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

The drawbridges come up: the dream of a interconnected context ecosystem is over

dbreunig.com

Post reply on HN
Print view View original

61–70 of 85 posts

Re: The drawbridges come up: the dream of a interconnected context ecosystem is over

#61

Post by TeMPOraL »

_heimdall wrote: post #29
seydor wrote: post #10

I m optimistic, because LLMs can understand plain language. MCP won't last as the article correctly states, but you will always be able to say to your AI to open your email and search whatever. And companies cannot block you from doing that as long as it is your own PC / Phone. If we do allow companies to block AI agents from accessing our own computers and data, then the users are to blame for falling again into ano…

MCPs are, in part, a response to the difficulties LLM companies had when trying out LLMs interact online by visually navigation the screen. They need APIs for it to be efficient. For whatever reason they didn't choose to use accessibility tooling to automate agents, and we haven't written REST APIs for 20+ years - they're left hoping a newly designed protocol will fix it.

> For whatever reason they didn't choose to use accessibility tooling to automate agents

That surprises me too. It's arguably the only way forward that has a chance of surviving for more than a moment, because accessibility actually has a strong cultural and (occasionally) legal backing, so companies can't easily close that off.

Re: The drawbridges come up: the dream of a interconnected context ecosystem is over

#62

Post by ImPostingOnHN »

MichaelZuo wrote: post #51
__MatrixMan__ wrote: post #50

Earlier quoted context omitted.

Are you proposing that interoperability is not an overall net positive? If it's getting a bad rap right now it's just because it's not always simultaneously a competitive advantage. But that line of thinking is a race to the bottom. I mean, why not just kill your competitors? Then your product, however bad, would be the only one. Clearly a net negative, but a competitive advantage. What has changed is that we've rece…

Privacy, reputation risk, etc., seem like huge disadvantages… so it’s not clear at all if it’s a net positive overall. Someone has to actually do that analysis in the first place. It doesn’t just automatically become true.

What are the privacy and reputation risks for me as a user, if I'm able to access my data via API?

If you're referring to 'net positive [for facebook]' rather than [for users] or [for society], then the point is conceded that facebook can make more profits abusing their users versus being more considerate of them.

Re: The drawbridges come up: the dream of a interconnected context ecosystem is over

#63

Post by _heimdall »

TeMPOraL wrote: post #61
_heimdall wrote: post #29

Earlier quoted context omitted.

MCPs are, in part, a response to the difficulties LLM companies had when trying out LLMs interact online by visually navigation the screen. They need APIs for it to be efficient. For whatever reason they didn't choose to use accessibility tooling to automate agents, and we haven't written REST APIs for 20+ years - they're left hoping a newly designed protocol will fix it.

> For whatever reason they didn't choose to use accessibility tooling to automate agents That surprises me too. It's arguably the only way forward that has a chance of surviving for more than a moment, because accessibility actually has a strong cultural and (occasionally) legal backing, so companies can't easily close that off.

I was genuinely (maybe naively) impressed when google pushed for https everywhere. Maybe there were nefarious reason behind it that I missed, but it did a lot of good for the average web user.

LLM companies could easily have made a similar impact by leaning on accessibility tooling. Pushing companies to better support ARIA standards online would have made a huge impact for the better.

Heck, throw a little of that LLM money towards browser vendors to even better support ARIA - personally I'd love to see a proper API for directly accessing the accessibility tree.

Re: The drawbridges come up: the dream of a interconnected context ecosystem is over

#64

Post by TeMPOraL »

_heimdall wrote: post #63
TeMPOraL wrote: post #61

Earlier quoted context omitted.

> For whatever reason they didn't choose to use accessibility tooling to automate agents That surprises me too. It's arguably the only way forward that has a chance of surviving for more than a moment, because accessibility actually has a strong cultural and (occasionally) legal backing, so companies can't easily close that off.

I was genuinely (maybe naively) impressed when google pushed for https everywhere. Maybe there were nefarious reason behind it that I missed, but it did a lot of good for the average web user. LLM companies could easily have made a similar impact by leaning on accessibility tooling. Pushing companies to better support ARIA standards online would have made a huge impact for the better. Heck, throw a little of that LLM…

> LLM companies could easily have made a similar impact by leaning on accessibility tooling. Pushing companies to better support ARIA standards online would have made a huge impact for the better.

If anything, they'd have the reverse impact, unfortunately. The thing is, the companies whose sites/apps/resources would be accessed by LLMs don't want this. That's the entire point of the article we're discussing.

All I'm saying is, accessibility is the only interoperability wedge they can't just close off, without a huge community backlash and in some cases because of compliance reasons.

Re: The drawbridges come up: the dream of a interconnected context ecosystem is over

#65

Post by MichaelZuo »

ImPostingOnHN wrote: post #62
MichaelZuo wrote: post #51

Earlier quoted context omitted.

Privacy, reputation risk, etc., seem like huge disadvantages… so it’s not clear at all if it’s a net positive overall. Someone has to actually do that analysis in the first place. It doesn’t just automatically become true.

