HNHacker News
TopNewBestAskShowJobs

lhorie

9,880 karma · joined September 5, 2008

submissionscomments
lhorie··on A Case Study on the Uber Alumni Network
Superpumped, by Mike Isaac.

Mind you, he was a journalist working off of leaks, presumably from disgruntled employees (of which there were many in the high attrition era), and he obviously had a journalitically oriented profit motive of his own.

So take it with a grain of salt, I guess.

lhorie··on A Case Study on the Uber Alumni Network
Uber is a ship of Theseus.

Lots of early people left since the Travis days. Since then, fixing the reputation damage has been a big focus in the company. Safety center, electric fleet commitments, DEI, Ukraine initiatives, etc, are all newer initiatives in that vein.

"Do the right thing" is now a company core value plastered all over the offices.

Even the "off the shelf stuff doesn't meet our needs" culture alluded in the article has given way to adopting 3rd party solutions more (e.g. Slack replacing uChat, Oracle partnership, etc). Lots of things have changed over the years.

lhorie··on Launch HN: Second (YC W23) – AI bots that add features to web apps
Are you using this to build the product itself? Seems like a prime candidate for dogfood.

If yes, where has it been useful, and if not, what does it not handle well?

lhorie··on Want an unfair advantage in your tech career? Consume content for other roles
I can't speak for every company, but the impression I get is that recruiters at the FAANGs of the world seem to search for various coarse rubrics that may or may not correlate with whatever they are looking for, be it a high number of years of experience as a potential correlation to seniority or certain keywords as correlation for expertise in some subject matter. "Cloud" and derivative keywords are probably the ones landing your linkedin profile in the search results of these recruiters.

Not everyone is getting recruiter spam because not all keywords are created equal and not all profiles look promising, but those of us who do get spammed, get a fair amount of it.

The Amazon recruiter spam in particular is bad enough to be a running joke.

lhorie··on Want an unfair advantage in your tech career? Consume content for other roles
OAs aren't super widespread in "elite" companies for a variety of reasons.

One is that sourcing recruiters are not equipped to handle technical screening logistics. They can't tell, for example, if a solution works but has O(n^2) complexity, or if the candidate failed because the tool was being stupid in some way or the question was phrased ambiguously.

Another is that OAs can be gamed much more easily than face-to-face phone screen sessions (see, for example, reports of indian "study" groups).

Another issue is that the evaluation criteria for OAs is often so coarse that it does not provide a reasonable level of filtering, compared to having a combo of a good sourcing team + an engineer "guarding the gate" so to speak via the phone screen session. An engineer's technical screening skill is considered non-automatable enough that companies would fly them into school fairs to vet internship candidates.

Teams may not even trust OAs that they didn't devise themselves, nor have the actual inclination to develop/maintain their own OAs.

OAs also tend to gravitate towards trivia-style quizzes that are often associated w/ "lesser" companies.

There may also be concerns about being respectful of candidates' time.

lhorie··on Want an unfair advantage in your tech career? Consume content for other roles
> wouldn’t it behoove you to practice that too

I mean, sure, the more you do the merrier. But that goes back to what I was saying: if "unfair advantage" means extracurriculars in high school to get into an ivy league + the ivy league degree to get spotted by recruiters + internships in STAR format on resumes + grinding leetcode + etc etc to increase your chances as much as possible, is any of that really a clever life hack as the "grind LC" meme implies, or is it merely just going way above and beyond what everyone else is able to/willing to put up with in the traditional career path?

If anything, it sounds more like going through a non-traditional career to build up the "street smarts" is the shortcut to big tech riches, given that the alternative is to try to stand out from a sea of millions of resumes all listing C++ and python internships/roles, potentially never landing big tech at all.

lhorie··on Want an unfair advantage in your tech career? Consume content for other roles
Many of us went through circuitous routes. My title is SWE at a big tech company and I interview SWE candidates, so I think I'm relatively close to how the sausage is made in terms of the pipeline through which most engineering candidates go through to get into big tech.

