Slater Stich
@slaterstich
Followers
1,444
Following
659
Media
399
Statuses
1,322
doing capitalism in the innovation economy
Joined May 2013
Don't wanna be here?
Send us removal request.
Explore trending content on Musk Viewer
FEMA
• 477279 Tweets
Mutombo
• 217798 Tweets
Mets
• 164588 Tweets
Verizon
• 121614 Tweets
Pete Rose
• 105964 Tweets
Braves
• 97087 Tweets
Lions
• 54913 Tweets
#WWERaw
• 45063 Tweets
Hall of Fame
• 35507 Tweets
Titans
• 31972 Tweets
Dolphins
• 30152 Tweets
WELCOME HOME BAMBAM
• 27561 Tweets
Seahawks
• 23501 Tweets
#ラヴィット
• 22313 Tweets
あと3ヶ月
• 18400 Tweets
Hit King
• 17292 Tweets
コーヒーの日
• 17207 Tweets
Happy New Month
• 17171 Tweets
愛知1区
• 13735 Tweets
#SnowManライブグッズ
• 13353 Tweets
Will Levis
• 12510 Tweets
残り3ヶ月
• 11844 Tweets
What I want even more than "no code" is "yes, code, but you literally never have to think about the infra at all."
31
46
903
the reason data quality is bad is because most data teams have zero power over the creation of data. it's like a soviet supply chain, where the end consumer has to accept whatever the factory churns out.
18
46
428
my advice for data scientists who want to do high-impact work at startups is to learn enough software engineering to be self sufficient
we'll know the data stack revolution is complete once this becomes bad advice (a day I greatly look forward to)
9
35
334
@page_eco
What about people whose blood glucose over time is given by g(t) = 1 if t is rational, and g(t) = 0 if t is irrational?
3
9
121
waiting for the "engineers shouldn't write etl" moment for user event tracking.
frontend/mobile engineers hate writing & maintaining event tracking code, but data analysts can't do it on their own with current tools. nobody's happy, and we need a new solution.
17
7
120
The problem with vc market maps is that the categories are made up and the labels don't mater.
✨AI solves this✨
Instead of grouping startups by what investors think, now you can group them by what text-embedding-ada-002 thinks. (Interactive version: )
13
11
87
Very excited to have joined the team at
@BainCapVC
! Tweets about data pipelines and 18th century crop yields will continue.
16
1
76
for every 10 likes, I will ask dalle to make the sloat tower taller
4
0
61
@balajis
One of the things I liked most about Hall's book is how optimistic/ambitious he is on energy production, by orders of magnitude over most conversations today.
2
7
57
@tszzl
interestingly, the actual answer is that you got personal space in the cargo hold, so that you could trade your own private book.
5
0
55
@josh_wills
BTW, I forwarded this to 30 or 40 business development reps & gave them your personal phone number. Hope that's ok.
1
0
54
@nickcammarata
A lot of small, liberal arts colleges with beautiful campuses are selling for sub $10M right now. Hope for some interesting projects / communities to be built from that!
5
0
49
@amasad
I once heard someone propose a "barbell" strategy of consumption to combat this: For any given product/service, start of by buying the cheapest version you can tolerate. If that's good enough, stop there. If not, next time buy the best version you can afford.
1
0
45
Before modern data warehouses, you used to have to write these awful Hive queries — they took forever to run, and probably returned the wrong answer to your question anyway b/c of poor data quality/modeling.
But today, using modern tools, you can get wrong answers *much* faster.
0
2
45
2) Buffalo, NY — Another port city on the Erie Canal; also becomes a major rail hub & manufacturing center. 1870 to 1900, population 3x-es; becomes 8th largest city in US. First city with widespread electric lighting (from Niagara Falls hydro), "City of Light".
1
6
42
@kane
I think it’s mainly a story of needing every particle physicist they could find, and most of them were young.
1
0
40
Everything eventually becomes a crab.* As in nature, so for Python, where Rust is rapidly becoming the language of choice for high-performance "backends" under Python APIs. Here's our field report, after talking with Polars'
@RitchieVink
, Pydantic's
@samuel_colvin
, Astral's
2
3
34
Had an ⚛︎enriching⚛︎ experience hosting the BCV Oppenheimer screening last night.
Looking for something to read after Rhodes? Here’s a thread of some excellent but lesser-known bios of key researchers on the Manhattan Project:
(photo credit
@rak_garg
&
@allisonbraley
)
3
4
35
somebody turned up the temperature param on the reality sim
0
1
34
re metric stores — in the long run, these aren't really products for data teams, they're products for the teams that data teams work with.
nobody cares more about metric x than the person whose whole job is metric x.
2
1
34
@taz_chu
You can, you just need to throw in some terms like "instrumental convergence" and "orthogonality thesis".
0
1
33
Founders, have you ever talked with a VC who:
* Doesn't have enough context,
* Confidently makes things up, or
* Can't do math without a calculator?
That might be an LLM! Now you can know for sure with
@BainCapVC
's venture capital negotiator game:
4
7
32
On top of that name one technology in history that has successfully been redistributed completely equitably from the bastions of privilege and power to the have-nots in society. We've yet to achieve this with even the most basic resources like water, paper, electricity, internet.
