It should be good. It's the third time he's written this exact same announcement, including "taking the blame" and "making the difficult choice to cut a large group instead of smaller cuts over time" and "thanking the expendables that got the company where it is."
I've actually had really positive responses. I'm fairly senior (~20 years of experience). I was laid off by Meta in 2022, started at Block 3 months later. Laid off by Block in 2024, started at a smaller company 1 month later. Decided to leave that company in early 2025, contacted one company from a HN Who's Hiring post and took that job. That ended up being a poor fit, and I went back to a FAANG around July of 2025.
In the last three transitions I applied to a grand total of 5 companies.
Also, looking at the recruiter emails I've been getting, they've been ramping up over the last few months, and I'm back up to one or two cold emails per week.
But again, I'm fairly senior, and I have deep domain knowledge in a few key areas. I understand the market is brutal if you're early career or your knowledge isn't "T" shaped.
FSD automatically shuts off and relinquishes control to the user in an emergency. I bet the real number of FSD accidents is far far higher, but they're using this loophole to claim it's lower than it is. If you call them out on it, they just hide behind their "the driver must be in control at all times" legal shield.
I would have loved to see some discussion of mono- versus poly-repos and steps towards improving this situation.
I prefer mono-repos personally, but they have the downside of being very busy and bulky because of that. Doing a partial clone by folder is very clunky at best, and there is no per-folder notion of permissions, which makes this a total non-starter for my team. The result is that we have thousands of repos and tooling that helps you check them all out into the correct place to make builds across them work. And inevitably they get out of sync and the builds break.
Git submodules seem like a solid approach, allowing you to cobble a bunch of poly-repos into something that kind of looks like a mono-repo, but the tooling support around it just makes it super painful, and you still don't have the ability to have an atomic commit across the repos. And people *hate* submodules.
I don't know the solution, but I wish this was on the radar of people who know git better than I. For now I'll just bumble through the repos I need to touch.
I feel like there are some key differences between the companies though.
The second one outlined for Meta is:
> Heavily-redacted undated internal document discussing “School Blasts” as a strategy for gaining more high school users (mass notifications sent during the school day).
This sounds a lot like Meta being intentionally disruptive.
The first one outlined for YouTube is:
> Slidedeck on the role that YouTube’s autoplay feature plays in “Tech Addiction” that concludes “Verdict: Autoplay could be potentially disrupting sleep patterns. Disabling or limiting Autoplay during the night could result in sleep savings.”
This sounds like YouTube proactively looking for solutions to a problem. And later on for YouTube:
> Discussing efforts to improve digital well-being, particularly among youth. Identified three concern areas impacting users 13-24 disproportionately: habitual heavy use, late night use, and unintentional use.
This sounds like YouTube taking actual steps to improve the situation.
> This sounds like YouTube taking actual steps to improve the situation.
The issue I take with statements like that is that they are saying one thing while doing the opposite. This document [1], for instance, shows that YouTube knew as early as April 2025 that infinite feeds of short form content can "displace valuable activities like time with friends or sleep", but that hasn't stopped them from aggressively pushing YouTube shorts everywhere.
The most charitable interpretation I can think of is that there are two factions, one worried about the effects of YouTube in teens and a second one worried about growth at all costs. And I don't think the first one is winning.
I think the reality for any product that has >7,000 employees working on it is that some people's job is to prioritize growth at all costs, some people's job is to prioritize the effects of on vulnerable people, and the vast majority of them have other jobs to be doing. This sounds appropriate to me; not everybody can be worried about mental health at all times, and somebody needs to focus on growth.
There are plenty of examples that the mental health people aren't being completely steamrolled. Parental controls allow you to block Shorts for your kids. That doesn't sound like a "growth at all costs" mindset.
> I think the reality for any product that has >7,000 employees working on it is that some people's job is to prioritize growth at all costs, some people's job is to prioritize the effects of on vulnerable people, and the vast majority of them have other jobs to be doing.
