shostack

I am dreading what this future looks like as there is now an obfuscated middle layer between me and search results. At least when I search Google I can use my finely honed senses around what is SEO garbage vs a reputable site. And I can infer a lot from the site itself in terms of if they are biased in how they are pushing something.

But with models calling for results and not seeing what Google or others are serving back, and then having the AI give you a summary of reviews or rankings for shopping for example, it is really scary how much more opaque that will get.

More realistically, I expect the AAA titles to move to streaming only cloud computing services, which is what they wanted to do anyway.

I'm hodl with my 2018 Intel mbp for a truly local model capable machine.

I keep seeing that subreddit mentioned in a context that appears trying to push those against this to violent solutions. Almost as if it is an attempt at driving some sort of stochastic action.

How did you set up the auth? Is this a local Android app or a web service connected to the Web API of Messages?

I am desperate for something like this on Android and cannot for the life of me fathom why the messages app does not have a simple keyword-based filter.

I am going to attempt to wire something up with Tasker because it is getting too bad to cope with.

Everyone focuses on Amazon here, but there is an entire industry of e-commerce tools and products and platforms and services designed around optimizing e-commerce sites for sales and other uses.

Agents invalidate a lot of that and so it is an existential threat to many of them. I expect the near future will result in a lot of different results tailored to agents that may actually result in higher priced products for agents browsing versus humans.

Vikunja is a decent project management layer for this I've found.

I'm not paying to read beyond the click bait but if this article but if this is about using hashed identifiers, this is like.... Something that has been used by every major and minor ads platform and advertiser for years. It is nothing new.

What matters is the signals that are joined against the hashes and if what privacy controls they have in place for preventing reidenitfication.

But for those unaware, hashed identifiers are why every publisher has moved to 1p data and logins. Cookies are dead and unreliable so 1p identifiers are the only real way for an advertiser to target a given user with any reliability.

If anyone has access to the full post and wants to share what is novel about this situation, please do. From the bit I could read I would be more surprised if they were not doing this as it is a critical component of every major publisher platform's audience and measurement stack.

Their margin is Bezos' opportunity. I don't know how he could have been more direct.

You seem surprised that a nation who invaded Ukraine unprovoked and has regularly gone out of their way to target civilians and has been found guilty of violating the Geneva Convention would carry out exactly the kind of hybrid warfare they have a reputation for. Curious.

Also a squarely middle class income in a vhcol area like the Bay Area. But a super luxurious lifestyle.

I have been wanting a Bluetooth ring myself so I can just tap my finger and have it trigger my airpods which would initiate live voice or dictation on my phone or computer.

I have not considered input for editing functions but that makes a lot of sense. Maybe a swipe gesture or something with a little capacitive pad on the ring.

YouTube transcripts are increasingly if not completely blocked now. It used to be great to grab content from long tops. I did not have time to watch but wanted to learn about. Now. The only way to do it is to manually copy and paste the transcript information or use computer use. But I cannot easily do it on my VPS it seems unless I'm missing a trick.

I see everybody freaking out about unfettered payment access to one's bank accounts and I keep wondering why people don't jump to the obvious solution of simply only giving it a single purpose or limited fund credit card number from privacy.com or something.

Am I missing something with the concern about ability to constrain the blast radius?

Counterpoint: I have had my own Muse like harness with Hermes and Matrix for over a year and all I want to do with it is seamlessly long-in and purchase stuff without worrying about secure credential injection or payment confirmation.

Credentials and payments for agents and local proxy browsing is a non-trivial problem.

And I promise you my Hermes agent is not sitting here trying to get me to buy things.

Don't forget my favorite:

"Project" for remote connections are local to the device they are created on. So if you create a project on your laptop for a remote server with a session in it, and want to access that same connection on your phone with the same project folder structure and session grouping, you need to manually recreate the project on that new device and add the session.

