What it seems like has happened, is that most or all Product Manager oversight was removed from the Heroku project, and an engineering team was given ownership of the whole thing, for the purpose of ongoing maintenance.
But, paradoxically, this has given those engineers free rein to make whatever improvements they deem fit - including things they may have been blocked from working on in the past due to Product meddling and/or corporate bureaucracy.
(Not speaking authoritatively - this situation just, from the outside, appears to have a lot of parallels to teams I've been on that owned "Legacy" services.)
I have built Cuber (https://github.com/cuber-cloud/cuber-gem) a few years ago as a replacement for Heroku and now we use it to deploy all our Rails applications on DigitalOcean Kubernetes. Extremely lower cost, better performance, less bugs, better support...
The blog author isn’t understanding it but it’s quite simple: the product only matters in the context of large enterprise customers.
The large customers still get what they want as long as the ask isn’t too big and that’s why you see new features even though the product is in maintenance mode.
I understand Judoscale is a customer with apprehensions and is asking for clarity. That will definitely raise anxiety.
However, Heroku said they were changing focus. It’s entirely possible to change focus away from something and still do some of it. A focus on things other than new features doesn’t mean, necessarily, no new features at all. Heroku could probably save their customers and partners a lot of anxiety by being clearer and more explicit what they mean.
What a weird article that's microanalysing language in Heroku's blog posts. I mean times are such that pivot-churn is becoming business as usual for most outfits these days so I wouldn't put any stock on C-Suite verbiage.
I think the downhill slide started when they introduced the "Private Space Peering". It is a wrapper on top of AWS VPC, but it was something like $1000 a month several years ago. It also was gating larger instances and other important features.
So few people used it. I guess this provided a negative signal to their management about the adoption rate of new features. And then everything eventually just died.
It’s just in coma, slowly dying away on a respirator. Some relatives irrationally keep paying the hospital to keep the patient alive, but the doctors just wait until they can finally pull the plugs and use the bed for someone with actual chances of survival.
I think its impossible for the Herokus and the digital oceans of the world to survive in the cloud world. They might be able to create better experience for customers but noone can match the networking that AWS, GCP and Azure can provide. Low latency will always win over better developer experience.
True, it can't compete with AWS/GCP/Azure if you're large scale. But most of us are not large scale, we just need a no frills experience instead of dealing with 27 nested panels just to spin up a VM.
Heroku runs on AWS though, doesn’t it? They just package it.
I don’t think it’s impossible for them to survive. Salesforce bought them more than 10 years ago and did little to support growth of Heroku. And yet they’re still around and people still ask „is there something new with comparable customer experience?” because they don’t mind paying more
on the other hand modern tech stacks can process insane amounts of req/s for typical websites/services in a single shared vserver core. not your 2010 ruby snoozefest anymore. plus I can't even remember when a few decade old droplets needed anything from me and still host some things just fine with zero issues or friction or nagging at all. DO is the number one pick for me in 2026 still when the problem fits a droplet style deployment, full stop.
I've never found cloud anything to beat the speed (and price) of a well placed server.
DO has always been a bit rich for my blood though, and even a low cost hetzner VPS has less cores than I remember seeing at the same price a decade ago. I could be wrong there though I usually use Vultr for their SYD region.
What it seems like has happened, is that most or all Product Manager oversight was removed from the Heroku project, and an engineering team was given ownership of the whole thing, for the purpose of ongoing maintenance.
But, paradoxically, this has given those engineers free rein to make whatever improvements they deem fit - including things they may have been blocked from working on in the past due to Product meddling and/or corporate bureaucracy.
(Not speaking authoritatively - this situation just, from the outside, appears to have a lot of parallels to teams I've been on that owned "Legacy" services.)
Five bucks it’s this:
Management: “we’re going into maintenance mode”
Devs: “You mean we get to work on whatever we want?!”
100% management is gone, engineers took over
Next week: “we are right-sizing the organization”
[dead]
I have built Cuber (https://github.com/cuber-cloud/cuber-gem) a few years ago as a replacement for Heroku and now we use it to deploy all our Rails applications on DigitalOcean Kubernetes. Extremely lower cost, better performance, less bugs, better support...
