HNHacker News
TopNewBestAskShowJobs

carlineng

656 karma · joined March 12, 2015

Working on the Malloy query language at Meta. Our mission is to replace analytical SQL. Check us out at https://malloydata.github.io/documentation/index.html

[ my public key: https://keybase.io/carlineng; my proof: https://keybase.io/carlineng/sigs/SxKhzboFFXEbC-x-9-NGEtke2DKLBS2wJCzYxL2vPV8 ]

submissionscomments
carlineng··on Microsoft Wallops Amazon in 2020 Cloud Revenue
From the article:

"I suspect that many still inaccurately equate or conflate “cloud” with “IaaS,” which is a perspective at least a full decade out of date. Without question, Amazon created and still leads the IaaS segment of the cloud—but the cloud of today looks nothing at all like it did 10 years ago when “cloud = IaaS” was a reasonable equation."

The "Cloud Revenue" number includes things like Dynamics, 365, etc.

carlineng··on Self-Directed Pandas Crash Course
DuckDB is a really promising project for just that: https://duckdb.org/
carlineng··on Databricks is an RDBMS
Not quite -- Snowflake allows you to query data directly from S3 without having to ingest it. You can either query the S3 file directly [1], or define an external table on top of it [2]. If the data is JSON, Avro, XML, or some other "semi-structured" format, then Snowflake's "variant" data type makes this type of data very easy to work with. For raw text or log lines, you can still load to Snowflake and process using regex or UDFs. For unstructured or "complex" data like images and PDFs, then you may need to reach for a different compute environment.

[1] https://docs.snowflake.com/en/user-guide/querying-stage.html [2] https://docs.snowflake.com/en/user-guide/tables-external-int...

carlineng··on Loon’s final flight
GCP
carlineng··on Ask HN: Is frugality underrated in startups?
Minimum rate of return is only one measure of cost. If we look at cash flows, the story is much different. Startups are usually cash-constrained, and equity financing is a way to raise cash without negatively impacting future prospects. Debt financing causes a drag on a company's cash flows and reduces flexibility, since now the company must divert a portion of its cash flow to interest payments. For a young company with low revenues and no profits, and thus unable to make tax deductions on interest, debt financing is actually a highly unattractive proposition.
carlineng··on No worldview can account for everything that will ever happen
I found this thread originally from Stratechery's daily update for January 13th, 2021. The article itself is behind a paywall, but the Twitter thread is insightful enough that I think it deserves more exposure.

https://stratechery.com/2021/internet-3-0-and-the-beginning-...

carlineng··on Visa and Plaid Abandon Merger After Antitrust Division’s Suit to Block
With public market valuations of B2B companies today, I could see Plaid being worth considerably more than the $5 billion that Visa agreed to. I think it's less likely buyer's remorse than just overwhelming evidence of anticompetitive behavior. The original DOJ complaint [1] has a lot of direct quotes from top level Visa execs. See paragraphs 9 and 10.

[1] https://www.justice.gov/opa/press-release/file/1334726/downl...

carlineng··on How to Become a Data Engineer in 2021
The CEO of Chartio recently gave a talk advocating for wide tables [1], but I don't fully agree.

Having a wide table can have some performance benefits, but I think the primary benefit of dimensional modelling isn't related to performance, but having standardized definitions, or "conformed" dimensions. For example, having a "calendar" dimension with attributes you can pull into your analysis such as fiscal quarter or reporting week.

In a data model with wide tables, if you need to change the definition of a dimension, it's going to be painful to track down all the different places where that old dimension is used.

[1] https://www.youtube.com/watch?v=3OcS2TMXELU

carlineng··on Quitting a New Job
"working on a product that would take many years to reach mass-market adoption due to pesky little things like manufacturing and road safety."

I know that sentence was probably meant to be humorous and not taken too seriously, but it comes off as smug and disdainful. Manufacturing is very difficult, and road safety is anything but trivial.

He joined the company because he believed in the mission, but then decided the lifestyle tradeoff wasn't worth it? His conviction must not have been very strong to begin with.

carlineng··on Bitcoin is a disaster
Bitcoin total market cap is ~$500 billion, so it's still a pretty small niche, and I think there's quite a ways to go before it becomes "too big to fail".

For comparison, the market cap of gold is ~$10 trillion. The majority of the US economy is not hooked into Bitcoin, and very few institutional investors have any significant exposure to Bitcoin. Contrast that with the '08 financial crisis that was based in real estate -- the US real estate market at the peak of the crisis was ~$23 trillion [1], with nearly every major financial institution heavily exposed.

