a story about it once. the protagonist is named Noah or something like that.
4
0
79
@Aella_Girl the older you get the more politics becomes real and actually about peopleās identities. Itās one thing to talk about communism as two college students, another when one is a starving grad student/artist and the other is a millionaire capitalist.
6
1
62
One thing I noticed in the most recent @ycombinator batch is how many of the fastest-growing companies sold AI agents. Theyāre going to be huge. Jaredās tweet thread is a phenomenal breakdown of how theyāre made, and what sets the best ones apart.
CaseText is one of the first vertical AI agents to be deployed at scale. It's an AI legal analyst used by thousands of lawyers. Oh, and it was bought for $650M just 2 months after launch. Here's @Jacob_Heller's playbook for building vertical AI agents that actually work:.
4
9
64
@Mjreard Honestly, Iām pretty sympathetic to ālook, EA is just this thing my husband is really into.ā.
2
0
59
My fun and irrelevant @geoffreyhinton take of the day is that we're distantly related š. In 1938, my great-grandfather Colin Clark moved from the UK to Australia, which is where most of my family remains today. His sister, Margaret, married Howard E. Hinton and stayed in the UK,
4
1
54
the most fucked up thing Iāve learned today is that the modern Neopets economy is composed of working women who slave away for years to buy rare items from an insanely wealthy male arbitrageur elite (h/t @tracewoodgrains)
3
0
47
This app is *so* good. It is probably the closest thing to an āai friendā Iāve tried so far. Itās been a friendly shoulder to lean on, a health & wellness coach and an excellent bouncing board for important decisions. I highly recommend giving it a try!.
Announcing @elysian_labs first product today: Auren!. Auren is a paradigm shift in human/AI interaction with a goal to improve the lives of both humans and AI. Here's a clip of what our iOS app is like and a thread on why this app is so important: š§µ
4
1
40
this is a lovely reflection on the wisdom one gains while growing older. I agree with much of it - my late twenties have been far better than my early twenties. When I was younger I was suspicious of most older people, especially in work settings. It felt like many of them had.
I'm terrified of old people. I used to be extremely confident in myself. I was barely 20 years old and I would tell people how to sleep [1], how to make friends [2], and how to live their lives [3]. I started a nonprofit aiming to literally rebuild the institutions of science.
0
1
32
I also noticed this part of the Cowen/Haidt podcast (summarized excellently by @thezvi). At least for my generation of internet children, HPMOR/Worm/LessWrong/SSC posts did this ā I can tell when someone spent their childhood consuming this. What will it be for the next
3
2
33
When I first read @benskuhn ās āStaring into the abyss as a core life skillā I didnāt really get it. I reread it recently. Now I do!. Before, Iād make poor choices. Iād think about them rarely, then feel trapped and depressed. Now? I still make poor choices. But I face up to
0
0
29
@tailwiinder I think 2019 was the last good year to graduate into SWE. 2020+ had jobs pulled/internships cancelled, and now weāre in the AI era and nobody wants to hire juniors.
0
0
27
@slimjimmy_dev this is true of LLM abstractions too!. if you want to use Claude, use Claude. If you want to use R1, use R1. Theyāre different models with different capabilities, and you should understand them deeply to make the most of them.
3
2
26
@kipperrii honestly even for college-trained software engineers taking operating systems is a useful positive signal imo.
1
0
23
@chandelog The visual feedback was huge & you can see it took me a long time to ramp up. But once I had an objective to measure I could start tweaking things ('oh I focus better if I switch workspaces every day,' 'oh, I am better on concentrating on writing if I do it at night', etc.).
1
1
22
anything Ricki runs is worth going to š.
san francisco house parties full of sober people got you down? wish they were an immersive markets-themed costume party in an enchanted forest? . TRADEGALA is tonight. there's a puzzle hunt, live trading, a surprise musical guest, and vibes. costumes are mandatory.
0
2
24
@matsonj I started playing Stardew Valley again and then realized it is female-coded Factorio. So that's a live option too.
