vineyardmike

Ah yes, because I am also forced to buy the same amount of electricity and water from my regulated utility regardless of need.

I've literally never met anyone in real life who used a computer that didn't already have a browser running 24/7

This is why the subscriptions are important. When the usage is (vaguely) unmetered, the provider has an incentive to make usage cheap on marginal use.

It aligns the incentives for faster, cheaper, terse and more reliable models, because the model providers pay the wasted tokens and electricity costs.

That's so interesting because where I work, the push was to "add one more API" to existing services, turning them into near monoliths for the sake of deployment and access. Still a mess of util and helper functions recursively calling each other, but at least it's one binary in one container.

I don't own a car, I don't drive a car, and I will never live in a city that has been made for a car primarily. That is not a majority-opinion nor majority-behavior in the US.

Cars are obviously useful at moving things around places (eg. Taxis are great, Fedex would never make sense as a service with public transit), but shaping our civic society around them have been terrible for the humans within it.

AI is obviously useful, and has proven clear use-cases that generate value. I think it'll be terrible if we make civic society for it like we did the car.

You can't build a business on per-seat subscriptions when you advertise making workers obsolete. API pricing with sustainable margins are the only way forward if you genuinely think you're going to cause (or accelerate) reduction in clients' headcount.

Additionally, the value generated by the best models with high-thinking and lots of context window is way higher than the cheap and tiny models, so you need to provide a "gateway drug" that lets people experience the best you offer.

Using an LLM to handle a task for you seems a lot like letting a car move you. Cars will make you “fat and lazy” if you never move your body otherwise, but it’s fairly clear to see that this is avoidable.

The research seems to always get (intentionally?) misconstrued at headlines that LLM is “bad for you” as opposed to more mundanely stealing opportunities for exercise and practice of mental activities if you let it.

Re: product strategy

Honestly, it probably led to long conversations. The tokens/GPU time for one long conversation is more expensive than multiple short conversations. They’re trying to shore up their finances, and they’re moving away from the consumer market and towards enterprise, and students were probably a bad demographic to sell to.

Don't worry, at least it has a discord server linked.

This is basically why the big companies can sell subscriptions for cheaper than API costs. First priority can go to API users, lower priority subscription users get slotted in as space/SLO allows, and then sell the remaining idle GPU to batch users and spare training. Oh and geography shift as necessary for different nations working hours.

> The official White House Android app has a cookie/paywall bypass injector, tracks your GPS every 4.5 minutes (9.5m when in background), and loads JavaScript from some guy's GitHub Pages (“lonelycpp” is acct, loads iframe viewer page).

Doesn’t seem too crazy for a generic react native app but of course coming from the official US government, it’s pretty wide open to supply chain attacks. Oh and no one should be continually giving the government their location. Pretty crazy that the official government is injecting JavaScript into web views to override the cookie banners and consent forms - it is often part of providing legal consent to the website TOS. But legal consent is not their strong suit I guess.

I don’t buy it.

First, I think there is value in the “rent seeking” Amazon marketplace because how else would the models “go straight to the seller”, another centralized search engine? Why not just use the Amazon one then?

Second, one of LLMs big weaknesses is judgement on what to trust. I would not trust the judgment of an LLM to determine “the seller is legit”… unless we outsource trust verification to a third party marketplace (who will want a cut).

Finally, OpenAI has been aggressively pushing for this so they can take a cut of the transaction. So it’ll just be another middle man.

> If you (a manufacturer) apply, they want information regarding corporate location, jursidiction, and ownership. They want a bill of materials with country of origin and a justification for why any foreign-sourced components can't be domestic. They want information about who provides software and updates. And they want to hear your plan to increase US domestic manufacturing and progress toward that goal.

Wow NGL this sounds great if you ignore the reality that it'll be used as a partisan backdoor to enriching the administration.

When I was younger, adults used to answer the phone with "Hello, this is MyName, who am I speaking with?"

Pragmatically, even basic words from your voice can be used to estimate your age, gender, and geographic region (local accents).

But also read other comments, people are saying they answer their phone by stating their name, so plenty clearly use it as a greeting.

I clear my throat. Do it loud enough that the other end can tell someone is on the line, so they'll know to start the conversation. It isn't a rude "WHAT" type answer, but it doesn't directly acknowledge the caller, and is not inviting a conversation. Its a common enough act that it's not suspicious nor weird to hear during a conversation, and therefore its not off-putting if extended family or clients called from an unknown number.

It doesn't share my voice (for fingerprinting, demographic leak, etc, smh).

Also works as a bot filter - Humans tend to start with a "hello..?" because they're not sure anyone is there, while robots use the non-zero audio as a signal to start talking with full confidence.

You're missing the point. No one cares about the datasets you've created in a commercial context.

The effort being discussed is a volunteer effort among a community of tech enthusiasts, who are disproportionately privacy-oriented vs the average person. This will undoubtable skew towards middle-aged male audiences, and will be extra-selective against children. It's a best-effort collection, they're probably not turning anyone away, and it's only what they can get, they're (AFAIK) not paying anyone to collect underrepresented demographics.

Well, we'd stop having to spend so much taxes on redistributive efforts, again, like subsidized internet. It's up to voters and politicians to actually change the tax rate to save the money. It'd reduce government debt at least.

Electricity would be cheaper. Here in California, a significant amount of the (very high) electricity costs are used to maintain rural power lines. If rural people moved away, we'd be able to decommission them and no longer maintain the lines.

It wouldn't happen immediately, but as more people become urbanites, we'd be able to move gas subsidies and government road maintenance spending to the urban environment, where we'd spend on more drivers-per-mile roads, OR shift to public transit funding, or simply reduce that government spending.

