Share this page!

Enter URL or ID to Unroll

You can paste full URL like: https://x.com/threadreaderapp/status/1644127596119195649
or just the ID like: 1644127596119195649

How to get URL link on X (Twitter) App

  1. On the Twitter thread, click on or icon on the bottom
  2. Click again on or Share Via icon
  3. Click on Copy Link to Tweet
  4. Paste it above and click "Unroll Thread"!
  5. More info at Twitter Help
Dan Luu Profile picture
Dan Luu
@danluu
Active on https://t.co/WG71NrsDQk; also trying out https://t.co/DBk2OnBVL1. No longer read replies or notifications here now that tweetdeck is gated.
7 subscribers
Dec 5, 2022 • 6 tweets • 3 min read
I feel like this is true for lots of kinds of conversations and not just tech interviews.

People are correctly pointing out that, if you dig into the logic of basically anything, it falls apart, but that's also generally true of actual humans, even experts. Sure, is ridiculous, but have you tried asking an expert coach on almost any topic why you should do X?

E.g., try listening to one of the top paddling coaches in the world explain *why* (the what is good, the why is nonsense)
Nov 20, 2022 • 4 tweets • 1 min read
What are examples of items/categories where you're really getting your money's worth at the high end, not necessarily in terms of utility, but in terms of the difficulty of producing the item more cheaply?

I find the contrast between these vs. "brand" items fascinating. An example of a category that doesn't qualify but where some items qualify would be high-end fashion, where you're quite often mostly paying for the brand (e.g., an expensive Theory shirt) but there are plenty of items where you're paying for the item (e.g., a $5k Kiton suit).
Nov 16, 2022 • 5 tweets • 2 min read
Lots of people in my mentions saying things like "Elon is cleaning house! Lazy bums are getting what they deserve!", as if Twitter employees are getting a much deserved comeuppance.

Since people don't seem to understand what the bums at are getting, here's a short primer: If you look at the people most responsible for Twitter's state, leadership, they had golden parachutes worth tens of millions of dollars

We can debate whether or not they deserve the money, but if you think someone is a lazy bum, cursing them to receive a $10M+ payout seems odd
Nov 16, 2022 • 19 tweets • 7 min read
One of the things that I think is sad about the decimation of Twitter eng is that Twitter was doing a lot of interesting (and high ROI) engineering work that, at younger companies, is mostly outsourced to "the cloud" or open source projects

A few examples off the top of my head: Twitter is, of course, mostly on prem.

The now gutted HWENG group was so good at designing low power servers that, in a meeting with Intel folks, discussing reference designs vs. what Twitter was doing, the Intel folks couldn't believe the power envelope Twitter achieved.
Nov 14, 2022 • 5 tweets • 2 min read
Nice thread about the misconception that major tech companies run systems that can run without intervention because they're automated

The example comes from Google, which is more automated compared than most major companies (MS, etc.), but still quite manual in an absolute sense t.co/diqwJ3RHZH One thing that's been interesting about recent events is seeing how people imagine big companies operate, e.g., people saying that Twitter is uniquely bad for not having a good cold boot procedure.

Multiple $1T companies didn't or don't have a real cold boot procedure. @ matthew_d_green Pour one out for Twitter’s cold boot pla[QT previous screenshot] @ahidalgosre Alright. Fine. I’m
Nov 14, 2022 • 4 tweets • 2 min read
An interesting thing about this claim is that not only is the implication wrong, Twitter probably has better evidence of its wrongness than any other company in its size class could have. There are very few companies that have a better distributed tracing setup w.r.t. getting actionable insights on the backend and the ones that have a better setup are much larger (Google, FB, etc.)

Twitter client tracing also punches above its weight.

Nov 12, 2022 • 10 tweets • 3 min read
It's amazing how much faster Mastodon is than Twitter, e.g., it takes 22s vs. 86s to load the same thread (mastodon.social/@danluu/109310… / ) on an old cell phone with a mediocre connection in Toronto. Twitter page load benchmark showing 86s LCP, 4.8MB payload wMastodon page load benchmark showing 2s LCP, 2.2MB payload. Mastodon isn't exactly hyper optimized, but Twitter is anti-optimized despite quite a few very good engineers having done serious optimization work to speed up Twitter.

Unfortunately, that was always a 1 step forward 2 steps affair due to leadership's priorities.
Nov 8, 2022 • 40 tweets • 9 min read
Now that this prediction has come to pass, I should explain why I thought this was a reasonable prediction and why a lot of other predictions about what might happen were or are implausible.

Some necessary background for this is the state of Twitter over the past few years: A lot of people are attributing almost all of Twitter's current problems to Elon (and others are attributing almost all the problems to current and past employees), but a lot of what's happening is attributable to past leadership.

Jack misran the company such that extreme
Nov 8, 2022 • 4 tweets • 2 min read
I'm surprised by how much money individuals make from video.

At the bottom of the high end, there are examples like , where the #137th streamer on twitch has been offered $10M/yr multiple times.

More surprising to me are numbers for "normal" streamers, e.g, @TechWithLucy says her 50k sub YT channel pays her more than AWS paid her: .

If you think of the top streamers as analogous to movie stars, maybe those numbers aren't surprising, but a 50k sub channel pulling in AWS eng money seems quite surprising.
Nov 7, 2022 • 7 tweets • 2 min read
The project led by the PM director who bragged about how many hours they work and sleeping in the office hit their deadline and shipped.

The product charges people who sign up but it doesn't actually work. But they technically shipped. The best kind of shipped. I'm curious how adults can really think that skipping sleep adds productivity, even in the short term.