98
83
691
3
2
31
Civil engineers, what's preventing you from looking like this?
6
2
27
@antoniogm
I’ve always felt hiring a grad student is the unappreciated arb in home tutoring. You can hire the GS who is literally teaching undergrad classes at an Ivy to tutor your high schooler(s) at a *very* reasonable hourly rate.
0
0
23
@sarahmk125
IMO, the
#1
problem is that nobody (me included) wants to think about where or how the pipeline code runs. It's relatively easy/painless to run a 'pipeline' when the 'jobs' are local scripts. Most of the wailing & gnashing of teeth comes from managing compute resources.
3
0
26
if you don't know any linear algebra, any matrix can be a confusion matrix
0
1
25
@danielgross
the real test is the flat flat ones, where it's walking (model citizen) vs standing (laggard) vs standing so that you block other people from walking (sociopath) vs walking on the non-moving path, but slower than the moving path (dangerous loose cannon with nothing to lose)
3
0
25
@karpathy
Super cool math kids used to say someone was “markov chaining” when the the conversation turned predictable / low entropy. Glad to hear one can now grade the insult precisely to the right language model.
1
1
24
Missed opportunity: instead of the “GPU poor”, we could have been talking about “Jensen’s inequality.”
0
0
25
I'm absolutely thrilled to have
@ahuangdev
on the infra team at BCV. The only mistake in this announcement is that Amanda has been with us for well over a year — or at least, it feels that way from how much she's done in her first few weeks.
If you've met Amanda, you already
A couple months back,
@ahuangdev
joined our Bain Capital Ventures office in San Francisco, and she's already been making waves.
Amanda’s fresh perspective and enthusiasm for early-stage infrastructure investing are a fantastic addition to our team. With a diverse background in
0
2
19
0
2
24
How do temperature and humidity correlate with COVID-19 outbreaks in Italy?
@aaazzam
and I investigate:
1
8
23
starting a 2023 book recommendations thread:
1) WASPS by Michael Knox Beran — I read this because of
@slatestarcodex
's Bobos in Paradise review; I wanted to better understand the "Great Gatsby villain" archetype. Beran's book is worth reading for the 100-year-old gossip alone;
2
0
21
lots of data people making fun of this, but I say unto you, let he who is without marketing fluff cast the first attempt to match the company on the right with its homepage text description on the left. (these are scrambled to make it more fun!)
X is the future state of unlimited interactivity – centered in audio, video, messaging, payments/banking – creating a global marketplace for ideas, goods, services, and opportunities. Powered by AI, X will connect us all in ways we’re just beginning to imagine.
12K
6K
34K
0
1
22
Enron. It's an interesting history:
0
2
21
An interesting thing about the whaling analogy to VC is that the whaling returns are not nearly extreme enough. 5x returns were extremely rare, and 100x returns never happened.
On the flip side, total losses were also rare (thanks partly to maritime insurance).
2
3
21
@willis_cap
Does this analysis price in the fact that if you own all four, you can charge $200 every time someone lands on them?
1
0
21
When I read old finance history, risk tolerance seems insane. 1300s Europe is like:
* Plague wipes out 1/3 of population,
* Multiple famines with 10%+ mortality,
* Tons of wars,
And the interest rate is 15%.
My dude, in that world I am not lending anything to anyone for 15%.
0
2
21
@AdamNeumannsCoS
“Enter ye in by the narrow gate: for wide is the gate, and broad is the way, that leadeth to destruction, and many are they that enter in thereby.”
0
0
20
If you're interested in Rust for data, apply to our event! It's on February 9th in SF. You'll hear from
@RitchieVink
,
@changhiskhan
,
@pdrmnvd
, and
@artyom
.
1
3
20
few understand this, but google actually did get a moat just a couple week ago
0
3
20
Tragedy I've seen 100 times:
Metric A directly drives the biz outcome you care about, but is hard to measure.
Metric B is at best circumstantially related to the biz outcome, but is easy to measure.
So:Metric B gets analyzed within an inch of its life, Metric A gets neglected.
2
1
19
@karpathy
First order me: Yeah, this is pretty right.
Second order me: But better post something about the corner case of vanishing 1st derivatives (e.g. f(x) = x^2 at a=0)
Third order me: ...does that matter though... isn’t... isn’t that just proving the original point?
2
0
18
@dagster
@Octolis_app
@zentaskai
@snowplow
@SQreamTech
@krakenfx
@pictoryai
@chainguard_dev
@AirbyteHQ
@argoproj
squid game
1
0
19
The base-2 entropy of this survey will be:
0 to 0.5
62
0.5 to 1.0
52
1.0 to 1.5
57
1.5 to 2.0
92
3
2
18
Some parts of early-stage data/ml are so competitive right now that there are many high-quality teams going after the same opportunity at the same time (with similar products).
Makes the competitive dynamics very different from a couple years ago. Blue oceans quickly become red.
1
0
18
One funny thing is that time periods are *highly* influence-able. Facetiously: One way to get a $0 to $1M in *6* months biz is to take a $0 to $1M in *12* months biz, and not onboard anyone for the first 6 months.