Is Heroku that expensive these days? What are extremely lower costs? In my experience DO is hella expensive.
Northflank is by far the cheapest alternative
The related discussion on one of the mentioned blog posts:
An update of Heroku
https://news.ycombinator.com/item?id=46913903
The blog author isn’t understanding it but it’s quite simple: the product only matters in the context of large enterprise customers.
The large customers still get what they want as long as the ask isn’t too big and that’s why you see new features even though the product is in maintenance mode.
I long ago switched to Fly.io. It feels like the old days of Heroku.
They might be dumping the last of the stuff that was already in the pipeline.
I don't understand who is still using Heroku today.
The experience of "you push, provision databases & dependencies in 3 clicks, and it just works" is sadly still unmatched.
Yep, even in the things that tried to copy Heroku
Maybe true in 2010 but not today.
2 replies →
I understand Judoscale is a customer with apprehensions and is asking for clarity. That will definitely raise anxiety.
However, Heroku said they were changing focus. It’s entirely possible to change focus away from something and still do some of it. A focus on things other than new features doesn’t mean, necessarily, no new features at all. Heroku could probably save their customers and partners a lot of anxiety by being clearer and more explicit what they mean.
What a weird article that's microanalysing language in Heroku's blog posts. I mean times are such that pivot-churn is becoming business as usual for most outfits these days so I wouldn't put any stock on C-Suite verbiage.
It's an ad for the author's service, that's how all of these engineering blogs generally are.
Heroku > Vercel. Try again HN
???
[dead]
[flagged]
[flagged]
Where have you been the last year? It’s much more explicit elsewhere on social media.
Heroku has been going downhill ever since Salesforce bought them.
I can't believe that it has been over 15 years ... https://news.ycombinator.com/item?id=1982489
Strong disagree. They didn’t even invent buildpacks until 2011, the year after the acquisition.
Momentum??
2 replies →
Actually the opposite: they came into their prime after the acquisition. Probably not due to Salesforce, but still.
I think the downhill slide started when they introduced the "Private Space Peering". It is a wrapper on top of AWS VPC, but it was something like $1000 a month several years ago. It also was gating larger instances and other important features.
So few people used it. I guess this provided a negative signal to their management about the adoption rate of new features. And then everything eventually just died.
It’s just in coma, slowly dying away on a respirator. Some relatives irrationally keep paying the hospital to keep the patient alive, but the doctors just wait until they can finally pull the plugs and use the bed for someone with actual chances of survival.
I think its impossible for the Herokus and the digital oceans of the world to survive in the cloud world. They might be able to create better experience for customers but noone can match the networking that AWS, GCP and Azure can provide. Low latency will always win over better developer experience.
DigitalOcean is the Arduino of cloud.
True, it can't compete with AWS/GCP/Azure if you're large scale. But most of us are not large scale, we just need a no frills experience instead of dealing with 27 nested panels just to spin up a VM.
Heroku runs on AWS though, doesn’t it? They just package it.
I don’t think it’s impossible for them to survive. Salesforce bought them more than 10 years ago and did little to support growth of Heroku. And yet they’re still around and people still ask „is there something new with comparable customer experience?” because they don’t mind paying more
Vercel is the chosen replacement, I believe.
1 reply →
on the other hand modern tech stacks can process insane amounts of req/s for typical websites/services in a single shared vserver core. not your 2010 ruby snoozefest anymore. plus I can't even remember when a few decade old droplets needed anything from me and still host some things just fine with zero issues or friction or nagging at all. DO is the number one pick for me in 2026 still when the problem fits a droplet style deployment, full stop.
I've never found cloud anything to beat the speed (and price) of a well placed server.
DO has always been a bit rich for my blood though, and even a low cost hetzner VPS has less cores than I remember seeing at the same price a decade ago. I could be wrong there though I usually use Vultr for their SYD region.
> has less cores than I remember seeing at the same price a decade ago
Less cores but probably 5x more performance per core now.
5 replies →