[1]: https://www.federalreserve.gov/releases/z1/dataviz/z1/balanc...

carlineng··on Oracle Moving HQ to Austin
Another fun fact, the San Carlos airport nearby has an FAA identifier of SQL: https://www.airnav.com/airport/KSQL
carlineng··on Wagnerism
For those like me who don't know much about the technical aspects of music discussed in this comment (motif, texture, modulation, etc.), I recommend checking out Craig Wright's lecture series on "Listening to Music" offered by Yale Open Courses: https://www.youtube.com/playlist?list=PLhYB41gNRjT1L2CyJRwtS...
carlineng··on Scaling Datastores at Slack with Vitess
Citus was acquired by MSFT in 2019, hence the integration. Vitess is primarily a MySQL-based solution, and I imagine that AWS would rather have its customers on Postgres-flavored DBs, given that MySQL is currently owned by Oracle.
carlineng··on Wish S-1
From p153:

> Our advertising costs to acquire new users constituted 96% of our sales and marketing expenses, and sales and marketing expenses constituted 91% of our operating expenses, in 2019.

That's pretty astounding. In 2019 they spent $1.46 billion on sales and marketing, of which $1.41 billion went to new user acquisition. According to the quarterly breakdown of metrics on p108, their LTM Active Buyers went from 61m in March 2019 to 63 million in March 2020, implying they spent nearly a billion and a half to increase their purchasing base by 2 million users... even looking at their peak of June 30 2020, they still only have 70 LTM buyers, implying a CAC for "quality" users of at least $150, and very likely more. Their discussion of CAC payback and LTV / CAC only includes numbers from 2016, which looks like their best year in terms of LTM Active Buyer growth, both on a percentage basis and on an absolute basis... For me to have any faith in this business, I would need to see some kind of analysis of "organic" purchasing behavior. If they turn off the ad spend, will people continue to use the product?

carlineng··on Airbnb S-1
I don't think the labels of "tech" vs "hospitality" are exclusive in any way. I think there's a good argument to be made that we should consider Airbnb a "hospitality" company, but ultimately the label of "tech" company is shorthand for a set of characteristics of the business that other sibling comments have called out -- large amounts of R&D, low marginal costs, and low CapEx. Their product is hospitality, but the company has operating characteristics that have more in common with Facebook than with Marriott.
carlineng··on Airbnb S-1
A few random observations I found interesting:

- Looks like they are significantly leaner now as a result of COVID. Cost of revenue is at an all-time low in Q3 (17% of revenue on p164). There's a lot of focus the drastic cuts to Sales and Marketing, but it looks like there's a fair amount of increased discipline across the board.

- It's interesting that Operations and support looks to be very similar in both the first 9 months of 2019 and 2020 despite having drastically different revenue numbers. They had to spend roughly the same amount (600 million) to support 30% less revenue

- Looks like China is a huge part of their future strategy, so much so that it is explicitly called out in their Executive Compensation bonus calculation (p271). I can think of exactly one other American consumer tech company that has had success in China.

carlineng··on Strava raises $110M, growth rate of 2M new users per month in 2020
> Predictably, not many people subscribed.

How do you know this? I don't believe I've ever seen them (or anyone else) publish actual subscription numbers... closest thing I've seen is some statistical analysis done by a VC way back in 2015.

carlineng··on NoisePage – Self-Driving Database Management System
Somewhat related is Pavlo's musings on Naming a Database Management System [1]:

"In my opinion, the best DBMS names from the last thirty years are Postgres[4] and Clickhouse.

These two names did not mean anything before, but now they only have one connotation. There is no ambiguity. There is no overlap with other DBMS names or non-database entities. They are easy to spell correctly[5]. Everyone refers to them by their full name. Nobody mistakingly calls them "PostgresDB" or "ClickhouseSQL."

After reflecting on what makes them so good, I realized what the secret was to them. They are a two-syllable name that is derived from combining two unrelated one-syllable words together (e.g., Post + Gres, Click + House). Each individual word has its own meaning. It is only when you put them together does it mean the database.

Given this, I henceforth contend that the best way to name a DBMS is to combine two one-syllable words."

Noise + Page

[1]: https://www.cs.cmu.edu/~pavlo/blog/2020/03/on-naming-a-datab...

carlineng··on Lots of overnight tragedies, no overnight miracles
The CEO of CircleUp stepped down and published an email he sent to a former board member accusing said board member of poor/damaging behavior. The board member is suspected to be from Collaborative Fund: https://news.ycombinator.com/item?id=24793170
carlineng··on Data engineering learning path with recommended resources
Security and privacy are last on the list and marked with an "essentiality" score of 1/3. I think as an industry, we need to do a better job of emphasizing and prioritizing those topics early and often throughout the educational process, or else the perpetual cycle of data misuse, leaks, and breaches is bound to continue.
carlineng··on Palantir S-1
There are rules around revenue recognition [1]. These rules provide guiding principles, and there's some human judgment involved in ironing out the details. Its almost certain that the revenue would be spread over the life of the contract, but it's possible that the contract has some growth baked in, so the latter years may be more heavily weighted than the early years.

