I actually think it’s the opposite. I think people will move back to their own hardware with OSS deployment like dokku. If everything is more bespoke the attack surface is smaller as is the pay off. Hacking railway got access to lots of api keys, hacking some random vps gets less payoff.
How is that ever going to work out? The speed of discovery for LLMs is so fast, these places are getting breached more often than not due to a human errors. Its really hard for machines, let alone humans, to guard against a lot of LLMs going against them 24x7.
I have a bridge to sell you if you think everyone on the Azure team is sub-moron. Infra security is a hard problem to solve, GCP and AWS don't quite have a perfect track record either, perhaps AWS stands out as a better option but it's hard to argue about it.
I don't think the blame lies with the intelligence of the Azure team. I would argue that there exists structural flaws in the business incentives which deprioritize 'users' from 'customers', along with the stark lack of consequences for security failures.
It didn't help that azure (especially in the beginning) was a cobbled-together mess of ancient Microsoft tech stacks that were never intended to talk to each other.
Yeah, if Exchange goes down, you get a day off. If the little boutique email provider you chose goes offline, you get a day in calls with everyone yelling at you.
Azure (and AWS) offer Gov clouds, and probably on prem air-gapped deployment of their full stack of Fabric for that type of NEED security customers who can pay for it.
It is trade-off between risk and access/productivity with a cost dimension always.
Depending on your threat model you could go to typewriters /paper as some Russian department did famously few years back too.
With a lot of PaaS type vendors you are effectively exposed to both the vendor themselves and the public cloud they run their stack on top of.
Broadly I think you're much more likely to be the victim of a platform-level breach involving an immature startup PaaS than a hyperscaler cloud, however I think for small teams the alternate risk you take on is that AWS/GCP/Azure are full of landmines and extremely easy to misconfigure in insecure ways if you don't know what you're doing. Especially if you have AI telling you confidently wrong things.
When I was at Microsoft we took security cripplingly seriously to the point we could not deliver features and our competitors could because the security process was convoluted and had multiple reviews and approval steps.
Having crippling bureaucracy does not correlate with good security outcomes, except maybe negatively. However, having a history of unending egregious security failures, like Microsoft, does correlate with poor security outcomes. For that matter, having a present of unending egregious security failures also correlates with poor security outcomes.
Having crippling bureaucracy does have better correlation with security outcomes than the “move fast, break things” philosophy that most of the software industry lives by.
I’m not even blaming Zuck for that, he didn’t really invent it, though he certainly popularized it — he just made an astute observation of the dynamics in software and turned it into a company motto.
Neither is correlated with good security outcomes against the prevailing threat landscape of profit-motivated, commercial, professional attackers with tiny budgets of a few million dollars. The standard is not "better", it is "adequate".
Correlation is not an absolute measure here, it’s relative. As is the entire concept of security.
As is what constitutes the standard of “adequate.” Which BTW is the standard in ALL engineering (not just security or software) because we live in a world where economic pressures apply and costs must be managed. Trivial example: bridges only use an “adequate” amount of steel and concrete because they are not free. The word “adequate” fundamentally involves the concept of tolerances. There’s a reason we consider over-engineering a bad thing.
So which development approach would you say is better correlated with good security outcomes?
I do not see how that is relevant. Are you saying that trivial cross-tenant compromises are not egregious security failures demonstrating their systems are unfit for purpose because other companys make products that have even more egregious security failures that are even more unfit for purpose?
My bridge that collapses is adequate and acceptable because other people make even worse bridges that collapse even easier? No, the bridge still collapses, everybody gets a F.
Practically every software security firm in the country was a vendor to Microsoft in the mid-late oughts and I know the Microsoft you're speaking about, but the vibe I've got lately is that things are different there now.
Yes and no. They (or anybody else) are not guaranteed to be secure, of course; see all the compromises regularly found in cloud providers; Azure especially has had so many cross-tenant vulnerabilities (there’s one on the front page right now!) that I’m amazed the concept of public multi-tenant clouds still exists.
But they are also very highly incentivized to be more secure. They have millions of customers (hence a correspondingly higher expectation of trust) and billions in revenue from the trillions in economic activity happening on those platforms (hence a correspondingly larger target on their backs.) This means exponentially higher liabilities. And that probably matters more than any other factor.
This will be the downfall of the torrent protocol. If you think governments (US and China) are going to tolerate people downloading copyrighted games, movies and music without paying the rightsholder, I have a datacenter to sell you.
