dfajgljsldkjag

Hate to agree, but with the underhanded tactics that Google is pulling with RCS, Meta is actually the sane and ethical company in the room.

> plus the fact that Google actually shoehorns RCS in countries where they think they can get away with it.

This is the real thing that nails down "RCS" as a totally google thing. Google will forcefully enable RCS for people on carriers that want nothing to do with it. And in that case Google controls the entire process every single step of the way.

> it also looks like this needs to be agent-initiated, ie. you can't add a "Text us" button that will take you to this experience.

This is correct as of when I researched RCS capabilities. Not sure if it's changed at all but it was a deal breaker for me.

Twilio, Bandwidth etc. are gatekept by Google and are basically just a middleman reselling Google's product.

Every single customer on the planet reachable via "RCS for Business" is through Google's servers and the Google Messages app.

> They may not be able to register with Google, but they can make an RCS app nonetheless.

They can make a "standards compliant" RCS which will be able to connect to literally zero carriers or servers on the planet.

In fact Google Messages "RCS" doesn't even implement the RCS standard. They use a proprietary protobuf api exclusive to Google. Google messages can't connect to any "RCS" servers except Google's.

> Twilio has RCS too

Twilio, Bandwidth etc. are gatekept by Google and are basically just a middleman reselling Google's product.

It's unfortunate that RCS is basically a de-facto Google walled garden in most countries. On android you can only use RCS in the Google Messages app and trying to build your own is explicitly blocked. [1] And of course the bonus that rooted devices are banned from RCS. All the propaganda Google spread to get Apple onboard (only with Google's blessing of course, Google could kick apple off if they wanted) was such a big win for them.

As a business I wanted RCS to be a simple upgrade to SMS, but instead they came up with this mess. Businesses using RCS for Business can send messages to anyone, but customers wanting to get in touch with your business can't. They can only reply to a message you sent first. And of course Google is the gatekeeper for anyone to be allowed to use it.

1. https://github.com/microg/GmsCore/issues/2994

I never really thought about cows having the ability to problem solve like that. It makes sense that she learned this just because she had a better setup than most cattle. It really makes you wonder what else they could do if they had more objects to interact with.

It looks like a simple way to schedule things without all the annoying ads you usually see. It is nice that you do not have to make an account just to set up a quick meeting.

The performance gain from using a single shared vector for the nodes is pretty crazy. It just goes to show how much allocation overhead can slow things down if you are not careful.

The example about the founder losing a hundred million dollars to taxes is painful to read. It makes sense why Cayman is becoming the standard now that big investors are accepting it. The old default of opening a Delaware C Corp seems like a huge trap for Latin American companies.

It is interesting to see how the Japanese market validates the idea of sharing information instead of hiding it. I get annoyed when companies hide their features behind a contact form because I just want to do my own due diligence. Giving people the information they need to research independently seems like a much better way to sell software.

I find it so annoying that they disable the NFC readers on their terminals just to force people to scan a QR code. It makes the whole checkout process feel completely outdated compared to every other store.

I think it is kind of scary that they bypassed the stalking protections just by changing the device ID constantly. It is really clever engineering to use random peoples phones as a data mule but it feels like a security hole.

It is annoying that the library does not just give you the query duration by default. I like that this fix does not require messing with the library internals but it is still a bit of work to set up just to get basic logs.

It is such a relief to see someone say you do not have to delete your accounts to be healthy. I find myself opening these apps purely out of habit when I am bored or anxious. The idea of pausing to ask yourself why you are logging in is something I definitely need to try.

I was under the impression that once you have a vulnerability with code execution, writing the actual payload to exploit it is the easy part. With tools like pentools and etc is fairly straightforward.

The interesting part is still finding new potential RCE vulnerabilities, and generally if you can demonstrate the vulnerability even without demonstrating an E2E pwn red teams and white hats will still get credit.

Interesting they are releasing a tiny (30B) variant, unlike the 4.5-air distill which was 106B parameters. It must be competing with gpt mini and nano models, which personally I have found to be pretty weak. But this could be perfect for local LLM use cases.

In my ime small tier models are good for simple tasks like translation and trivia answering, but are useless for anything more complex. 70B class and above is where models really start to shine.

This is a really cool example of transferring domain knowledge from one field to another. It is impressive that the same math for lasers can help you fold a beetle from one sheet of paper. I need to find a way to pivot my own side projects like that.

It is really good that they are taking steps to remove this stuff. You can usually tell right away when something was not written by a human.

Agents can complete an impressive amount of tasks with just this, but they quickly hit a bottleneck in loading context. A major reason for the success of agentic coding tools such as Claude and Cursor is how they push context of the problem and codebase into the agent proactively, rather than have the agent waste time and tokens figuring out how to list the directory etc.

Companies doing SEO for lead generation have been ruining search results for years now. I've searched for basic car questions so many times and gotten low effort articles from some random dealership in Kentucky or Ohio [1]. No, Toyota of Louisville, I am not a lead and I don't want your opinion any more than I want the opinion of any of the other 9000 local car dealerships in the country. And this pattern applies to home maintenance[2], legal questions, etc. Screw those guys.

We don't click because those guys have made it so there's nothing worth clicking on.

1. https://www.google.com/search?q=how+often+to+do+oil+change

(this particular search has a few high quality results ranked at the top, but it illustrates what random dealerships from who knows where are doing to ruin the results.)

2. https://www.google.com/search?q=how+to+wash+upstairs+windows

The author is right that usage is dropping, but that really has no bearing on whether or not it is open source. Technologies get replaced all the time regardless of the license they use.

I like the idea that we moved from cathedrals to megachurches because it explains why everything feels so corporate now. It is easy to forget that the messy bazaar is still underneath all the shiny tools we use.

It is annoying how often our offline metrics look perfect while the actual A/B test shows zero lift. The article explains that gap really well by framing recommendations as an interventional problem rather than just an observational one. I guess we really need to start looking at counterfactual evaluation if we want our offline tests to actually mean something.

This is basically just a massive database of litter that documents what people threw in the canal over the centuries. It is interesting to see the materials change as you scroll from the older dates up to the present day.

I had assumed the delay was technical but it turns out it was mostly about finding a business model that worked for everyone. It is good they finally settled on a shared infrastructure approach so they do not have to crowd the tunnels with extra equipment.

The article mentions 7-OH also known as feel free, which shockingly hasn't been banned and is sold without checks at many stores. There are quite a few Youtube videos talking about addiction to it and it sounds awful.

https://www.youtube.com/watch?v=TLObpcBR2yw

The guardrails clearly failed here because the model was trying to be helpful instead of safe. We know that these systems hallucinate facts but regular users have no idea. This is a huge liability issue that needs to be fixed immediately.

It is pretty disappointing to see legitimate engineers getting mixed up in these crypto schemes just because the money is there. The article makes a good point that buying the coin does literally nothing for the software itself, so it is basically just gambling. I guess this is just the newest way to grift off the hype around AI development.

The comparison to artisanal furniture really lands because we do spend way too much time on the basics. It would be nice to just assemble the pieces we need and move on to the interesting work.