And you cannot mix local and remote sessions in a single project. So if you need to do anything locally for a remote project, get in the habit of naming projects things like "Project Foo [Local]" and "Project Foo [Dev]".

It's funny that you say you want a more user-friendly lower maintenance version of something than Muse but then say something better than openclaw or Hermes.

What exactly is high maintenance about Hermes? I agree the setup could be a little more user-friendly, but I have it running on my VPS with Matrix and the actual maintenance has been very minimal.

Spending a lot of time on harness engineering to try to compensate for deficiencies with Deepseek v4 flash because I'm too cheap to use better models on the other hand...

They also pair that with increased prices unless you use the app. It is particularly insidious.

I am curious what your take is on skill packages like Obra/Superpowers and grillme and the like. On one hand I find them extremely useful for asking me questions. I wouldn't have thought of myself or that I don't feel I would have necessarily gotten in a structured a format just by having the AI interview me. On the other hand, I am not sure how to tell when they have become obsolete because I'm not sure where to even start with generating evals for such a thing.

Last I checked they still offer no training ZDR US based hosting. It is one of my three pinned providers for deepseek v4 flash along with Parasail and Deepinfra.

How timely. This is a great write-up on what I meant: https://news.ycombinator.com/item?id=49850305

These are effectively the next OS. I wonder if Microsoft realizes that and if so what their strategic play is.

TY, appreciate the thoughtful reply. Transparently, I have been so on the fence with moving more of my workflows and personal usage over from Hermes+VPS+ZDR model provider, or implementing the "Hermes shell over Codex Subscription" pattern because I worry about:

1. Legal loopholes given OpenAI's advertising aspirations and model training needs

2. Data retention and rising threats of fascism that historically have not served the persecuted very well when fascist regimes get access to said data

3. Risk from centralized collection of that data with a company whose software I do not control in a world where enshitification and lock-in is the norm.

I really wish OpenAI did more to espouse exactly this: "When I say no training, I mean no training. No gimmicks around data vs derived data, synthetic data, preference data, etc." and ideally provide technical reasurrances that this is impossible (eg: certain technical ZDR approaches, etc.).

Do you happen to have a favorite reference to point me at that would document some of those official assurances to the nuanced detail we've discussed here?

Ted can you confirm your choice of words here to be precise for an audience who is familiar with the nuances, when you say "no training" or "training (opt out)" for personal... Is that inclusive of "sanitized" (or pseudonymized) data?

Your response to the original question is using generalized terminology when there is a very important distinction the OP made by the use of "sanitized."

People want to know to that extent derivatives of their data are being used. Synthetic data has been proven to be effective at generating training data and AI is very good at shuffling context such that you have something where you don't have to say it is "user data."

But there are many shades of gray there for people versed in how the sausage is made. I'm sure you'll appreciate then why your response leaves additional questions in light of that "sanitized" distinction.

So effectively, trying to undercut their competitor to reduce its wealth and power?

Where are all the articles stating how switching to a Flock competitor is just as bad and isn't making the issue go away?

Tech priests reciting blessings to the Omnisiah. Time to do the blessing of checking how much 5h usage I have left.

The negative comments, including the one I'll make here, are from fans who want the game to be more than it is and see the potential but continually get disappointed with its biggest flaw.

That flaw is that for all the features they bolt on for free, the actual depth never feels that much deeper. It feels like the shallow pond got wider.

The worlds and universe still feel repetitive and empty. There is very little emergent gameplay. Mechanics are not complex compared to simulator sandboxes like E:D or more in depth survival crafter games.

And so you end up with what often feels like a kids toy with the edges filed off. Its very colorful, and they keep giving us me shinier versions for free, but it still feels like playing with Duplo when I want it to be LEGO or K'Nex.

And I say the above with nothing but respect for Sean and the team and what they have accomplished. I will be a launch day LNF player as well.

Your comment is slightly reminiscent of the response to dropbox when it was first announced on HN.