SOTA open models need tens of thousands of dollars worth of hardware to run.
I get your analogy, and it feels right, but the compute requirements to run these models are extreme. That analogy falls apart when it necessitates having a corporate 3rd party run the models for you. They will not tolerate breaking laws.
If China says no more, then the best the masses are gonna get on their high-end setups are the 30B class models. Which are decent, but not "hack my bank", errr, "protect our bank" decent.
Piracy flourished because you only need an operating laptop to participate. LLMs you need a server rack of GPUs to participate.
I have news for you. A lot of enterprises are moving a lot of workloads back on prem and are expanding their datacenters and on prem GPU compute. Also replacing a lot of SaaS with internal tooling.
I ran some stats on producthunt a few months ago, there are 22 times more daily SaaS launches now than in 2020, the AI impact on the industry is hard to fathom. The blog I wrote (AI-assisted) if you're curious: https://www.mahl.me/blog/product-hunt-is-swamped-now/
I feel like these breaches started to accelerate after Anthropic wrote a full post about how easy it is to abliterate glm 5.3 to make a frontier-level cyber model.
Now, to be clear, abliteration was no state secret, but that blog post was pure accelerationism.
That was just 30 days ago. This has been happening for months. Part of the "Black October" narrative is just vulnerability research; Calif didn't abliterate 5.3 to find PHP2Shell.
Yeah IIRC calif research on php2shell preceded all of that but I still stand by the thought that Anthropic has been the most accelerationist AI company despite its lip service to the notion of safety.
The idea that attackers were waiting for Anthropic to tip them off on this is kind of risible. It's been a huge topic amongst practitioners for the better part of a year, and practitioners are usually a step or two behind dedicated criminal or state-sponsored attackers.
Right, that's kind of my point. Also: it's a good blog post (the Anthropic one), but it observes abliterated 5.3 models were already public. It's not like people who break into computers didn't know that already. They're not waiting for Anthropic to tell them about open weight models; they already live in the world of open weight models.
You are running a machine with tons of vendor provided agents/scripts to do god knows what on a cloud provider which itself runs tons of agents/scripts to manage that infrastructure and then you expect to be secure.
And no, before the labs there was a degree of fear for everyone to hack into some one else system. With the labs changing the narrative, anyone can just blame the agents for the breach, whoever did it would only get fame and no blame
How these Paas (Railway before) are being breached? Poor choices or something else? Scary Nonetheless. I thought AI advancement will make the self hosting painful. If PaaS are no good, then what can be done?
Stuff made during the move fast and break things era now doesn't hold up very well in the AI era. Or to say another way: startups are very vulnerable now - this new environment is unforgiving towards cutting corners and/or having inexperienced devs
Modern PaaS is tough because there's a general expectation that users (and agents) are able to see and read back environment variables in their management UI or CLI - which means that you can't segregate credential values from the public web environment. See e.g. the visibility icons on https://northflank.com/docs/v1/application/secure/inject-sec...
And sure, you can encrypt at rest, and perhaps even enforce that only re-encrypted versions are ever seen by the management/API web server. But fundamentally this display requirement means that if your management/API web server is compromised, users' secrets can be obtained by a process of impersonation.
Why would ai advancement make self hosting painful? Apart from increased cost of hardware (which you implicitly would pay for in PaaS as well anyways).
If anything, AI makes running linux boxes much easier than 4 years ago; it's now faster and the learning cliff is gone to get the correct CLI commands and digging into system logs.
When you build something small, you can easily see all the doors and properly guard them.
As you build something bigger and bigger, quickly, with several people building, guarding all the doors becomes exceptionally difficult without a kind of aggressive discipline which is contrary to the "go fast" culture.
Cybersecurity in practice until now, for most companies, has been an insurance problem, not a quality problem.
Such companies are quickly discovering that insurance is a post-hoc partial correction and not a proper solution.
This situation will not improve anytime soon for most companies.
The solution does not require any new magical tools.
Run agents and everything they build. Secure execution from the first sandbox
Getting hacked while in the business of selling a secure sandbox isn't really a good look. It seems like can't even trust anything these days. The other day OpenCode got hacked and leaked everyone's secrets inside .env to attackers and now this. Sigh.
