steviesands

API gateways are primarily used for HTTP traffic coming from clients external to your backend services eg. an iOS device (hence the term 'gateway' vs. 'mesh'). I don't think they support thrift or grpc (at least aws doesn't, not sure about other providers). https://aws.amazon.com/api-gateway/

Counterpoint to the Clarendon vs. DC. IMO Clarendon bars are chock full of college frat bros with next to no diversity, while DC is more of a mixed scene.

As someone who moved from DC proper to NYC, DC is nowhere near NYC in terms of expenses. In 14th St or Logan circle you can get a brand new 2BR for the price of an old tiny, dilapidated 1BR in an equivalent area in NYC. You can likely even get a 3BR older row house for the same price as a 1BR in Manhattan or prime Brooklyn.

Not to mention, in NYC it's very difficult to buy a home as they are ~30-40% more expensive than renting, with high maintenance fees. For example, a 2BR in Upper East or West side (not near central park) will be ~6-9k a month with ~3k a month in taxes/maintenance.

The (made up) cocktail price index is also illustrative and in NYC one cocktail is ~$17-20 and DC is ~$11-14.

If you are interested in the genuine impacts and not a hyperbolic, emotionally charged discussion best left to pundits, then below are some areas I would suggest researching and conducting your own thought experiments. A good exercise is to pretend if you were a CTO presenting to the CEO the tradeoffs and risks in cost cutting.

- impact on reliability (sum across users for key metrics, not anecdotal evidence)

- systemic risk changes, measured using probability of a rare event

- feature velocity, volume of features, 'width' of features (how_many_bets x feature_surface_areas)

- the durability of systems engineered for reliability or "did we build systems that will survive if everyone leaves"

- revenue

- advertiser satisfaction with the technical platform

- impacts to internally maintained systems

- impact of loss of institutional knowledge

- impact of further attrition post-RIF

- compliance

- rate of progress vs competitors as an advertising platform

- rate of progress vs competitors in social media

- (related to above two) ability to maintain competitives advantages with reduced headcount

- probability of success for new revenue streams

There has also been recent reporting on Twitter: https://nypost.com/2023/01/18/twitters-daily-revenue-plunged... https://www.axios.com/2023/01/29/fidelity-cuts-twitter-valua...

I read it a while ago and while I can't recall what was said exactly, I remember how I felt while reading it and how strong it was. The mark of a good book.

How did you achieve 4 days a week? Curious to hear about your strategy.

If your goal is maintenance mode for the company, milk profits for 10 years, and to be eventually replaced by competitors then I think you can run a company very slim. The other question would be, if you have captured a large market share (100s of M/yr to Billions in revenue), how many employees do you need to prevent an upstart from overtaking you through improved tech, product, or strategy?

For example, if FB never invested in ML they would have had even larger margins (fewer GPUs and ML engineers), but now that investment may pay off by fending off tiktok through copycat products and also rebuilding ad attribution after ATT. To complicate matters, before it happens, you don't know in what area your competitor will arise (ML? Product? Paradigm shift?). Similar examples with Google vs. OpenAI, ~2010s Kubernetes wars between cloud providers, Snap vs. FB/Twitter, etc.

Can you elaborate on budgets realigned? Wouldn't that have happened in Q4 for the new year?

Interesting, 80% seems a bit on the higher end nowadays though? For example, Google quantified this as the "datacenter tax" and through their cluster wide profiling tooling saw that it was 22-27% of all CPU cycles (still a huge amount). They go a different route and suggest hardware accelerators for common operations. Datacenter tax was defined as:

"The components that we included in the tax classification are: protocol buffer management, remote procedure calls (RPCs), hashing, compression, memory allocation and data movement."

https://static.googleusercontent.com/media/research.google.c...

A few thoughts. The first is, are we asking the wrong questions? Should it be, "If I spend 10m on hardware for predicting ads (storage/compute) that generates 25m in revenue, should I buy the hardware?". Sure, we can "minify" twitter, and it's a wonderful thought experiment, but it seems devoid of the context of revenue generation.

The second is, it's interesting to understand social media industry wide infra cost per user. If you look at FB, Snap, etc. they are within all within an order of magnitude in cost per DAU (DAU / Cost of revenue) of each other. This can be verified via 10-ks which show Twitter at $1.4B vs. SNAP 1.7B Cost of Revenue. The major difference between the platforms is revenue per user, with FB being the notable exception.

Also would you summarize the patent/architecture? The link is a bit opaque/hard to read.

Note: Cost of Revenue does also include TAC and revenue sharing (IIRC) and not just Infra costs but in theory they would also be at similar levels.

eg. SNAPs 10-k https://d18rn0p25nwr6d.cloudfront.net/CIK-0001564408/da8288a...

Is this how it works at most companies where new budget leads to new headcount and hiring in January? Is it due to an annual process?

I checked a few days ago and the revenue per employee at big tech is eerily similar to "Biglaw" and non retail banking (Jones day, >200k entry level, goldman is similar) at 1-2mil per employee. One could argue the market for IB/trading has been saturated by applicants for years but they pay is still well above norms ~>150k entry level. Pretty interesting.

Interestingly revenue per employee (1.2 - 1.8 mil per employee) at biglaw companies is similar to FAANG although i'd guess biglaw margins are higher as you have less other costs (infra).

https://www.law.com/law-firm-profile/?id=178&name=Latham-%26...

Does this logic work for lawyers? "Biglaw" firms are similar to FAANG in that it is a handful of companies that offer very high comp (235k for first year). There is a deluge of folks that apply from all universities to biglaw but generally they primarily hire from top tier law schools and even then not all make it.

Similar for Big 3 consulting firms (BCG, Bain, McKinsey) in that there is a huge application pool for positions but they are difficult to get and provide high pay.

Maybe you could provide links, examples, or helpful context instead of snark?

Twitter has ads serving infra, recommendation systems (timeline, notifications, events, users), user generated events, prediction systems (ads), user graphs. The complexity is from processing and persisting exabytes of data in company owned datacenters. eg. Twitter stores images, videos, user events, user data, tweets/replies. WhatsApp has little persistence outside of metadata maybe? But your messages are not stored in a FB datacenter and if they are I'd be concerned. You can read about their infra in their blog. Comparing p2p messaging versus a distributed social media site with mountains of data and years of iteration in ML systems does not make sense.

https://blog.twitter.com/engineering/en_us/topics/infrastruc...