Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Temporary Cloudflare accounts for AI agents

blog.cloudflare.com

Post reply on HN
Print view View original

151–160 of 173 posts

Re: Temporary Cloudflare accounts for AI agents

#151

Post by simonw »

jedberg wrote: post #117
simonw wrote: post #47

Looks like Cloudflare still haven't shipped the most valuable possible feature for Cloudflare Workers though: hard billing caps. I want to set a cap of $100/month and know, for sure, that if something untoward happens my apps will all stop serving traffic rather than me getting hit with a bill for $1000s. The safest way to use Workers is on the free tier, which will shut off after 100,000 requests/day: https://develo…

I've dealt with this before. The problem with doing a billing cap is that now you are bringing an out-of-band batch system (billing) in-band. The only way to do a dollar cap is to constantly calculate your bill. Capping it at 100K requests is a lot easier, because it's a single number that can be incremented easily in a distributed way. It's the same reason it took AWS forever to offer billing caps, and even today, t…

I'd be happy for the vendor to eat the difference if their accounting systems let me spend a little over my configured cap, personally.

(I get that they're not happy to do that.)

Re: Temporary Cloudflare accounts for AI agents

#152

Post by manwe150 »

londons_explore wrote: post #120
jedberg wrote: post #117

Earlier quoted context omitted.

I've dealt with this before. The problem with doing a billing cap is that now you are bringing an out-of-band batch system (billing) in-band. The only way to do a dollar cap is to constantly calculate your bill. Capping it at 100K requests is a lot easier, because it's a single number that can be incremented easily in a distributed way. It's the same reason it took AWS forever to offer billing caps, and even today, t…

Seems like a weak excuse... If you build a billing system which can calculate bills with low latency 99.9% of the time, but perhaps fails when there is a net split for example, then you can simply let the company write off the loss for that 0.1% of the time the billing system is down.

This seems the real answer. The question of whether the billing team can set accurate maximums is largely disjoint from the question of whether the engineering team shuts off service precisely at that moment.

Re: Temporary Cloudflare accounts for AI agents

#153

Post by mannanj »

scotty79 wrote: post #98
mannanj wrote: post #87

Earlier quoted context omitted.

Particular model or LLM that you use?

Local ones (through ollama, probably will use something more performant when I move fully to it). When I touched it the last time I was a fan of gemma4 https://ollama.com/library/gemma4 Since then I grew really fond of Qwen3.6 (it's super capable in coding against the DOM) so I'll probably try to move to it with the next iteration. https://ollama.com/library/qwen3.6

Great stuff. Thank you!

Re: Temporary Cloudflare accounts for AI agents

#154

Post by CamperBob2 »

scotty79 wrote: post #121
CamperBob2 wrote: post #102

Earlier quoted context omitted.

So this actually runs within ollama, which you then leave up full-time? Or only when you manually want to refresh your news feed?

The prototype runs through ollama on my laptop while I browse. I usually have few hundreds stories ready from the previous session, as new stories come in they get processed in the background while I browse the old ones. It processes stories faster than I consume them. When I move it to the server I might consider waking it up periodically to pull and analyze new stories and perhaps notify me if something absolutely…

Very cool, so you just time-shift the good content while removing the rest. This program would have an audience of at least two if it were public.

How do the vector embeddings fit into the picture?

Re: Temporary Cloudflare accounts for AI agents

#155

Post by johanyc »

winterqt wrote: post #104
simonw wrote: post #25

Earlier quoted context omitted.

I just tried this out: % npx wrangler deploy --temporary wrangler 4.103.0 ──────────────────── You must accept Cloudflare's Terms of Service (https://www.cloudflare.com/terms/) and Privacy Policy (https://www.cloudflare.com/privacypolicy/) in order to continue. By typing "yes", you agree to these terms. Type "yes" to continue. … yes Solving proof-of-work challenge… Temporary account ready: Account: Educated Celery (c…

Interesting, it’s still live!