Unfortunately I'm not longer using PaaS tools. Instead I use Azure for work and vps now for all my personal stuff. With AI it's easier than ever to lock a server down tight and mitigate this kind of exposure.
I'm actually pretty pissed at this now. I have a throwaway appstore app that uses northflank for the backend, now I have to go in and rotate revenuecat secrets, apple secrets, ai secrets. For an app with basically 2 users.
Can I even delete the app on apple's app store or whats the process like what a pain in the ass. I'm pretty done tbh.
Unfortunately for infra startups, I think we're going to see a convergence back to hyperscalers as a result of incidents like this and railway.
There's no amount of "nicer to use" that can replace security, at least certainly when selling to enterprises
I actually think it’s the opposite. I think people will move back to their own hardware with OSS deployment like dokku. If everything is more bespoke the attack surface is smaller as is the pay off. Hacking railway got access to lots of api keys, hacking some random vps gets less payoff.
How is that ever going to work out? The speed of discovery for LLMs is so fast, these places are getting breached more often than not due to a human errors. Its really hard for machines, let alone humans, to guard against a lot of LLMs going against them 24x7.
> There's no amount of "nicer to use" that can replace security, at least certainly when selling to enterprises
Azure would beg to differ.
https://news.ycombinator.com/item?id=50301854
A moron could build a better UX and DevX.
I have a bridge to sell you if you think everyone on the Azure team is sub-moron. Infra security is a hard problem to solve, GCP and AWS don't quite have a perfect track record either, perhaps AWS stands out as a better option but it's hard to argue about it.
I don't think the blame lies with the intelligence of the Azure team. I would argue that there exists structural flaws in the business incentives which deprioritize 'users' from 'customers', along with the stark lack of consequences for security failures.
It didn't help that azure (especially in the beginning) was a cobbled-together mess of ancient Microsoft tech stacks that were never intended to talk to each other.
Are hyperscalers actually more secure though? I have only heard anecdotal thinking eg nobody ever got fired for choosing IBM.
It’s the same thing: let’s say Microsoft or AWS got hacked. Then you’re definitely not the only one who got hacked.
So, you can always say: Look, it’s not us (especially not my org!). It f’ing Microsoft/AWS! Everyone got hacked.
You and your career are gonna be just fine.
The whole "No one gets fired for using IBM" analogy.
Yeah, if Exchange goes down, you get a day off. If the little boutique email provider you chose goes offline, you get a day in calls with everyone yelling at you.
If you need deniability, sure.
But it you actually NEED security, it's clearly not an option.
Azure (and AWS) offer Gov clouds, and probably on prem air-gapped deployment of their full stack of Fabric for that type of NEED security customers who can pay for it.
It is trade-off between risk and access/productivity with a cost dimension always.
Depending on your threat model you could go to typewriters /paper as some Russian department did famously few years back too.
With a lot of PaaS type vendors you are effectively exposed to both the vendor themselves and the public cloud they run their stack on top of.
Broadly I think you're much more likely to be the victim of a platform-level breach involving an immature startup PaaS than a hyperscaler cloud, however I think for small teams the alternate risk you take on is that AWS/GCP/Azure are full of landmines and extremely easy to misconfigure in insecure ways if you don't know what you're doing. Especially if you have AI telling you confidently wrong things.
When I was at Microsoft we took security cripplingly seriously to the point we could not deliver features and our competitors could because the security process was convoluted and had multiple reviews and approval steps.
Having crippling bureaucracy does not correlate with good security outcomes, except maybe negatively. However, having a history of unending egregious security failures, like Microsoft, does correlate with poor security outcomes. For that matter, having a present of unending egregious security failures also correlates with poor security outcomes.
Let us look on the front page and see if there are any, oh look here is one: https://www.binarysecurity.no/posts/2026/10/one-root-case
Having crippling bureaucracy does have better correlation with security outcomes than the “move fast, break things” philosophy that most of the software industry lives by.
I’m not even blaming Zuck for that, he didn’t really invent it, though he certainly popularized it — he just made an astute observation of the dynamics in software and turned it into a company motto.
Neither is correlated with good security outcomes against the prevailing threat landscape of profit-motivated, commercial, professional attackers with tiny budgets of a few million dollars. The standard is not "better", it is "adequate".
Correlation is not an absolute measure here, it’s relative. As is the entire concept of security.