... it's not at all costs though, that would be easier, because then the situation would be more obvious (legibility is important, so is plausible deniability)
so of course "growth hackers" (or whatever the folks responsible for growth are called nowadays... other than CFOs and CEOs), simply they are the ones whose judgement and "worldview" regarding whose responsibility is to manage the negative consequences of their increased revenue is very skewed, in other words they mostly have elaborate self-serving explanations (excuses)
and many times that overlaps various user freedom arguments, arguments against paternalism, etc...
My YouTube use definitely isn't healthy, but it's still the only social app that asks me to take a break if I use it too long or late at night. That should be standard in any of these apps.
Does it recommend taking a break? Mostly I've seen it ask if I'm still watching. I've always assumed this is not for user benefit, but in order to not spend bandwidth on a screen that is not being looked at.
The only site I'm familiar with that has somewhat decent self-limiting functions built in is HN's no procrastination settings. But that's of course because HN isn't run to make money, but as a hobby.
Youtube isn't doing that for your health. It's so they're not wasting ads and bandwidth on users who aren't watching anymore to maximize their profits. The sole purpose of a for-profit corporation is to generate revenue for its owners and shareholders. That's it. Nothing more, nothing less. Do not expect anything different. Corporations only care so long as its profitable.
No, it sounds like youtube being fully aware of the consequences of their offering but couched in terms that allows them to pretend they were not. 'could' indeed.
Not realistic to reply to all your replies re:youtube, but they've absolutely added some features to mitigate bedtime use and at least for me they were opt-out rather than opt-in.
I guess some companies try to limit the harm they do to children while profiting, and some companies try not to know the harm they do to children while profiting. What remains to be seen is how much harm we allow to be done to children in the name of profits. Maybe we even insist that things need to be a positive influence. Less profit, but maybe better to the economy over all. And the kids, if they matter.
I believe the whole point is that some people inside acknowledge the issue, made leadership aware of it, yet, youtube still pushed sorts aggressively. The documents are prof of awareness, so they can't pretend they were unaware of the issues.
The glass of the window does not have a frame. You want the glass to go into a rubber seal to really prevent air from getting in and whistling at high speeds. If there's a frame around it, then no problem, the seals move with the glass when you open the door. But if you don't have a frame then opening the door without retracting the glass will cause it to pull at the rubber seals. At best it'll wear the rubber faster, but eventually it'll pull the rubber seal out.
This is very common on cars where the windows don't have a frame. Before I had a Tesla I had a convertible Mustang. Because it was a soft top it didn't have the same kinds of seals. Instead it used lateral pressure to hold the window against some rubber. At freeway speeds the window would flex and let air in. Eventually the soft top started blocking the passenger side window from meeting the rubber, and there was always a 1/4" gap unless I rolled the window down a bit and then back up.
I'm a fan of this too, I think it's a very clever design. But I also think it'd be pretty trivial to make in a six axis CNC. Maybe even a 4 axis if you're clever with your mounting.
The algorithm for the checksum (the sixth digit) is subject to one of the most common human errors, swapping adjacent digits. The UPC checksum algorithm handles this without significantly more complexity. They have you multiply all of the numbers in odd positions by 3 and then add up all numbers. The last digit is chosen to make the sum a multiple of 10.
To use your example: 51076, you'd do `5*3 + 1 + 0*3 + 7 + 6*3 = 15 + 1 + 0 + 7 + 18 = 41`. The sixth digit would be 9 ((10 - (41 mod 10)) mod 10). If you were to transpose any two adjacent numbers the checksum would be off. 3 is chosen because it's the smallest number that is co-prime with 10.
I'm a software engineer who does woodworking in my spare time. But I've never experienced that satisfaction. I've never made something that is perfect. Every time I look at something I've made all I can see are the flaws. Most of my things are smaller, but I can look at a project I completed 4 years ago and know exactly where the tear-out is that I had to hide, or the errant marking knife line that I tried to sand away, or the snipe from the planer that I didn't have enough spare material to be able to cut off, or the piece of wood that is perfectly shaped but there's a knot that just doesn't look quite right there.
At least with software I can go back and edit my past transgressions.
reply