2
1
24
@demiurgently not a friend but I did something similar. imo, if you can, save a large amount in NY while itās easy and then move to the Bay. Cash is relatively scarce here as people largely get paid in other things. So it can go a long way here. But if you run low itās harder to replenish. I.
0
0
23
@GrantSlatton Interesting, I think my approach differs a lot!. I tend to want to do the āhardest partā first ā eg if the goal is to implement against an external api, Iāll write the hackiest possible proof of concept to make sure I understand it. In this example it would be some command line.
1
0
23
@PradyuPrasad I will say they are walking the walk here - from what I hear the bar for hiring has gone up a *lot.* People who would have gotten jobs a year ago wonāt now.
0
0
23
"fundamentally, the reason you're miscalibrated is because you think business is about making money" - @tmychow, to me.
3
0
23
@deedydas to be fair, Claude Code has had access to 3.7 for a while b/c Anthropic was using it internally and it built for it. Whereas the Cursor team hasnāt had any time to update their scaffolding to adjust to 3.7ās quirks. Itāll be a more fair comparison in a few weeks. I still prefer.
2
0
22
great write up on scaling by Nolen!!. also he is being modest about his prior experience lol he was one of the best devs at Jane Street and built multiple key systems from scratch.
alright, we're well north of 100 million boxes checked. today has been smoother. We just passed yesterday's check count :). Some more thoughts on scaling this up (I'll try to respond to questions as I can!). The biggest upgrades and bugs have felt obvious in retrospect:.
1
0
22
@pragdua honestly this is good advice for building a good relationship with anyone - colleagues, significant other. e.g. so important to me to be in an office environment where I can come in and say, 'Good morning!' and leave and say 'Good night!' to friendly folks.
1
0
21
@allgarbled sometimes this sort of impersonal, technical work is extraordinarily good for getting your mind off of personal grief.
0
0
21
@nosilverv reminds me of a conversation I had with a male friend once. him: āwho buys stuff from Facebook ads? have you?ā.me: āyeah wait what? all the time!ā.him: ā??ā.me: ālike handbags and dresses and makeup⦠insta is where you find the cool up and coming brands now, before they go badā.
0
0
20
@uncatherio Once I asked a senior exec who was known for being a āreally good managerā what his secret was. āI taught at childrenās summer camps a lot when I was a teen,ā is what he said.
1
2
19
@ShakeelHashim I'm so upset that "PDF in the style of an academic paper" is how we share ideas now.
1
0
19
@GregFeingold @AnthropicAI You know itās a cult when they give all new joiners little r̶e̶d̶ blue book.
0
0
18
I was pondering this morning how OpenAI's $6.5b raise is pretty incredible -- and also less than California has already spent on a high-speed train to nowhere. It's actually a little crazy that government capital is so misallocated in the US. Not sure why it happens.
I didn't realize how quickly China could scale towards AGI if they choose to centralize. "China can have a bigger model than any of the labs by next year". The US as a whole gets more GPUs than China, but they're split over half a dozen different labs. If Xi got scaling
0
1
17
this one is a neat trick for parties. if you stand by the wall with your arms crossed you will feel like a wallflower. if you stand confidently in the center of the room and relax, you will have a good time. works surprisingly well!.
posture is a big one for me lately; are my arms folded? do i have one foot turned away like i'm about to run, am i fidgeting? or am i relaxed and open?. this one is cool bc shifting posture to look more relaxed when you realize you're feeling anxious will help reduce the anxiety.
0
0
17
@dwarkesh_sp At some point I realized it was useful to write up my ideas out of something. almost like spite. "you won't give me the time to explain this in conversation so I'm going to write it out!".