Over time, we'd be less reliant on cars, which reduces everyones costs, but will mean we aren't so desperate to protect oil interests, so we'd be able to stop paying for wars in the middle-east. Honestly this alone has so many positive side-affects it'd be hard to actually enumerate.

> What if labor organizes around human work and consumers are willing to pay the premium?

We could start today, but sweat shops and factories dominate the items on our shelves.

But I’m sure people will draw the line at human made software…/s

Many of the aspects of life "outside the city" are subsidized by the city. It's affordable because of this, and the cities are extra unaffordable as a result.

There are many small towns who will never generate the tax revenue to cover their $50M highway off-ramp and associated infrastructure. The thread was about internet, which has also been subsidized. We subside oil so driving long distances is cheaper. We subsidize food production. Electricity and water distribution is subsidized by urban customers. Even health care is subsidized.

If rural people actually had to pay market-rate for these resources, it wouldn't be cheaper than the city.

In 2026 we need to update our mental model of Google. Google has been wildly successful at adding diversification. Around 40% of Google’s profit (depending on the quarter) comes from non-search income.

They build a wildly successful cloud platform, they’re expanding their subscription services, they’ve got enterprise offerings, etc

The trick is that Google accepted that none of their other business would likely have the margins and volume that search has, but they did it anyways.

Generally no. Big tech companies have gotten good at locking down devices to the boot loader. Some of the signing keys for certain OTA versions have leaked, but you can’t rely on that.

Some of the devices contain browsers, and people have set up hacky ways to turn them into thin clients through that, but it’s not particularly reliable IME.

I heard some Chinese brands which made similar hardware for Chinese consumers don’t lock their devices down, letting you flash an open install of Android on them, but I haven’t seen anyone try that IRL.

I have no firsthand knowledge, but I’d strongly bet that the home-assistant effort to donate training data is mostly get adult males, and nearly zero children.

> You could have said same about Transformers, Google released it, but didn't move forward, turns out it was a great idea

Google released transforms as research because they invented it while improving Google Translate. They had been running it for customers for years.

Beyond that, they had publicly-used transformer based LMs ("mums") integrated into search before GPT-3 (pre-chat mode) was even trained. They were shipping transformer models generating text for years before the ChatGPT moment. Literally available on the Google SERP page is probably the widest deployment technology can have today.

Transformers are also used widely in ASR technologies, like Google Assistant, which of course was available to hundreds of millions of users.

Finally, they had a private-to-employees experimental LLMs available, as well as various research initatives released (meena, LaMDA, PaLM, BERT, etc) and other experiments, they just didn't productize everything (but see earlier points). They even experimented with scaling (see "Chinchilla scaling laws").

Respectfully, I’ve never forgotten a call site, but also yes. In a hypothetical HelloWorld service, the HelloRequest and HelloResponse generally aren’t used anywhere except a rpc caller and rpc handler, so it’s not hard to “remember” and find the usage.

Some callers may not need to update right away, or don’t need the new feature at all, and breaking the existing callers compilation is bad. If your caller is a different team, for example, and their CICD breaks because you added a field, that’s bad. Each place it’s used, you should think about how it’ll be handled, BUT ALSO, your system explicitly should gracefully handle the case where it’s not uniformly present. It’s an explicit goal of protos to support the use case where heterogeneous schema versions are used over the wire.

If a bug is introduced because the caller and handler use different versions, the compiler wasn’t going to save you anyways. That bug would have shown up when you deploy or update the client and server anyways - unless you atomically update both at once. You generally cannot guarantee that a client won’t use an outdated version of the schema, and if things break because of that, you didn’t guard it correctly. That’s a business logic failure not a compilation failure.

> Minor thing that bothered me disproportionately: the constant syntax in the docs (x = 600) doesn't match what the parser actually accepts (x: 600).

You’re a better man than me. If the docs can’t even get the syntax right, that’s a hard no from me.

Also, fwiw, you’ve got a few points wrong about protos. Inspecting the binary data is hard, but the tag numbers are present. You need the schema, but at least you can identify each element.

Also, I disagree on the constructor front. Proto forces you to grapple with the reality that a field may be missing. In a production system, when adding a new field, there will be a point where that field isn’t present on only one side of the network call. The compiler isn’t saving you.

Fresh is more honest than better, and personally, I wouldn’t change it.

I assume this is less "if they're lying" and more "if you've independently collected this data". It doesn't require you to challenge the user-indicated age, it requires you to use your own signal before that of the OS.

As a silly example, tax software probably has your full birthday, including year, which is more precise. Many social networks collected this data, as did a lot of major tech companies that implemented parental controls already.

I don’t use ad block.

I find that when it messes with the layout or formatting of a website it’s really annoying, and I consider the volume and type of ads an important signal for a website’s trustworthiness.

Oh and plenty of devices don’t have easy access to ad block, like my work computer.

Not allowed on my work computer. Which I do use the internet on.

Also you can put ad block on Apple devices.

Fresh eyes are always better equipped to ask “Why” to poke at (no longer necessary) Cheshire Fences. Their ignorance (by virtue of being new) gives them cover to question things. This doesn’t necessarily mean inexperienced people, but that does give even more cover to ask the very necessary “why” and get answers, and excuses their lack of familiarity to intuit the situation.

Junior people also tend to have the humility to know they’re inexperienced and lacking insight, so they tend to be more willing to “follow” the senior employees rather than attempt to leave their mark as a leader. “Too many cooks in the kitchen” and all that jazz.

I’m not exactly fully bought into the idea (for many practicality reasons) but it seems easier to build many (and replace) ground stations than data centers.

Additionally, StarLink et al are now able to directly communicate with cell phones. It therefore should be possible to route entirely in space between “data center satellites” and communications satellites and communicate directly with an end user device, avoiding the entire terrestrial internet.