parliament32

The future is already pretty much here. Note the recent stories about Chrome adding a local model, not to mention the Googlebook demo (if it works as advertised, there's a 0% chance you could get that kind of latency with a non-local model).

They will still need to increase costs for enterprise to be profitable, they're just going to be more greasy about it. Claude 5 will cost 20% more but not be 20% better, more shenanigans with "oh no we had a bug in our cache system :^)", or this gem from the current enterprise pricing page: "Opus 4.7 uses a new tokenizer... may use up to 35% more tokens for the same fixed text".

Moviepass (afaik) was an attempt at the exact same playbook, it just failed.

Anthropic will also fail when the competition is.. near-equivalent-capability DeepSeek/Qwen/Llama on a $1k GPU with a break-even of 5 months of subscription costs. The value is simply not there for what they would need to charge to become profitable.

It's okay, at this rate Anthropic will be the only ones left using Bun.

This is the Extinguish phase of the process, right?

Given how terrible their in-house support bot is, I don't think they've figured out jack shit. If you truly had a half-decent LLM, you'd think the first place you'd put it to demonstrate the capabilities is handling support.

It really is the doordash/uber playbook all over again eh? Sell at a massive loss, gain userbase, then gradually boil the frog by adding fees, removing features, and increasing prices. Except instead of doing this a few years down the line, they're speedrunning the tighten-the-noose phase.

Unfortunately the competition is nipping at their heels so there's a good chance this blows up in their faces.

Well, in an ideal scenario, you would have exported anything you wanted to keep, and you'd import it if you subscribe again.

How long would you expect them to keep your data for? Do you really expect them to pay storage costs for your data indefinitely just because you paid them $20 once upon a time?

And on the inverse side, would you really want your data to be compromised when they inevitably get breached just because you had a sub there once?

These are the reasons data retention policies exist.

I guess it's not a termination, but a downgrade to the "free" tier. But that still makes sense, given Claude Design is limited "to Pro, Max, Team, and Enterprise plans". He's not on that plan anymore so.. what commercial reason could they possibly have to keep his data?

Google Workspace seems to halt access immediately[1] and purge data within 60d[2]. For comparison, Atlassian leaves you access for 15d, and purges data at 60d[3]. 365 gives you 90d[4] before purging.

This is a pretty regular thing across the industry.

[1] https://knowledge.workspace.google.com/admin/billing/cancel-...

[2] https://support.google.com/a/thread/345697828/recovering-dat...

[3] https://support.atlassian.com/security-and-access-policies/d...

[4] https://learn.microsoft.com/en-us/compliance/assurance/assur...

So.. you unsubscribed from a SaaS and expected them not to purge your data? Why would that make sense?

Anthropic may be a bunch of skids but it sounds like they did the right thing here. Pretty much all SaaS applications, especially in B2B, are required by compliance to remove customer data within X amount of time at the end of the contractual relationship.

When AI respects my license, sure.

"The customers love our AI so much the DAU is 1:1 with the platform!"

> Many people use the wrong scorecard.

Correct. "Revenue" is the wrong scorecard when they're selling 20$ bills for 15$. I too can make a bajillion dollars in revenue with that strategy.

Show me a company not speed running the uber/doordash playbook and we can talk.

Note Clickhouse is quite old (2010ish?) but they've always been a "web server access log analytics" solution. The pivot to "we do observability too" is new, we'll see how that plays out. Not terribly optimistic given how badly a similar pivot went for Elastic, but who knows.

FWIW we've also tried all sorts of different things, and honestly the very vanilla (prometheus -> central thanos, fluentbit -> central loki, grafana) ends up on top. The resource consumption is surprisingly minimal (for a sense of scale, we run about 200k eps for metrics and 1k eps for logs). For all these solutions, I find myself asking the same question as you.. what problem are you trying to solve? Is there anything actually different about your product other than less stability than the battle-tested stack?

Azure is a compliance requirement for us.

Haven't had anything impacting in GovCloud, but if you're not there yet I'm sure there's shenanigans in the consumer version.