Personally I think "grind LC" is a very common meme perpetuated by folks on the candidate side, but it lacks a lot of context in terms of how modern big tech interviews are actually done. Focusing exclusively on code, for example, is IMHO, a counter productive way of hyper-focusing. I personally didn't even know what leetcode was when I joined my big tech company, for example. I did well in the loops because they evaluated various other things that I could demonstrate thanks to previous open source work (e.g. community building).

What you're basically seem to be saying is that you fell into big tech thanks to a fortuitously aligned set of previous experiences, as many of us did. But that's the exact opposite of grinding LC as a newbie. Grinding LC is one way of hyper-focusing, and going by the numbers, it's no longer necessarily sufficient to land a big tech job.

lhorie··on Want an unfair advantage in your tech career? Consume content for other roles
Then you joined big tech at the peak of the pandemic boom, when it was completely fair game to for recruiters to be poaching even foreigners due to the outsized growth outlooks. You should be aware that the economic environment has basically taken a 180 since then, and your experience might be very different from someone joining the workforce today with no work experience.

To counter the "cram LC" thing, the issue is that there are plenty of new grads doing exactly that who don't even get past ATS screening. You just don't hear about it much because places like cscq are echo chambers w/ high degrees of survivorship bias and there's qualitative levels of difference to "cramming leetcode" that can't really be captured in online discussions.

Ultimately the generalization seems to be "don't pigeonhole yourself, set yourself apart in some way", which is, IMHO, not exactly a super insightful revelation. That has always been the "smart" way to play the game.

lhorie··on How to hire engineering talent without the BS
Yes. These days onsites have at least three categories of interviews: leetcode style, system design and behavioral. For L5 level (senior) and above, the latter two tend to weigh more heavily.

It's also more common now to have two coding sessions, and even two system design ones, under the rationalization that it reduces interviewer bias.

Google explicitly links to leetcode/hackerrank/etc as suggestions for practice resources in their pre-interview email, implying that practicing artificial DS&A quizzes is now table stakes.

I would be curious as to whether these changes have been data driven, or whether the bar is being increased purely out of necessity to weed out candidates more strictly.

I certainly have seen my share of candidates that do fine in leetcode sessions but then go on to fail in the other sessions for various reasons that allegedly correlate to work performance.

lhorie··on A senior engineer's guide to the system design interview
The problem isn't demonstrating understanding of what a cache does conceptually, in fact any SWE worth their salt ought to be able to explain caching. The problem is if the candidate's decision to add a cache causes the system behavior to become obviously incorrect upon an ounce of further inspection, because this demonstrates a lack of foresight about local maxima vs global maxima, and systems are all about trade-offs at the macro scale.

Caching is an easy thing to blurt out in an interview setting, but not all problem spaces benefit from caching and caching often isn't the only available solution, or the most ideal one.

lhorie··on A senior engineer's guide to the system design interview
The way it typically fall aparts for my candidates is I ask about optimizations, they say X could be cached, then I ask about cache invalidation and lo and behold, it didn't even occur to them that the cache needs to be invalidated at some point, so they blurt something along the lines of doing the entire expensive operation again and checking it against the cache, or something similarly nonsensical.

More common than I'd like to admit.

lhorie··on Launch HN: Moonrepo (YC W23) – Open-source build system
I guess I'll echo some of the other comments re: extensibility. For us, Bazel's Starlark language is useful because it allows us to integrate w/ non-JS things like protobuf schema definition files. I suspect most companies are going to have hybrid stacks instead of Node.js-only ones. FWIW, Bazel vs Buck have some history wrt adoption wars, and Bazel largely pulled ahead due to a better ability to flourish an ecosystem (hell, the entire JS ecosystem around it is done via the extensibility system). It very much reminds me of how Typescript won against Flow.
lhorie··on Launch HN: Moonrepo (YC W23) – Open-source build system
IMHO, if you're targeting the Javascript ecosystem, this area is already fairly crowded, with Turborepo, Nx and various open source tools providing various degrees of functionality (Bazel, Pants, Lerna, etc) already competing in the space.

