A major distinction here is that it is very cheap to host content on the internet and VERY EXPENSIVE to build things like a separate road network in the real world.
Who is actually hurt if I publish an llms.txt or MCP in addition to my existing content?
Lucky for you that if you remove the "ai" subdomain here then you get a traditional "human-first" website.
Really though, how is this all that different from making candidates type their resume into a form then filtering in their ATS? Seems like a nice ergonomic approach if they're actually set up to use MCPs in candidate sourcing (probably won't be the case for at least another year).
The size of the country has almost no bearing on the way we develop our towns and cities, subsidize car production, assume/require car ownership in public policy, etc.
It's also quite the stretch to claim the UK as the best country in the EU and even more to claim that it doesn't also have a car culture. And it's not just "America's worst state", the vast majority of US town and cities are car dependent.
Any time someone uses the excuse that “the size of the USA” prevents this or that, then ask why New Jersey can’t do this or that. Some states are the same size and density of the average European country.
NJ has an extensive train network. Its crap for the following reasons(only referring to main interstate rail, their city light rails are different):
1. The stock is divided into two sets: single decker trains that are falling apart but are still better than new stock because they were designed during an era where there was actual consideration given to comfort. They are old and run down but more comfy.
The new double decker trains are a design by committee nightmare: they have uncomfortable molded chairs made mostly out of plastic, super cold, noisier due to the terrible shifting whine of the electric motors, and now poorly maintained (god help you if you have to use their rest rooms).
2. They are constantly stopping due to priority given to amtrak and commercial rail. In the best case scenario the train takes ~45 mins to get from central NJ to NYC with stops. For about ~20 miles of track distance, that averages out to ~29-31mph. If Amtrack stops you that time is extended.
3. Very slow speeds due to old track technology.
4. Rising ticket costs for the same lousy service. (~$21 round trip from central NJ to NYC) I guess being able to have a digital ticket helps? (It probably helps them more than it helps you)
5. Pure depression because outside the main routes going into NYC and Hoboken, the dregs of society are the only ones riding these trains. This really grates on you if you take the train for years like I did.
6. Recently, they couldn't even provide working windows so people can see the 'beauty' of NJ. ( To be fair after all the TikTok meme videos called them out on this they finally got embarrassed enough to start replacing them: https://www.youtube.com/watch?v=j-a3r6i4hiQ )
It all priorities. People in the US just dont want mass rail over other priorities. The people who do are internet keyboard warriors.
Don't forget one of the most famous and visited destinations in the country is a walkable neighborhood served by great public transportation and uses a rat as a mascot.
You looked at that transportation recently? It is collapsing due to legacy, graft, and cost overruns. I don't presume you are European but I HATE when they use this system as an example or public transit that works in America. Its a dump. The worst trains in France and Germany run miles around it.
Your criticisms are absolutely right. I upvoted you, but just fyi I think your response may come off as too blunt/rude, and get “downvotes”. It doesn’t really matter here, but maybe helpful to know in general, if you didn’t already.
Fair enough! You're right it's a bit blunt. I didn't take very much time to edit it. Hopefully folks are able to look past any gruffness and find the substance in it.
> Can't we just have a set of best practices around API endpoints/functions? Like, imo we could just keep using Rest APIs and have a convention that an agent exposes endpoints like /capabilities, /task_status ...
To make this work at scale we all need to agree on the specific routes names, payloads, behaviors, etc. At that point we have defined a protocol (built on top of HTTP, itself a lower level protocol).
I don't sign a term sheet when I order at McDonalds but you can be damn sure they count how many big macs I order. Does that make them morally bankrupt? Or is it just a normal business operation that is actually totally reasonable?
This automatic sense of entitlement to surveil users is the absolute embodiment of the banality of evil.
It's 2025 - we want informed consent and voluntary participation with the default assumption that no, we do not want you watching over our shoulders, and no, you are not entitled to covertly harvest all the data you want and monetize that without notifying users or asking permissions. The whole ToS gotcha game is bullshit, and it's way past time for this behavior to stop.
Ignorance and inertia bolstering the status quo doesn't make it any less wrong to pile more bullshit like this onto the existing massive pile of bullshit we put up with. It's still bullshit.
You're making a huge jump from "gathering anonymous counters to understand how many people use the thing" to "harvest all the data you want and monetize it".