I remember learning this lesson in college. For complex (for college) projects, ~6 hours past my bedtime, I lost about 95% of my velocity and had increased error rate. Screenshot of PM "director" (this is what PM managNat Friedman @natfriedman: This is how great new things are
Oct 31, 2022 • 7 tweets • 2 min read
This reminds me of an "Ironic. He could save others from provisioning, but not himself" story about MS.

There were, of course, datacenter supply chain issues due to Azure's growth, but the most immediately noticeable issues on joining were things like desks, toilet paper, etc. I was in building 43, with most of Azure networking. The building was highly overprovisioned pending Microsoft remodelling the building to convert it to open offices, something that was happening building-by-building.

Meeting rooms had been converted to bullpen offices,
Oct 30, 2022 • 22 tweets • 7 min read
I often marvel at the dynamic range of productivity that's out there.

Once upon a time, a programmer wanted to make an online Dominion implementation, so they did it as a "weekend project".

To this day, it's the nicest interface I've used for any online board game, with terra.snellman.net, another spare-time online board game, being the only thing that's in the same league.

Anyway, the creator of the Dominion board game shut down the unofficial implementation at isotropic and sold the license to a VC funded company that had
Oct 28, 2022 • 14 tweets • 4 min read
This, but for cloud migrations. The quarterly "cloud is cheaper than on-prem when you account for all on-prem costs and everyone who thinks otherwise is clueless" thread is happening again.

Every time, people with serious on-prem experience will point out that's false. The reason these threads always get rebutted is that cloud advocates don't note that a common cause of cloud being superior is organizational dysfunction that prevents the company from running on-prem hardware effectively, which the thread creator incorrectly assumes is universal
Oct 26, 2022 • 6 tweets • 2 min read
I find it odd how often the evidence cited for someone being terrible is fiction.

Another example of this is Borat, which I've even seen cited as evidence of Giuliani's bad behavior and general American idiocy, but the Giuliani scene has clearly been misleadingly edited: time moves backwards due to cuts and things are spliced in the wrong order and/or repeated to make it seem like a movement was repeated.

There's plenty of evidence of bad behavior from Giuliani, so why cite this clearly misleadingly edited scene?
Oct 23, 2022 • 4 tweets • 2 min read
This is a nice illustration of how poor the most upvoted HN/Blind career advice is

The claim is that at FB and FAANG companies, the way to succeed is to do pointless work that juices metrics, but that's directionally opposite of what I tend to hear from very successful people How to succeed at Meta Meta • metamoty Joined about 5 mont How to succeed at Meta (teamblind.com) 245 points by donsu This is highly org-dependent, but I hear a lot more complaints about how metric-moving direct impact work isn't valued at all and "leadership" work is the only thing that matters from, e.g., people like my friend who graduated in 2015 and hit L7 in 2021.

Oct 5, 2022 • 4 tweets • 2 min read
I found it a bit absurd that so many people are passing around screenshots of poor quality scans of printouts of texts relating to the Twitter buyout, so I OCR'd them:

danluu.com/elon-twitter-t…

On reading it all, I'd say that most summaries I've seen are fairly misleading. In many cases, articles seem like they're misleading because the journalist went in with an agenda and pulled out quotes to support the agenda, but in other cases, journalists just don't seem to understand which texts are conveying important information?
Sep 30, 2022 • 8 tweets • 2 min read
Is there a podcast that discusses the tech industry at large and the impact on society where the host understands tech enough to call out "obviously" incorrect statements?

I've been trying out general interest podcasts that talk about tech and, so far, they all miss on this. Three examples from this week:

1. on a discussion about workplaces, the guest mentions that great employers do X and cites Cisco and what a great place Cisco is to work, as evidenced by Cisco winning some "best places to work" ranking four years in a row.
Sep 28, 2022 • 12 tweets • 3 min read
I find it funny that, looking at writing that's "about" productivity, there's overwhelmingly more written about tech/tooling choices than about emotional management (seems like at least 100:1) when emotional management dominates tech choices at typical margins in industry. Per , one of the highest ROI interventions I've seen for interns/junior engs is getting them to understand that it's ok to ask "stupid" questions, which is much higher impact than Haskell vs. Python or any other tech topic that has more ink spilled on it.
Sep 15, 2022 • 13 tweets • 3 min read
I think this "superhuman" framing is interesting because I distinctly remember when I went from thinking that people who got a lot done were really superhuman to realizing that it's really possible to do a lot more than most people expect and it was due to random events.

In HS, I happened to be classmates with @PoShenLoh, who described a textbook reading technique he called "the bulldozer method", where you do all of the exercises.

That seemed bonkers and I filed it away as something a genius like Po could do, but impractical for normal people.
Sep 14, 2022 • 20 tweets • 4 min read
Another area where, AFAICT, the best way to get information is how organizations work, what kinds of strategies make sense for large orgs, etc.

There quite a bit of public commentary on organizational strategies, changes, etc., but it's mostly quite bad. Being willing to provide detailed public commentary on "current events" in tech is strongly negatively correlated with having good information about what's going on because of course people with inside info aren't generally going to comment as things are happening, but,
Sep 13, 2022 • 4 tweets • 2 min read
Evaluating futurist prediction methods and accuracy:

danluu.com/futurist-predi… Ray Kurzweil: 7% accuracy R...Ray Kurzweil has claimed to... Just saw this other example that illustrates how obvious things can seem from inside a big company. In the outside world, this would've seemed like a fairly bold prediction, but at Google, at least three people were independently making the case for custom deep learning hardware Nabeel @nabeelqu Sep 6 Pete...