0
0
16
@ankurnagpal These private returns are *way* better than I expected. Ofc I notice itās the startup investing that did well - I guess if you sell a startup you do get good deal flow, which is the crucial part. (In general I ~mostly expect private investing to return zero or even negative. But
1
0
7
@Fissure256 @otis_reid I think youāre right that this canāt actually have inspired the flood myths. Wikipedia says it happened 5m years ago, and weāve only been around ~300k years. Sad ā it would be a nice meme if true!.
0
0
14
was reflecting recently on how much more Iāve learned from actually futzing around with LLMs compared to āhanging around at SF partiesā (or even reading blog posts). no replacement for actually just getting into the arena and trying things. (this tweet is actually about deepseek).
used to watch a lot of youtube explainer vids, but have since cut back a ton. they mostly give the illusion of knowledge. feels better to just be ignorant or commit to actually learning.
1
0
15
@DolphinMossad tbh I had this reaction yesterday when I went to Home Depot for the first time. And also when I went to Costco for the first time a month ago. (Iāve lived in the states for 5+ years, but mostly in NYC).
0
0
13
I wish something like @atlasfellow still existed, tho less expensive/selective. I know at least one teenager who would benefit from being around other nerdy teenagers at a summer camp run by college students. (if you would do this except that you need ~small funding, lmk!).
6
0
14
@daltonc Man, I knew Slashdot was old, but I didnāt realize it was *that* old. (Though I do remember discovering a cool new forum called Hacker News on it š).
0
1
14
@robkhenderson @antoniogm oh wow I just learned about this (!). I had previously noticed a disconnect between how well he writes and the quality of his ideas. I thought it was weird. I guess this explains it. A real shame because he clearly has taste in ideas - if he attributed it would be much better.
0
0
12
@ByrneHobart in both college and business, he climbed to the top of his local social environment!.
1
0
13
@Koschmar @Aella_Girl Thanks! Yeah Iāve generally been a fan of and am pretty aware of Aellaās background (and now consider her well in the wealthy capitalist groupā¦). Maybe I came off negative? Sorry. I was really just riffing. š¤·š»āāļø.
0
0
12
@michael_nielsen My college friends moved a 30-person group chat to Discord because we had too many side chats on different topics. Itās really helpful for organizing that.
1
0
13
same goes for code, fwiw. it is usually a mistake to open a fully-written and tested PR for review if you haven't talked about the high-level layout with the reviewer. better to have an initial PR with just the data schema and high-level function outline and aggressive stubbing.
some of the best writing advice I got came from a McKinsey consultant:. if you're looking for feedback on a piece, share it at the outline phase and not a finished draft. much easier to iterate on layout/ideas with others, and *way* less painful to cut and re-arrange things (to.
1
0
13
@mattshumer_ My go-to line now is: āmake this as simple as possible and DONāT SWALLOW EXCEPTIONS.ā.
0
0
9
@alexeyguzey I have found that my advice posts are a representation of the very best person I could be, not who I am day-to-day. The nice side-effect is that this means I can find my own advice posts useful too!.
0
0
12
@PradyuPrasad the simple answer is:. Opioids = bad.Psychedelics = not bad. Ulbricht gets glossed as psychedelics drug crowd not opioid crowd (correctly or not). this is the part of drug policy East Asia gets very wrong btw. Iām not gonna say psychedelics are *good* but they are far less.
1
0
12
@Jeffinatorator Property taxes would also fix this! Though maybe getting rid of Prop 13 is more of a pipe dream than a land value tax š.
0
0
10
This is actually really nice offer letter software from the guys at @CentralHQX! . I've helped many friends through confusing equity calculations, so I've definitely felt the pain. There's a slider that lets you see how much your equity is worth in different valuation
3
3
11
learned tonight that there are upper middle class professionals out there who follow @tylercowen *and* have no idea what @stripe is. (yes, yes, I need to get out of my bubble more. theyāre just so intertwined to me!).
0
0
11
@FT The actionable part for most people is this:. you can put your passive investing funds into an all-world index or just US stocks (the S&P 500). Right now, it's *way* better to be long US. This is a macro bet, but I think it'll be true for a long time.
3
3
10