That's one of the things I like about Azure, they don't overwhelm you with listing prices beside every individual item as you're creating it, but they seem to always present a price on things that could be expensive. It's a good balance, I have yet to be surprised by a charge.

It's AI but not in the way you think.

AI usage is getting expensive since Anthropic et al are turning the screws, and that money has to come from somewhere. Reducing AI usage is blasphemy of course, so cutting headcount is the only path forward.

> It should be illegal

It should be illegal to host insecure services, especially when you're dealing with PII. Breaches keep happening and nobody gives a fuck, because the worst that'll happen is you might lose a handful of customers and buy some "credit monitoring".

Incidents like this should be followed by an audit and charges being laid. Send corp officers to jail for negligent security failures. If you can go to jail for accounting fraud, you should be able to go to jail for cybersecurity-promises-fraud.

They claim to be compliant with a number of security standards [1]. I would love to see a postmortem audit of how much of this they actually implemented.

[1] https://www.instructure.com/en-au/trust-center/compliance

Agreed it's a weird comparison, but I'd argue SG&A needs to come out of gross margin too for a fair comparison. You need a warehouse/staff/utilities/etc to sell merchandise, you need nothing to sell a membership (whether it's worth anything is another question of course).

In their 2025 filing, gross margin on merchandise was $30B, but SG&A cost $25B (with membership fees at $5.3B).

Note that $2.6B of those membership fees will go back to members as membership rewards, which is interesting too.

Wait for the EU AI Act to require text watermarking in August. It will work, and it will be effective -- not because it'll be impossible to circumvent, but because all the big SaaSes will have to adopt it, and the hurdle of stripping it back out will filter out the vast majority of the sloppers.

No, they did not. Careful of falling for the psychosis.

> This finding was AI-assisted, but began with an insight from Theori researcher Taeyang Lee, who was studying how the Linux crypto subsystem interacts with page-cache-backed data.

https://xint.io/blog/copy-fail-linux-distributions

It sure has limitations, but "no protective benefits" is pretty wrong. In a real world example, if your containerized application has an RCE, you're preventing the attacker from executing binaries they tampered with or down/up-loaded. Combined with minimal distroless containers, it's a very effective attack surface reduction strategy, and works much better than the legacy scan-occasionally integrity-checking methods (rkhunter et al).

Look at the FedRAMP requirements around integrity protection, then look at how massive the list of complaint products is. I promise, pretty much everyone in regulated environments is. It's so prevelant Azure is even pushing a turnkey solution for k8s https://learn.microsoft.com/en-us/azure/aks/use-azure-linux-...

Remember when GPT-2 was "too dangerous to release"?

I'll believe it when I see it, but it's almost certainly just marketing drivel.

It's curious they're just "monitoring" rather than preventing.

In a serious environment you'd run IPE with dm-verity/fs-verity to ensure binaries are whitelisted and integrity-checked at every execution.

Everyone cranks out endless pages of slop, that everyone else then has to ingest. Anthropic collects a fee from all of you and is the only winner here.

I'm looking forward to the impending crash when the AI providers actually start charging what it costs to run these models. It's going to be a bloodbath, and it's going to be cathartic as fuck.

> I believe they already are slowing down

They certainly have. The only impactful improvement in the last year is "just run it in a loop until it gets it right" lmao

Which, of course, only works as long as the costs are subsidized by the companies vying for market share.

I assume you're talking about a local application? You don't care if a malicious image you downloaded pwns your PC then? Like CVE-2016-3714

> the front page of HN

I think I've "flagged" more links this year than my last 13 years on this site combined. I'm sure it's unproductive and doesn't really do anything, but it makes me feel a touch better. I'm so over the slop I think I'm actually visiting HN markedly less because of it.

On the plus side, there has been a (predictable) uptick in slop-flagging browser extensions over the last few months. Once a good locally-hosted version exists, I think it'll take its rightful place alongside ad blockers for tech-minded folks.

To the web? It's fantastic for the web, these are the kinds of fun projects that make the web a worthwhile place to be. To slop generators? Yes, absolutely harmful, and that's for the best.