Still live lol

Re: Temporary Cloudflare accounts for AI agents

#156

Post by scotty79 »

CamperBob2 wrote: post #154
scotty79 wrote: post #121

Earlier quoted context omitted.

The prototype runs through ollama on my laptop while I browse. I usually have few hundreds stories ready from the previous session, as new stories come in they get processed in the background while I browse the old ones. It processes stories faster than I consume them. When I move it to the server I might consider waking it up periodically to pull and analyze new stories and perhaps notify me if something absolutely…

Very cool, so you just time-shift the good content while removing the rest. This program would have an audience of at least two if it were public. How do the vector embeddings fit into the picture?

I was basically throwing stuff at the wall and seeing what sticks and makes a pretty picture for me. Vector embeddings were just one of the things I threw, in hopes of boosting news items thematically similar to the stuff I already liked. I can't really tell how much good or bad are they doing. All I know is that they are fast (embeddinggemma through ollama) and their output is worth few points in some combined heuristics that AI concocted for me.

Re: Temporary Cloudflare accounts for AI agents

#157

Post by Anamon »

zuzululu wrote: post #95
gwittel wrote: post #50

Earlier quoted context omitted.

Given how little they do now to stop malicious content hosted behind/by Cloudflare, the bare minimum if anything.

Cloudflare support is pretty active and very responsive to incidents. Here's a token of appreciation.

I report several phishing sites to them per week. They take ages to respond, often close to a week. And often, the response will be that they couldn't find any abuse, because I also report to the host which is often much quicker to respond.

Any process that doesn't take down phishing sites within a few hours at most is inadequate for protecting potential victims. Especially when I compare it to all the other providers I report abuses to, Cloudflare doesn't strike me as a company that takes abuse reports particularly seriously.

Re: Temporary Cloudflare accounts for AI agents

#158

Post by Anamon »

IncreasePosts wrote: post #60
Hackbraten wrote: post #15

Cloudflare: let's give the bots their own accounts so they can scrape harder. Also Cloudflare: let's send normal humans who are trying to go about their daily lives into endless Turnstile spinner loops with absolutely zero recourse, grievance, or support infrastructure.

I'm sure they don't want humans to have that experience - the issue is that the human behavior looks very bot-like. This is usually only experienced by people whose setup is peculiar

Yes, and using Firefox or having an ad blocker installed counts as extremely peculiar these days.

Re: Temporary Cloudflare accounts for AI agents

#159

Post by Anamon »

bornfreddy wrote: post #68
Hackbraten wrote: post #15

Cloudflare: let's give the bots their own accounts so they can scrape harder. Also Cloudflare: let's send normal humans who are trying to go about their daily lives into endless Turnstile spinner loops with absolutely zero recourse, grievance, or support infrastructure.

Just had 10 rounds of busses, motorcycles and fire hydrants with Google before I decided I don't actually want to see that page so much. So Cloudflare is unfortunately not the only offender here.

There's a certain type of Recaptcha challenge (the 4x4 single-photo one) that I have a 100% failure rate on. If Google decide to serve me 15 of those in a row, I will invariably fail 15 times until they decide to serve one of the other types for round 16. I'm very close to believing that that challenge is actually bugged and unsolvable for everyone, and simply nobody ever bothered to fix it.

Re: Temporary Cloudflare accounts for AI agents

#160

Post by zuzululu »

Anamon wrote: post #157
zuzululu wrote: post #95

Earlier quoted context omitted.

Cloudflare support is pretty active and very responsive to incidents. Here's a token of appreciation.

I report several phishing sites to them per week. They take ages to respond, often close to a week. And often, the response will be that they couldn't find any abuse, because I also report to the host which is often much quicker to respond. Any process that doesn't take down phishing sites within a few hours at most is inadequate for protecting potential victims. Especially when I compare it to all the other provider…

impossible to tell if you are trying to defame a company or telling the truth without seeing any evidence to back up your claims
Post reply on HN