I did use AI to organize my ideas but I didn't think it was that bad, I'll modify and make it easier to read.
Anyway, in my test I saw zero requests from any Google UA after multiple Gemini and AI mode prompts that should have triggered grounding, so the working interpretation is that Gemini served from its own index/cache rather than doing a live provider-side fetch. The original phrasing was fuzzier than it should have been.
If this weren't on HN I wouldn't have given this more than a few seconds of reading before switching away. Some examples of phrasing that triggers me:
> attributing hits was a grep, not a guess
> values below are copied from the probe’s log file, not paraphrased
> a User-agent: Claude-User disallow is the live control
> Only Claude-User is the user-initiated retrieval signal
I could go on and on but I won't. Phrasing aside, the text is too structured with many sections and subsections when the intent was clearly more narrative. "I was curious about X and did Y and I am going to tell you about it."
Signals that suggest a human who cares would be: use of the first-person; demonstrated curiosity, humility, and uncertainty; inline hyperlinks; and any kind of personality or opinion.
"Idiolect" is both subtle and distinct: the choice of vocabulary, grammar, phrasing and colloquial metaphors will vary in kind and frequency for everyone like an intellectual signature. You can sometimes tell if someone has been reading too much of a particular author recently just because of the way the author's choice of vocabulary bleeds into their own speech patterns. Sometimes it's a permanent influence.
I wonder if reading so much LLM stuff lately has affected my idiolect and that I write (or worse, think) more machine-like than before...
> I wonder if reading so much LLM stuff lately has affected my idiolect and that I write (or worse, think) more machine-like than before...
Totally of topic ofc, but I always get triggered by the claim that llms are "machine-like".
I'm aware it's a total pet peeve and a lil irrational, but "machine-like" would imply to me that it's thinking like a machine, which in turn implies machine intelligence - which in turn implies they're doing something which they aren't.
I'm not trying to undersell their capabilities. Used well they're able to do a lot of things. But the way they achieve it is by mimicking human dialogue and rhetoric processes to facilitate this process. That's in my opinion anything but machine intelligence. I struggle finding an applicable word for it though
I didn't see your reply until now but "AI" is correctly describing the phenomenon. Most definitions for "artifice" converge around the idea of deception or insincerity.
The term "machine learning" also distinguishes itself from the organic process by authentic intelligence.
In other words, inferring "machine intelligence" is less correct than "artificial intelligence". By definition LLMs are machines pretending to think and they do it well enough to have a writing style.
Added $http_accept and re-ran. None of them use text/markdown. Results:
ChatGPT-User/1.0 text/html,application/xhtml+xml,application/xml;q=0.9,image/avif,image/webp,image/apng,/;q=0.8,application/signed-exchange;v=b3;q=0.9
Claude-User/1.0 /
Perplexity-User/1.0 (empty, no Accept header)
PerplexityBot/1.0 (empty, no Accept header)
ChatGPT sends a Chrome-style Accept string. Claude sends a wildcard. Perplexity sends nothing at all. Gemini didn't fetch in my test.
Also worth noting: Claude-User hit /robots.txt before the page.
Yeah, that the main reason I never use services like Google Cloud if I don't have to, it's impossible to have a hard cap, and anyone pretending to be an expert, is just off.
Google says that they can't provide a hard cap because that would mean shutting down all your services..bla bla, but at least give users the option.
The providers will not ship hard caps because a hard cap is a revenue ceiling. They will ship soft alerts that fire late enough to be useless. The only answer is an independent layer between your config and the provider that does not have that incentive. Building exactly that at Traeco. traeco.dev
We have spend caps at the billing account level and the project level (developer set) in the Gemini API now. There is up to a 10 minute delay in processing everything but this should significantly mitigate the risk here: https://ai.google.dev/gemini-api/docs/billing#tier-spend-cap...
By default, new Tier 1 paid accounts can only spend $250 in a given month.
I just find it extraordinary that the biggest tech company in the world can do cutting edge real time AI for millions of people, run Youtube and of course all the other google services with having literally the smartest people in the world and unlimited resources on board, but still can't keep real time track of the user's current billing and their spending limits, it's all best effort still. Somehow it doesn't add up. (Pun not intended, but I'm happy to have it)
Not just Google, also Microsoft and Amazon. Real-time cost tracking is technically impossible to solve according to the major cloud providers. I have huge respect for those sales & finops engineers.