I'm a tech lead for the web monorepo at Uber. We talked to the Turborepo guy a few years ago, and he admitted that he wasn't sure if it could handle our scale in terms of all the bells and whistles that a repo of our size leverages - and his is one of the more feature packed commercial offerings in this space.

As a random example: we see thousands of commits a week, so invalidating the whole build graph when the lockfile changes is a complete non-starter, yet most turn-key solutions target small teams and are not equipped to cope with this problem. Phantom dependencies[0] are a problem with disjointed build graphs. Etc.

As someone who's evaluated a ton of these systems, I would love to see where your experience in this space is coming from. A new kid in the block will have a lot to prove.

[0] https://rushjs.io/pages/advanced/phantom_deps/

lhorie··on After Callous Layoffs, Workers Are Done with the Full-Time Work
Freelance is all about your network quality. When I did part time freelance to bootstrap my career I too got my dose of difficult clients. Then years later, I did a switch from FTE to consulting, with a single high quality client, and it was basically just like any other job.

What I came to realize is that freelance is actually the lower end of the spectrum for consulting agencies. A client base for such work can range from mom-and-pop businesses to veterinary chains to big telcos, with stability and size of contracts being relatively proportional to how good your sales/business development pipeline is. Needless to say, it's all work up and down the spectrum, so finding that mythical work/life balance liberation is an akin to mastering an art form.

lhorie··on Tech CEOs Should Be Held Accountable, or Even Fired, Amid Layoffs
> There’s zero introspection or root causing

You mean unlike the Experian breach post mortem, or the numerous "everything is green" Slack outages? IMHO, Costco talking about container costs is introspection. I'll grant that taking analogies too literally does indeed lead one to conclude the other party's arguments are bullshit, but I think that's just a matter of perspective and unwillingness to engage in good faith discussions.

lhorie··on Tech CEOs Should Be Held Accountable, or Even Fired, Amid Layoffs
Earnings calls are post mortems for the performance of the previous quarter. A post mortem itself is just an analysis, it doesn't even have to be about a screw up. Screw ups are something else. When you think about cases like WeWork or Theranos, then I think we get into more familiar territory: gross negligence is a widely acceptable reason for firing, no matter if you're the CEO, Joe developer or a janitor.

An operational mistake by a SWE might be something like picking the wrong technology, and for a CEO, an operational mistake might be something like a bad acquisition. Just as a technology is a tool for the dev, an acquisition is a tool for the CEO. Both of these are examples of getting bad outcomes from their tool usage decisions, but they're still categorically operational mistakes, as opposed to gross misconduct.

So my two cents is that if one wants to argue that operational mistakes categorically warrant firing personnel, then yeah, I think I'm gonna have a problem with that line of thinking.

lhorie··on Tech CEOs Should Be Held Accountable, or Even Fired, Amid Layoffs
I'm seeing a lot of this "screw up = get fired" narrative lately and I think it's a bit ironic.

Talk about screwing up in engineering contexts and suddenly the tune changes: now people will talk about improving processes, blameless postmortems, and how the individual is not at fault.

Imagine a youtuber saying googlers should be fired because some decision was made that affected the youtuber's income negatively. Or imagine an Uber driver saying that. Or a small business owner dealing w/ a Stripe mishap or whatever. If you are the one being targeted by these calls for loss of employment, then comes out the "I'm not the responsible person here" card.

So on the one hand, you hear people with pitchforks about how CEOs or whoever need to take responsibility in some movie-esque "bad guy ending" fashion, but turn the tables and nobody wants to step up to be responsible for their own shit shows.

Even if we entertain the idea of firing CEOs, is the implicit assumption that companies can run fine without one? By that logic, anyone could just give a bunch of money to a bunch of fiverr code monkeys and that business would run just fine? I don't really get what the logical conclusion of this whole line of argument is supposed to be.

lhorie··on I’m now a full-time professional open source maintainer
As someone with more than 15 years of experience and who has leveraged my open source work to improve my own career, my two cents is that GP has a fairly sober view on OSS as a job: Filippo is an outlier among outliers. Most people simply don't have the opportunity to build a reputation from multi-year, full time, corporate sponsored open source work on a high visibility project to leverage when leaving a cushy job to pursue higher rungs in Maslow's pyramid.