What are the privacy and reputation risks for me as a user, if I'm able to access my data via API? If you're referring to 'net positive [for facebook]' rather than [for users] or [for society], then the point is conceded that facebook can make more profits abusing their users versus being more considerate of them.

How is this relevant? The user isn’t the one making the decision to implement it or not.

Re: The drawbridges come up: the dream of a interconnected context ecosystem is over

#66

Post by __MatrixMan__ »

MichaelZuo wrote: post #65
ImPostingOnHN wrote: post #62

Earlier quoted context omitted.

What are the privacy and reputation risks for me as a user, if I'm able to access my data via API? If you're referring to 'net positive [for facebook]' rather than [for users] or [for society], then the point is conceded that facebook can make more profits abusing their users versus being more considerate of them.

How is this relevant? The user isn’t the one making the decision to implement it or not.

In general there are more users than implementers, so their experience is more relevant to whether it's an overall net positive.

Re: The drawbridges come up: the dream of a interconnected context ecosystem is over

#67

Post by armchairhacker »

pixl97 wrote: post #56
armchairhacker wrote: post #49

Today, an external camera can record your computer screen and audio, AI can extract the data and metadata, and a 2D contraption can physically move your mouse to interact. In the future, these will probably become more effective and cheaper (eventually the AI becoming possible to run locally, though even today it’s possible with a good GPU on simple UIs). Lots of other comments argue for regulation mandating open API…

The biggest problem here is preventing said scraping from shutting down the sites with cost. Over the years most of the problems I had with sites getting overloaded were from valid 'scrapers' like Google. Quite often providers were adding more memory/cpu just to ensure Google could index the site. While hosting costs are cheaper than ever, being on the internet can still get very expensive very fast.

In theory it could be solved by websites charging a very small fee (maybe crypto) for incoming requests, to pay operating costs. The fees from human browsing (even excessive like 1 site / second) would be negligible. APIs that use scraping would forward the fee to their users. Training or search (index) data would cost a lot to generate, but probably still insignificant compared to training the ML model or operating the search.

It already costs a small amount of electricity for clients to send requests, so maybe paying for the server to handle them wouldn’t be a big difference, but I don’t know much about networking.

Although in practice, similar things have been tried and those haven’t worked out, e.g. Coil. It would require adoption by most browsers or ISPs, otherwise participating sites would be mostly ignored (most clients don’t pay the fee so they don’t get access, they don’t cares because it’s probably a small site and/or there are alternatives).

Re: The drawbridges come up: the dream of a interconnected context ecosystem is over

#68

Post by ImPostingOnHN »

MichaelZuo wrote: post #65
ImPostingOnHN wrote: post #62

Earlier quoted context omitted.

What are the privacy and reputation risks for me as a user, if I'm able to access my data via API? If you're referring to 'net positive [for facebook]' rather than [for users] or [for society], then the point is conceded that facebook can make more profits abusing their users versus being more considerate of them.

How is this relevant? The user isn’t the one making the decision to implement it or not.

> How is this relevant?

How is it relevant to what? We were discussing "net positives" as in net positives for society, but it sounds like you're referring to net positives for Facebook alone, at the expense of all others.

> The user isn’t the one making the decision to implement it or not

This is true. Here is another difference between them: Facebook isn't the one being harmed by cutting off approved API access to 3rd parties.

It seems like we both agree that Facebook and users are different groups, in different circumstances, with different things to gain and lose from different decisions made here.

With that shared foundation of understanding: Why should we care about what Facebook wants, more than we care what users want? Why should we care about Facebook wanting to stifle startups and other competitors, more than we care what users want? Why should we care about Facebook's profit margin, more than we care what users want?

Of course Facebook is free to do what is good for them and bad for users, and they indeed chose (and continue to choose) to do so. Here we see the predictable result: Users criticizing them for it.

We are all smart and can comprehend Facebook's purely-economic "fuck the users, we wanna make money" decision criteria just fine, but we don't have to respect it.

Re: The drawbridges come up: the dream of a interconnected context ecosystem is over

#69

Post by MichaelZuo »

__MatrixMan__ wrote: post #66
MichaelZuo wrote: post #65

Earlier quoted context omitted.

How is this relevant? The user isn’t the one making the decision to implement it or not.

In general there are more users than implementers, so their experience is more relevant to whether it's an overall net positive.

More relevant than the people who make the decisions… according to who? Is there some arbiter that can overrule the decision makers?

If so… then we should discuss those instead.

Re: The drawbridges come up: the dream of a interconnected context ecosystem is over

#70

Post by skybrian »

cryptonector wrote: post #58
skybrian wrote: post #46

Earlier quoted context omitted.

Yeah, I don’t think it’s the wrong decision. Maybe I should have called it a design tradeoff. Edit: editing posts is nice to have.

> Yeah, I don’t think it’s the wrong decision. Maybe I should have called it a design tradeoff. How would you design Bluesky to prevent analog holes?

There's no way to do that, but it doesn't mean that people give up on privacy.

We could contrast Bluesky with the Fediverse: a maze of independent websites with uneven distribution and no systematic search or archiving. So, if you don't have much of a following and you delete a post, it's possible that nobody saved it. Some people prefer it that way.

Post reply on HN