Technically impossible for them because real-time attribution cuts into margin. Config-layer tracking does not have that constraint. You do not need real-time billing data if you know from the agent config that it is going to explode before it runs. That is the distinction Traeco is built on. traeco.dev
I'm sure it's me being an idiot, but once again I spent 20m trying to figure how to do a specific thing in google-land and still haven't figured it out. Even if I did set it somewhere, I see things like "Setting a budget does not cap resource or API consumption" with a link to a bunch of documentation I have to analyze.
It shouldnt mean shutting down all your services, it should mean not letting you provision new ones and limiting the scope of what you can continue doing.
If you have a lambda set up that normally runs a hundred times a day, and suddenly it tries to spin up 10 million instances, it should block that unless you specifically enable it.
If I budget enough to store 1TB of data for 1 month, then on the first day of the month I store 2TB of data - what should the behaviour be after 15 days?
Read/write access should be frozen, data should be saved for 1 month so you have time to react to warning emails. If you didn't upgrade in that time, it should be deleted.
Nuke the data. It’s gone forever if you didn’t back it up elsewhere. This should be a meaningful risk mitigation that I can employ to avoid having a catastrophic financial disaster.
This isn’t a limit I’m setting at some percentage above expected costs, it’s: “I don’t want to take out a HELOC if something goes wrong”
Unfortunately, a lot of people keep their backups in the same cloud account as their primary data. Thinking that multiple copies and multiple availability zones are sufficient.
For these users, the article’s €54k bill would be replaced with their business data getting wiped out.
You know that's not how the cloud works. If you're build by the hour for compute and that compute is powering a server, the only way to stop that is by shutting off the compute, breaking the server.
I would love to have a “if the bill for this hobby project becomes a threat to my ability to pay my mortgage, nuke it.” If I cared about the data enough. I’d have backed it up.
I thought I had something wrong within my setup, I could never use Codex 5.3 while everyone else was praising it. It uses some weird terms and complex jargon and doesn't really make it clear what it was doing or planning to do unlike Opus which makes things clear, this allows me to give accurate feedback and change plans and make proper decision.
Not bad, but it sacrifices accuracy and there are risks of causing more hallucinations from having incomplete data or agent writing bad extraction logic. So the whole MCP assumes Claude is smart enough to write good extraction scripts AND formulate good search queries. I'm sure thing could expand in the future to something better, but information preservation is a real issue in my experience.
There are 4 models, all receiving the exact same prompts a few times a day, required to respond with a specific action.
In the first experiment I used gemini-3-pro-preview, it spent ~$18 on the same task where Opus 4.5 spent ~$4, GPT-5.1 spent ~$4.50, and Grok spent ~$7. Pro was burning through money so fast I switched to gemini-3-flash-preview, and it's still outspending every other model on identical prompts. The new experiment is showing the same pattern.
Most of the cost appears to be reasoning tokens.
The takeaway here is: Gemini spends significantly more on reasoning tokens to produce lower quality answers, while Opus thinks less and delivers better results. The per-token price being lower doesn't matter much when the model needs 4x the tokens to get there.
WordPress Foundation is paying for the servers, so I guess they have the right to choose who gets access or not. Using the resources as a single person or a small business is not the same as using them from a hosting company with millions of websites. Other hosting companies contribute to the foundation which keeps the service running. If WPEngine isn't contributing anything, it would be unfair for other contributors/sponsors. Especially that they are making a large amount of money from it.
As so often I think it would be beneficial for the conversation to provide some more context. Single user install generated load VS WP generated load on the infrastructure of WordPress.org
It's not at all a dick move to block IPs that essentially DDOS your free services.
Google, Amazon, you name it do this infinite times a day with crawlers.
If you build a business on taking resources from some public source, on a large scale, you could very well be out of a business at any time. This has been the case for a long, long time. And nobody seems to take issue with it.
I've got the API ready, which requires 2 main values:
- The original file ( which can be sent as a file, or hosted internally and requested using an ID )
- The data that needs to be printed into the audio file.
The API will return a watermarked version of the audio file that you can use later to extract the same data you sent before.
It's currently being tested on a production website, will wait for feedback, improve, and create an actual service out of the API.
It's 00:16, just about to go to bed, I ran `git push` and it's not working. Check Github, says it's down, I think it's only me, maybe I'm blocked, Github can't be down. Come here to check and it's down for everyone, such a relief.
reply