"Possible" is very different than "likely".

lhorie··on Ask HN: Do you use Flow (JS type checker) in 2023?
We use it extensively but are migrating away from it at Uber.

Its strictness is a double edged sword, it often trips people up with checks that are pedantic but not fully sound. For example, TS will gladly accept that `if (typeof foo.bar === 'string') doStuff(foo.bar)` narrows down `foo.bar`'s type to string. Not so with flow... because Javascript supports getters and Flow's type system isn't strong enough to keep track of what is a getter and what isn't. So it will nag about it maybe not being a string, even when it isn't warranted (which is like 99.9% of the time this happens). There's also a long standing bug where await doesn't unwrap a Promise type in some cases.

Perhaps the biggest pain is the ecosystem. There's no tool to generate libdefs and the ones in flow-typed are often crap or non-existent, so people stub things all the time.

Another sore point is lack of transparency wrt project direction. It's a Facebook-first project and it shows. Things break between patch versions. One really bad side effect is that it's essentially impossible to make semver guarantees for libraries when the type system breaks if the consumer project uses a different version of flow or even simply different configurations.

For us, migrating from earlier versions to flow's Types-First architecture is such a monumental task that we actually decided that migrating our 10+ MLOC monorepo to Typescript would be cheaper.

Also despite the claim of higher strictness, it can still lie to your face. One time, I had to submit a fix to flow-typed of all projects because the maintainer did a refactor to fix a type checker nag and accidentally broke the runtime in doing so.

lhorie··on Making sense of TypeScript using set theory
See the sibling comment from oblosys for an example where the exact same code can give different results in vanilla JS vs after a tsc pass. As you can see there, the problem is that an identifier can simultaneously represent a value and a type.

This is different than JSX or hashbang, where the set of non-JS syntax cannot legally overlap with existing syntax/semantics.

lhorie··on Making sense of TypeScript using set theory
> Typescript is designed to accept any valid Javascript program

That's not strictly true though. `a<b,c>(d)` is valid Javascript but Typescript treats it as a different syntactic construct[0].

[0] https://www.typescriptlang.org/play?#code/C4TwDgpgBARlC8UB2B...

lhorie··on WTF with high school questions?
I've heard reports that recruiters might spend all of 15 to 30 seconds per resume during a screening exercise. Never mind not having technical expertise to ask technical probing questions, sifting through the top of the funnel is more of an exercise in coping w/ high volumes of resumes than it is about validating depth of expertise at the bottom of the interview funnel.

IMHO, complaining about the status quo is easy and giving armchair hot takes about circumstances one does not understand is also easy. If one wants to play the smart game, there are some facts that can leverage to one's advantages:

1) job postings being above the fold vs long tail is a thing. There are tons of companies/roles you will never hear about from by perusing linkedin search.

2) following from that, interview criteria often needs to be adjusted based on various factors and plenty of companies have to make more streamlined funnels in order to have a fighting change of capturing any talent at all.

3) even for high visibility/demand companies, recruiters have incentives to think outside the box to find truly valuable talent, because these people are not even in the market for a new job, let alone willing to put up w/ bullshit. The catch that many people miss is that just accruing a high number of years of experience doesn't necessarily translate to being valuable, especially on the fast changing world of tech.

lhorie··on Japan was the future but it's stuck in the past
I've lived in three and find it amusing how there's two camps arguing with a straight face as if they could generalize anecdotes/conclusions about large amorphous entities. Both Norway and Venezuela are western countries. Tokyo is very different from rural Hokkaido. San Francisco is very different from New Orleans.
lhorie··on The gotcha of unhandled promise rejections
Not sure what `Promise.race(Promise.all(promises), loop());` is supposed to be doing. Promise.race only takes one arg, no? Also not sure what partialUpdate is supposed to be doing. Is that assuming some React-like diff reconciliation?