You can see how much prepaid money is "waiting" to be recognized by looking at Deferred Revenue on the balance sheet. $263 million at the end of 2019.

[1] https://www.investopedia.com/terms/r/revenuerecognition.asp

carlineng··on Palantir S-1 shows the company has 125 total customers
Average revenue per customer of $5.6 million per year. Previous articles mention that 91% of revenue growth comes from existing customers, and this is probably a large reason why. With such enormous contract values, their sales cycle is probably hugely long, complex, and expensive.

Also very interesting that 60% of revenue is generated internationally, which seems counter to the narrative that they're simply a glorified Federal systems integrator. That's undoubtedly a huge part of their business, but clearly they've been able to expand beyond that.

My take after reading all this is there's probably not a huge amount of upside remaining in Palantir. They'll probably be hard to displace in their existing accounts, but given how slowly they're acquiring new customers, my guess is that their product is not very competitive in the market anymore.

carlineng··on AMD Reports Q2 2020 Earnings
"yet even in their most recent results Intel revenue/earnings still dwarfs AMD's."

Paradoxically this is likely WHY they are getting punished so much by the market. Their current revenue/margins are too juicy to give up, despite the fact that it's causing them massive pain on the technological front.

carlineng··on D.E. Shaw and how computer geeks and English majors transformed Wall St. (2018)
I'll let dcaisen speak for himself, but I don't get the impression that he believes DE Shaw is actively harmful; just that the service they provide (market liquidity, low spreads) comes at a relatively high cost. "How?" is a perfectly legitimate question, and it looks like a problem he's actively working on as the CEO of Proof Trading.
carlineng··on D.E. Shaw and how computer geeks and English majors transformed Wall St. (2018)
GP is in agreement with you on the last point, and the highlighted conflicts of interest are not related to DE Shaw:

"brokers and exchanges are much worse [than DE Shaw] since they're fiduciaries and semi-regulatory entities, respectively, and riddled with conflicts of interest."

carlineng··on Fivetran Is the Cloud’s Newest Unicorn with $1.2B Valuation
Their Series B was only 9 months ago [1], and based on the amount raised ($44 million), I would guess they were valued at around 400-500 million. This would mean they've more than doubled their valuation in less than a year. Congrats to the Fivetran team for all the success!

[1] https://techcrunch.com/2019/09/24/fivetran-hauls-in-44m-seri...

carlineng··on Palantir to File IPO in Weeks for Possible Fall Debut
Super interesting that their funding round in 2015 valued them at $20 billion. In late 2015, public SaaS revenue multiples were in the 6-7x range [1]. According to the article, they are looking to hit $1 billion in revenue this year. If we assume a 25% compound annual growth rate from 2015 to today, this would put their 2015 revenues at ~$350 million, for a multiple of >55x (Zoom's current EV/NTV revenue multiple) at the time of the raise. That's pretty astounding, and makes me think there must be more to the story here (and not necessarily in a good way).

[1]: https://www.meritechcapital.com/public-comparables/enterpris...

carlineng··on Slack drops 10% after its revenue growth guidance fails to impress
> ... investors had expected the firm to more aggressively beat expectations

I know this is the norm nowadays with companies sandbagging guidance, but the above statement is still absurd.

carlineng··on Behind tech layoffs lay cash flow-negative companies
Until recently, Yelp had a giant stockpile of cash, but an activist investor successfully launched a campaign to get Yelp management to return much of that cash via buybacks [1]. By the end of 2018, Yelp had a cash balance of ~$837 million and proceeded to return almost $500 million in cash to stockholders via buybacks in 2019 [2].

Hoarding cash may seem like a good idea in hindsight, but public market investors will often raise a huge stink if you try.

[1] https://sqnletters.com/content/uploads/2019/01/Yelp-A-Fresh-... [2] https://ycharts.com/companies/YELP/stock_buyback

carlineng··on Show HN: Census – Export for data warehouses
Excited to follow your progress! I view this problem as the one of the biggest gaps in today's "Cloud Data Ecosystem". Tools like Stitch and Fivetran make it super easy to extract data from source systems; next-gen cloud data platforms like Snowflake make storing, transforming, and querying that data a breeze (especially with the help of tools like dbt and dataform); and there are a ton of powerful and easy to use BI tools for visualizing and digesting that data. But the minute you need to send that data to other systems, it's back to painful, failure-prone, and mind-numbing scripts.
← PreviousPage 3 of 4Next →