If they were tracking my identity across sites and actually selling it to the highest bidder that's one thing that we'll definitely agree on. This is so so far from that.
You're welcome to build and use your own MCP browser automation if you're so hostile to the developer that built something cool and free for you to use.
The supply chain vulnerability in any extension is obvious. The problems with telemetry - any at all - are wide ranging and it's crazy to me that people don't see this.
Any covert, involuntary, automatic surveillance of a person for any reason whatsoever should have a court order and legal authority behind it - it's gross and exposes the target to vulnerabilities they're not cognizant of.
For telemetry tracking user behavior to be useful at all, it's got to be associated with a user. The idea of telemetry anonymization is marketing speak for "we obfuscated it, we know deanonymization is trivial, but people are stupid, especially regulators."
Any anonymization done is sufficiently obfuscated such that corporate asses get covered in the case of any regulatory investigation. There's no legitimate, mathematically valid anonymization of user data that you could do without destroying the information that you're trying to get in the first place through these tools. This means that any aggregation of user data useful to a malicious actor will inevitably be compromised - the second Posthog or Amplitude become a desirable target, they'll get pwned and breached, and much handwringing will be done, and there will be no recourse or recompense for damages done.
The only strategy to prevent the dissemination of surveillance data is not to collect it in the first place. It should be illegal to collect the data without voluntary, user initiated participation, and any information collected should be ephemeral with regular inspection to ensure compliance. Any violation of user privacy should result in crippling fines, something like 5% of the value of the company per user per day of violation - if you can't responsibly manage the data, you shouldn't be collecting it.
This means all the automatic continuous development a/b testing intrusive corner cutting corporate bullshit would have to stop. Continually leaking surveillance data to malicious actors year over year with no repercussions has thoroughly demonstrated that people cannot be trusted with safekeeping data.
I will build and use my own automation if I need to, based on products that don't covertly, involuntarily, ignorantly surveil their users, without even being aware of potential for harm, and I'll continue to point it out when it shows up in random projects and products, because it's wrong and it should stop.
We should stop embracing the things that enshittify the world, and stop sacrificing things like "other people's privacy" for convenience or profit.
At a high level, the request format and endpoints. Instead of needing to write a bespoke connector for every type of context that matches their preferred API standards, I just tell my client that the server exists and the standard takes care of the rest.
Do you have similar doubts about something like gRPC?
> This looks like micro services crossed with AI.
Seems like a cynical take with no substance to me. What about a standard request protocol implies anything about separation of concerns, scaling, etc?
Because it's not a request format for LLMs, it's a request format for client software that is instrumenting LLMs. If you make a connector to say HomeAssistant to turn on/off your lights you're exposing a tool definition which is really just a JSON schema. The agent will present that tool to the LLM as one it's allowed to use, validate that the LLM matched your change_light_state tool schema and send off the appropriate API call to your server.
The spec is genuinely a hot fucking mess that looks like a hobby project by an overeager junior dev but conceptually it's just a set of JSON schemas to represent common LLM things (prompts, tools, files) and some verbs.
ChatGPT is available for free - you can find out for yourself!
It won't generate music for you (yet). Depending on your phrasing and how outlandish the request is, it might give you suggestions of actual bands to go check out or it may just describe the characteristics of what that would sound like.
When I asked "Led Zeppelin style music with djembe beats", it gave me a high level description of the guitar/drums/vocal sound + some Led Zeppelin songs that may adapt well to the fusion.
They experimented with music back in 2020 [0]. I suppose they didn't continue because the risk (work, training, possibly getting sued) is greater than the reward (splitting the market with Suno).
> ChatGPT is available for free - you can find out for yourself!
ChatGPT changes a lot. Somebody reading this comment thread in a year or five might well be interested in the answer but can't test it for themselves anymore.
What are you talking about? A "yaml formatted string" is just a string. And the Costco shareholder meeting notes/ballots that the GP posted is not YAML either, it's a pdf.
The Columbia logo on this craft is both. They're advertising the brand in a very cool and unique way AND they contributed significant heat shield technology to the craft itself.
Good point about the importance of checking your own page with your ad blockers and similar turned off.
It's something I imagine a lot of people would forget - I don't think the thought ever occurred to me for a personal site.
Part of having browsed the web with ad blockers turned on for so long is I tend to forget ads exist for the most part - I definitely forget how pervasive they are.
Who is actually hurt if I publish an llms.txt or MCP in addition to my existing content?