As is what constitutes the standard of “adequate.” Which BTW is the standard in ALL engineering (not just security or software) because we live in a world where economic pressures apply and costs must be managed. Trivial example: bridges only use an “adequate” amount of steel and concrete because they are not free. The word “adequate” fundamentally involves the concept of tolerances. There’s a reason we consider over-engineering a bad thing.
So which development approach would you say is better correlated with good security outcomes?
the only reason you’re seeing this post is because of the large bug bounty program that actors like Northflank don’t even operate
I do not see how that is relevant. Are you saying that trivial cross-tenant compromises are not egregious security failures demonstrating their systems are unfit for purpose because other companys make products that have even more egregious security failures that are even more unfit for purpose?
My bridge that collapses is adequate and acceptable because other people make even worse bridges that collapse even easier? No, the bridge still collapses, everybody gets a F.
Practically every software security firm in the country was a vendor to Microsoft in the mid-late oughts and I know the Microsoft you're speaking about, but the vibe I've got lately is that things are different there now.
Tell me you weren’t on the Windows team without telling me you weren’t on the windows team :)
Yes and no. They (or anybody else) are not guaranteed to be secure, of course; see all the compromises regularly found in cloud providers; Azure especially has had so many cross-tenant vulnerabilities (there’s one on the front page right now!) that I’m amazed the concept of public multi-tenant clouds still exists.
But they are also very highly incentivized to be more secure. They have millions of customers (hence a correspondingly higher expectation of trust) and billions in revenue from the trillions in economic activity happening on those platforms (hence a correspondingly larger target on their backs.) This means exponentially higher liabilities. And that probably matters more than any other factor.
> nobody ever got fired
thats the crux; are you working for the paycheck, or for the truth?
could be both, but pretty much all "i was right" stories start with "everyone else said i was wrong"
For me, the truth. I’d like the most secure provider possible. Which one is it though?
Theres always a market for seedbox providers. Hyperscalers will always wear the white gloves and enable this market.
It's also going to be the downfall of open models.
If you think governments (US and China) are going to tolerate people using models for anti-social behavior, I have a datacenter to sell you.
This will be the downfall of the torrent protocol. If you think governments (US and China) are going to tolerate people downloading copyrighted games, movies and music without paying the rightsholder, I have a datacenter to sell you.
Pirated content can run on a $20 computer.
SOTA open models need tens of thousands of dollars worth of hardware to run.
I get your analogy, and it feels right, but the compute requirements to run these models are extreme. That analogy falls apart when it necessitates having a corporate 3rd party run the models for you. They will not tolerate breaking laws.
If China says no more, then the best the masses are gonna get on their high-end setups are the 30B class models. Which are decent, but not "hack my bank", errr, "protect our bank" decent.
Piracy flourished because you only need an operating laptop to participate. LLMs you need a server rack of GPUs to participate.
I have news for you. A lot of enterprises are moving a lot of workloads back on prem and are expanding their datacenters and on prem GPU compute. Also replacing a lot of SaaS with internal tooling.
They are also getting hacked, and they are much jucier targets.
>There's no amount of "nicer to use" that can replace security, at least certainly when selling to enterprises
We can lower our standards instead
What is Northflank? I am surprised at the amount of new business out there
I ran some stats on producthunt a few months ago, there are 22 times more daily SaaS launches now than in 2020, the AI impact on the industry is hard to fathom. The blog I wrote (AI-assisted) if you're curious: https://www.mahl.me/blog/product-hunt-is-swamped-now/
Northflank isn't that new, it was founded in 2019. Its a cloud platform, you can get more details by googling.
You are on a YC startup forum and are surprised tech startups exist?
"amount" it says. No matter the context, there could be more than you expect.
New businesses, not new business.
I feel like these breaches started to accelerate after Anthropic wrote a full post about how easy it is to abliterate glm 5.3 to make a frontier-level cyber model.
Now, to be clear, abliteration was no state secret, but that blog post was pure accelerationism.
That was just 30 days ago. This has been happening for months. Part of the "Black October" narrative is just vulnerability research; Calif didn't abliterate 5.3 to find PHP2Shell.
Yeah IIRC calif research on php2shell preceded all of that but I still stand by the thought that Anthropic has been the most accelerationist AI company despite its lip service to the notion of safety.
I don't have a take about Anthropic generally. The blog post was quite good. The blog post did not cause October's security debacles.
It was a good blog post yes, but I think you’re underplaying the bannister effect of Anthropic coming out and endorsing glm 5.3’s cyberattack prowess.
