Hacker News Reader

Comment by jrockway | Hacker News Reader

What's the context behind this? I've used CircleCI at many jobs, and I guess people must have moved on to some better tool? What's the better tool?

View on HN

> What's the context behind this?

From the article:

> Right now, however, there is a ripple effect of uncertainty in our sector. Since the start of Q4, we’ve seen a dramatic shift in how every company’s performance is evaluated. Companies were once praised for growth at all costs. Very rapidly, market expectations have shifted. The emphasis now is on maximizing efficiency.

It sounds like they are cowardly using the Google/Amazon/Twitter workforce decimation campaign as a scapegoat to fire people.

I find their explanation pretty reasonable. In 2020 and 2021 money used to be close to free, with interest rates near zero and VCs throwing money at anyone with a growth story. Now both of these are no longer true, VCs are much more reserved and bank loans much more expensive. And with money being more expensive to acquire, everyone is more considerate about spending money (and thus manpower).

> I find their explanation pretty reasonable. In 2020 and 2021 money used to be close to free, with interest rates near zero and VCs throwing money at anyone with a growth story.

You're trying to put words in their mouth. Nowhere in their rationale they mention funding or money or interest rates or revenue or results. They only argue that they decided to base their HR policies on a fad. To me this looks like an awfully cowardly way to justify their own decisions on a pseudo-force majeure. Bullshit through and through.

Just about everyone is being more cautious about headcount and even if they're not net shrinking their workforce, they're looking harder at what people are working on.

I've heard about Twitter and Amazon campaigns, but not Google's yet. Would it be possible to provide a link for that?

CircleCI has gone the same way as Docker. They pioneered a paradigm but late entrants like GitHub/GitLab/Bitrise etc stole the show with their offerings. Most teams I’ve seen use either of these instead of CircleCI

Too many engineers getting caught up in the engineering of it and not the problem solving of it.

I love the concept of Docker, but I hate working with it any time I need to work with it. It's extremely opaque and unnecessarily hard to do basic things.

I don't know what language you use, but I figured out the success of Docker. I use Go, so my entire program is just one statically-linked binary. I can //go:embed any non-code files that I need at runtime. So I am naturally annoyed at the overhead Docker introduces with builds (throwing away my beloved and reliable ~/.cache for no reason), but I realize I'm not the target market, because most people don't program in a language with a build system. The target market is Python and Javascript programmers. Those languages can't even compute what modules you depend on without running arbitrary code from the Internet, much less produce a single file that will run on any reasonable Linux box. ("Hi! I got drunk and ran someone over with my motorcycle, but I'm out of jail and I need a job. Click here to give me one!" is an actual message that NPM will print in any large project, though it omits some details about the whole killing someone thing.)

Docker is a system for shipping your laptop to production, and because of the inadequacies in the tooling of those popular languages, it's in high demand.

> Docker is a system for shipping your laptop to production, and because of the inadequacies in the tooling of those popular languages, it's in high demand.

I know this is in jest, but I don't agree with this. Docker is a way to run code in an environment you fully control without knowing exactly what it will be running on.

Many of the popular tools can generate a single file, but they'll be built for a specific architecture or require a certain run time.

Yeah, that's a very fair interpretation. Honestly, a lot of programming languages need a lot of files to be in exactly the right place or your code will blow up. I like minimizing the number of files you need around, but making sure that you can have all of those files you need in production is a good strategy as well.

I am not sure about "better" but a lot of people who previously used Travis and Circle use GitHub Actions these days (and maybe azure in the meantime).

It does seem like CI/CD turned out to be less of a product and more of a feature.

I think CI in particular turned out to be something that people really, really don't like being unreliable and it's incredibly challenging to operate reliably and cost effective. For Microsoft providing that service is somewhat possible because they have the warchest and operation teams.

I think there is a different world where Travis (and Circle) could have been built differently and maybe with a slightly different operating model (eg: easier to operate self hosted runners) where they could have made it harder for GitHub to compete in the beginning. But not sure if that would have delayed the inevitable.

I would assume some combination of “VC funding is no longer raining from the sky” and “GitHub Actions are good enough for enough teams not to want a second procurement”.

Just because this comment and/or like of discussion could be perceived as insensitive on a day when people are getting tough news, can I jump in to say that I think it’s more likely to be symptomatic of the general global downturn and current wave of cost-cutting than it is any particular issue with product fit or competitors.

I’d also add something along the line of SaaS fatigue. The costs of managing vendors, securing & monitoring, etc. add up and it’s not uncommon for the overhead cost of another service to be more than the benefits of using it. CI falls into that category since there’s an upper bound to how much a dedicated service can give you, especially when the competition like GitHub or GitLab gives you a bunch of other features for the same price.

Reply on news.ycombinator.com