Personally, I'd go with something like this:

    async function showChapters(chapterURLs) {
      return Promise.all(
        chapterURLs
          .map((url) => fetch(url).then(r => r.json()))
          .map((p, i) => p.then(data => appendChapter(data, i), e => displayError(e, i)))
      );
    }
With this, you can control exactly where and how list items are rendered, regardless of order of promise resolution, and things are rendered as soon as they are available, instead of getting stuck behind slow requests.
lhorie··on Amazon to Lay Off over 17,000 Workers, More Than First Planned
No offense but this comment reeks of bias.

For one thing, if by your own analysis, these companies' interviewing processes aren't effective, there's logically a conclusion that yours also aren't, given that interview processes are fairly similar among big tech companies. I've even also heard of an interviewer trying to tell a director with a straight face that a high level IC candidate (L6/L7 IIRC) from a prestigious big tech couldn't solve a "simple" problem at one of those companies. Leetcode style questions aren't even necessarily good proxies for whether an engineer is actually effective at their job or not (and yes, a good number of FAANGers openly admit here and elsewhere that they needed to grind leetcode to pass interviews at these companies).

Another thing is that Uber and Amazon are fairly competitive w/ other more prestigious FAANGs in terms of compensation. A decent staff eng at Uber isn't going to be very interested in grinding leetcode for a equivalent role at Google, the compensation delta is just not appealing[0]

[0] https://www.levels.fyi/?compare=Uber,Google&track=Software%2...

lhorie··on Devpod: Remote development environment at Uber
Upgrades come in a spectrum. Some are able to automerge via dependabot or similar tools, some need minor codemodding and can be managed by tools like soucegraph's new offering. Some take months of dedicated effort and digging through the ripple effects across a web of ecosystem libraries (React comes to mind).

The problem monorepos attempt to solve isn't a technology problem, it's more of a people problem. For example, say you cron CI job fails. Then what? Someone needs to look at it.

It's easier for someone to fix things they currently have context for (e.g. if I upgrade Python or update some security-related config and something breaks, I can reason it was my change that broke it), vs an unsuspecting contractor getting around to some backlog task 6 months after the fact with no context.

Organizationally, we can shard tasks to match areas of expertise. We only need one Node.js expert, one tzdata expert, one JRE expert, etc, to upgrade each of these, instead of everyone needing to obtain above average familiarity w/ obscure FFI bullshit or whatever in each technology.

lhorie··on Devpod: Remote development environment at Uber
You use a build system. There are a number of open source ones, we use Bazel. In a nutshell, it keeps track of dependency graphs, so you can test only affected projects.

It can also cache execution of unchanged transitive steps, so you can skip builds/tests that were already run previously (e.g. you could skip most of a large 2nd CI run if all you did in a code review was edit one file.

You can also parallelize execution across cloud nodes.

lhorie··on Devpod: Remote development environment at Uber
It's the same convo as docker: can't one just use setup scripts for reproducible envs? In theory yes, but in practice, theory and practice aren't the same thing.

It's easier to apply best practices to a greenfield project written in a modern language (hence devpods) than trying to comb through a decade+ of tech debt written in bash.

lhorie··on Devpod: Remote development environment at Uber
One of the reasons we developed devpods was because our setup scripts were unmanageable.

Instructions to new employees would say things like "run this thing, scroll up past dozens of pages of stdout noise and manually deal w/ the errors buried therein by looking up relevant FAQs in some doc somewhere"

The scripts would touch every technology imaginable, from brew to npm to arc (phabricator's cli) to proprietary tools and no single person understands how setup scripts work in their entirety.

One exercise we'd get new employees to run through was get them to brainstorm about how some system ought to work. The lesson was that just about any idea they could come up with would have already been tried (and failed).

I'm told that devpods aren't even the first time we tried cloud dev envs. Presumably lots of lessons were learned from previous attempts at improving dev envs.

lhorie··on Devpod: Remote development environment at Uber
The joke is that we love them so much that we have many of them.

In seriousness, we historically used microrepos and just getting to language-specific monorepos was already a monumental effort.

← PreviousPage 2 of 34Next →