The idea that attackers were waiting for Anthropic to tip them off on this is kind of risible. It's been a huge topic amongst practitioners for the better part of a year, and practitioners are usually a step or two behind dedicated criminal or state-sponsored attackers.
They are using mythos-5.5, which is still better than an abliterated GLM-5.3.
Right, that's kind of my point. Also: it's a good blog post (the Anthropic one), but it observes abliterated 5.3 models were already public. It's not like people who break into computers didn't know that already. They're not waiting for Anthropic to tell them about open weight models; they already live in the world of open weight models.
You are running a machine with tons of vendor provided agents/scripts to do god knows what on a cloud provider which itself runs tons of agents/scripts to manage that infrastructure and then you expect to be secure.
And no, before the labs there was a degree of fear for everyone to hack into some one else system. With the labs changing the narrative, anyone can just blame the agents for the breach, whoever did it would only get fame and no blame
lol no, abliteration communities have existed and abliterated models routinely published long before Anthropic wrote that post last month.
What’s changing now is that models are getting more powerful.
How these Paas (Railway before) are being breached? Poor choices or something else? Scary Nonetheless. I thought AI advancement will make the self hosting painful. If PaaS are no good, then what can be done?
Stuff made during the move fast and break things era now doesn't hold up very well in the AI era. Or to say another way: startups are very vulnerable now - this new environment is unforgiving towards cutting corners and/or having inexperienced devs
Nope. The solution is to move even faster and thus break less things via AI.
The "move faster and break things" have been unbundled. SaaS moves faster. Hackers break things.
For new stuff, maybe. I'm talking about existing startups now absorbing the onslaught of AI.
Modern PaaS is tough because there's a general expectation that users (and agents) are able to see and read back environment variables in their management UI or CLI - which means that you can't segregate credential values from the public web environment. See e.g. the visibility icons on https://northflank.com/docs/v1/application/secure/inject-sec...
And sure, you can encrypt at rest, and perhaps even enforce that only re-encrypted versions are ever seen by the management/API web server. But fundamentally this display requirement means that if your management/API web server is compromised, users' secrets can be obtained by a process of impersonation.
Why would ai advancement make self hosting painful? Apart from increased cost of hardware (which you implicitly would pay for in PaaS as well anyways).
If anything, AI makes running linux boxes much easier than 4 years ago; it's now faster and the learning cliff is gone to get the correct CLI commands and digging into system logs.
platforms are better placed to guard against attackers
Certainly. They are also larger targets.
to me, this is a variant of security by obscurity and will not fare well in the LLM age
if you're still doing that manually in 2026 you probably should not be self hosting anything of consequence or commercial liability
What's "that"? Running your own hardware, using a CLI to manage servers, or touching anything below your application layer?
According to HN of the last year, "marketing stunts." Remember when Mythos preview was announced only 6 months ago
When you build something small, you can easily see all the doors and properly guard them.
As you build something bigger and bigger, quickly, with several people building, guarding all the doors becomes exceptionally difficult without a kind of aggressive discipline which is contrary to the "go fast" culture.
Cybersecurity in practice until now, for most companies, has been an insurance problem, not a quality problem.
Such companies are quickly discovering that insurance is a post-hoc partial correction and not a proper solution.
This situation will not improve anytime soon for most companies.
The solution does not require any new magical tools.
Does anyone know to what degree Sentry use Northflank?
Whatcha mean. If they do, it would likely be in Sentry’s list of subprocessors. Check there.
Good point. I don't see Northflank listed, https://sentry.io/legal/subprocessors/.
If I go on wayback a few days, I can see Sentry: https://web.archive.org/web/20261026032756/https://northflan...
Sentry is listed on the northflank site as a customer.
They've updated their website to remove their customers
i'm surprised at the growth of all these neoclouds given the current security risk climate...
So sad...
Unfortunately I'm not longer using PaaS tools. Instead I use Azure for work and vps now for all my personal stuff. With AI it's easier than ever to lock a server down tight and mitigate this kind of exposure.
I'm actually pretty pissed at this now. I have a throwaway appstore app that uses northflank for the backend, now I have to go in and rotate revenuecat secrets, apple secrets, ai secrets. For an app with basically 2 users.
Can I even delete the app on apple's app store or whats the process like what a pain in the ass. I'm pretty done tbh.
Azure? Really?