If you write code that powers an EV's 'self driving mode' - which makes calculated choices, sell it and deploy it, when that car gets into an accident under 'self driving mode', you may not be liable (depending on the case and jurisdiction - as proven in the past). The driver is.
There are many instances (where I am from, at least - and I believe in the USA), where 'accidents' happen and individuals are found not guilty. As long as you can prove that it wasn't due to negligence. Could "don't be an asshole" as instructions be enough in some arenas to prove they aren't negligent? I believe so.
I'd like to agree with this, but I sadly don't. I'd like to agree because we've been Heroku customers for about 18 years - which is wild to think about. I've used Heroku both personally and professionally day-in, day-out for over a decade.
We've been on self service and we've been on enterprise contracts. In the last 2 years I believe we've cycled through about seven account managers. Heroku as a concept might not be dead, but if you release an incredibly empty announcement saying there's no new enterprise contracts and existing ones may be renewed, enterprise Heroku is absolutely dead and I'd suggest it means Heroku as the current product is dead too.
Any Heroku user that has been at the level of an enterprise user before, or who currently is, would be ringing alarm bells at the current situation. It doesn't matter about the internal good will of employees - if you have a blog post hanging your enterprise customers out to dry (ironically as enterprise customers we have received zero communication from Heroku about this) after a year of terrible stability - you're really doing a great job of killing the whole thing.
> Heroku remains an actively supported, production-ready platform, with an emphasis on maintaining quality and operational excellence
Anyone that has used Heroku for a while will know that it is far less reliable today than it has been at nearly any point in its history (it's the least reliable since its first year of existence, imo). There is very little "operational excellence" left as an organization. All you need to do is look at how they communicated (or extreme lack-thereof) a critical outage that lasted for hours last year[1]
As an organization, we've put up with terrible reliability over the last couple of years, and swallowed cost increases every renewal and we've always been committed. That's changed in the last few days - we've tried out Railway and Northflank, and we'll continue to try out a few other services until we find the one that fits. We're lucky, we have about 9 months left on our contract and that gives us enough time to move.
Also it's important to note the topic is during school hours. There's a wealth of knowledge to learn at school, and there's also a wealth to learn outside of school. Knowledge about the world can, and will, happen in both. Many hours outside of school to 'grow your knowledge' through your phone.
Currently Claude etc. can interact with services (including AWS) via MCPs.
What the user you're replying to is saying the Bun acquisition looks silly as a dev tool for Node. However if you look at their binding work for services like s3[0], the LLM will be able to interact directly with cloud services directly (lower latency, tighter integration, simplified deployment).
I'm not too familiar with the history of a few that you've mentioned, but the Mustang (excluding the GTD), Supra, Carolla, S2000, Z never had track focused cars. For the time-poor and money-rich racing enthusiast, lacking a turn key package isn't really an option.
Your local track days where you see the likes that you mentioned - you'll also likely get a few Porches etc too. But there's absolutely a culture of racing higher end cars, the GT3's, Ferraris (488 Pista, 360 Challenge Stradale, starting to see XX's around too), the Valkyrie and Valiant etc. There's also those that grab older Cup cars and track them to the track. There's absolutely a market for turn key, track focused, road legal Ferrari's.
You can have an air gap between two physical items - it doesn't matter if those physical items are air tight or not. Air gapped doesn't mean the items are prohibited to intake air (i.e. air tight), it just means they're prohibited to intake things _apart_ from air.
It's also a good paper-trail if you have an SLA. Pointing to a ticket saying "this is when we knew" makes time calculation for compensation very trivial and transparent. Of course if your salesforce account manager is like most, they don't need such "evidence" but in the past it has helped me.
We host our app on Heroku - the core service and a few microservices. The heroku dashboard is down, and we can't use the cli due to the in-browser auth callback that it needs (which is also down).
The status page itself is either saying nothing is wrong, or points to an error page[0]. The incident itself[1] hasn't been updated, which is pretty frustrating.
We can't submit a support ticket because, well, it requires the authentication procedure as well.
We use worker queues, and the queues are getting blown out because heroku can't action anything. We're having our microservices yo-yo now, which suggests things are getting worse, now better.
I've always been a huge Heroku advocate, but the last 5 years have been death by a thousand cuts.
This feels like the last paper cut for me. It's easy enough to keep using Heroku because it's "easy for the team," even though it is overpriced. But when they can't even acknowledge an incident and there is no way to log in and file a ticket, it feels like they have failed.
Which is odd, heroku I'd think would be pretty good at keeping it's status page infrastructure separate enough to stay up. Must mean something pretty fundamental in their architecture is malfunctioning. :(
but when I am able to see the error page, it did say "Heroku continues to investigate and remediate an issue with intermittent outages" -- I would say it is acknolwedged. Yes, that message is 3 hours old. The fact that it's taking them over 3 hours to fix is disturbing, but getting contant progress communication isn't really urgent for me -- I know they know about it, I know they are working to fix it, I'd like them to fix it _quicker_ but I don't need a play-by-play, "can't even acknowledge an incident" is NOT a problem being exhibited, it's acknowledged.
We'll wait and see what it was. A good retrospective write-up goes a long way to increasing many people's confidence, including mine.
In general, I was getting death-by-a-thousand-cuts pessmisim about heroku, but I'm really optimistic about the new Fir stack, which will fix some of my biggest complaints.
Stability definitely still matters though, of course.
Not sure what country you're from, but "blockout blinds" are likely what you're looking for. They blockout (essentially) all light and are operated like normal blinds.
Have never seen blockout blinds which stop enough light during the day - they leak enough to be comparable to a reading lamp, since outdoors is so bright.
I used to work night shift and tried a few different things. Maybe the super-premium blinds block more light, but IME the most effective solution is blackout curtains. You'll need to hang a curtain rod that's wider than the window and get curtains that are wide enough, ideally just a single curtain rather than 2 if your window is narrow enough.
Mounted internal to the window frame, not external, works better for me. Internal can ride tighter to the window, so light can't go out the edges. With external frame mounting, you need much wider shades.
If the fabric itself isn't blocking light... You need better material. I have only ever had problem with light leakage in the edges, not in the fabric material.
I believe 'blackout thermal shades' is what to look for.
I have that. They have a magnet that keep the middle, where they connect light-tight.
Installation is key - they need to be oversized, covering the entire window AND the trim, and they need to be carefully installed so that they touch the trim.
The biggest difference is GitHub in your infrastructure is (nearly always) internal. Fly in your infrastructure is external. Users generally don't see when you have issues with GitHub, but they do generally see when you have issues with Fly.
There are many instances (where I am from, at least - and I believe in the USA), where 'accidents' happen and individuals are found not guilty. As long as you can prove that it wasn't due to negligence. Could "don't be an asshole" as instructions be enough in some arenas to prove they aren't negligent? I believe so.