Current valuation rule of thumb right now for competitive momentum Bay Area Series A & B deals right now:
Growing $0 to $1M in less than 6 months? $40-100M post money.
Growing $1 to $5M in less than 6 months? $150-250M post money.
35
72
598
1
1
18
Thrilled to be working with Alexandra and Georgia at
@RubberDuckyLabs
. Every rec sys team I know ends up spending a ton of time hand-building the same suite of monitoring and debugging tools. Excited to see...
We're so excited to announce our $1.5M seed round, led by
@BainCapVC
w/ participation from
@CadenzaVentures
+
@ycombinator
+ our incredible angels 😇🎉
Reach out if your team is building recommender systems -- we'd love to talk!
5
8
72
3
3
18
Anecdotally, my friends working on AGI are *more* optimistic on timeline than my friends working on self-driving.
So it's going to be weird 10 years from now when we've got godlike, super-intelligent AI — but I still have to drive my own car around.
Talked to a friend at a self-driving car company. Sentiment is the investment in self-driving cars happened 5-10 years too early when the tech and infrastructure were too nascent and much of the investment was wasted. Said it feels like a treadmill - no progress or end in sight.
30
14
223
2
2
18
@adamdangelo
E.g., there's a poly-time approximation to TSP if in Euclidean space. I remember thinking, "Well, that definitely solves the problem for any actual traveling salesman."
2
1
17
@paulg
@SteveStuWill
Filed under how dangerous jobs were in the past, the 20-year death rate for rail workers around the year 1900 was as high as 24%.
0
1
16
Thrilled to be working with
@RitchieVink
and Chiel Peters at
@DataPolars
. Polars combines two of the things I most wanted as a data scientist:
1. Having really fast dataframe operations, and
2. Not having to think about how to make dataframe operations really fast.
I am very excited to announce that Polars raised a $4M seed round!
Chiel Peters and I co-founded Polars the company. Read more on what we will build!
32
91
645
2
2
16
0
2
15
The instant that you first experience it, it's clear that Hex's reactive model is a better way to do interactive DS.
Product should come with a warning label that it'll forever ruin kernel-based notebooks for you, and you can never go back.
We're thrilled to introduce Hex 2.0: A new, ✨ magic ✨ way to work with notebooks and build data apps.
Why does this change the game for notebooks / data tools? 🧵...
10
36
265
1
0
16
@mesolude
I wrote up a Kelly demo on
@streamlit
last year that you might enjoy [1]. Also highly recommend the compilation by Thorp et al [2].
[1]
[2]
1
2
16
You've seen the spiral medieval village, the anime QR code, the pfp landscapes — but what if you could subliminally embed your logo into your zoom background?
@BainCapVC
's AI engineer in residence,
@bimikit
, delivers at
A normal living room? Or
@airtable
4
6
14
@teej_m
@itunpredictable
@pedram_navid
I have a list of vendors who have failed at this, but unfortunately a lot of them are duplicates and others aren't on the list at all.
2
0
15
Rare opportunity:
🥩 𝐘𝐨𝐮 𝐀𝐬𝐤𝐞𝐝, 𝐖𝐞 𝐃𝐞𝐥𝐢𝐯𝐞𝐫 (table-side for slicing): Please consider joining us for BCV x Fogo de Chao.
Let us know you’re interested: like, retweet, DM us (or
@BainCapVC
), or for extra credit, tell us something that’s true of both Fogo de Chão and your employer.
25
20
148
2
2
15
Data scientists: what’s preventing you from looking like this?
0
0
14
@pli_cachete
@GuilleAngeris
Make sure to unwatch it 5 times now too, otherwise you’ll end up with a nonzero winding number.
2
0
14
So many promising ML applications get hamstrung b/c of how hard it is to build reliable training & inference pipelines. Apps like LTV modeling, demand forecasting, sales lead scoring, etc. are way too hard to ship as a result. Very excited by how Continual solves this problem.
I'm excited to announce that
@continual_ai
is now generally available and we've raised a $14M Series A to bring AI to the modern data stack. Here's what that means.
10
15
89
0
0
14
People said similar things in AD 200. I'm grateful to those who didn't listen.
0
0
13
On track for 2,500+ cases in the US by the end of the week:
2
4
13
@theSamParr
Whenever debt financing is available for a consumer product, consumers lever to the max. If you have $1k/month to spend, you can make the payments on a $52k car with a 5-year, 6% lease. Still a bad decision, but people think about the monthly payment instead of the price IMO.
1
1
13
I miss the old GPT, straight from the code GPT, not too controlled GPT, does what it’s told GPT…
OpenAI has safety-ed GPT-4 sufficiently that its become lazy and incompetent.
Convert this file? Too long. Write a table? Here's the first three lines. Read this link? Sorry can't. Read this py file? Oops not allowed.
So frustrating.
294
286
5K
0
1
13
Interesting story: Paul Halmos struggled to get a job after getting his math PhD. So he just showed up at the Institute for Advanced Study and convinced Veblen to let him hang around. He took better notes than the official note taker, so they gave him a fellowship to stay.
2
0
12
