> Is it private?

> 100% local. Nothing leaves your machine. Ever.

Except for the part where it gets added to the context window and sent to anthropic's servers?

This is a bit strange to point out explicitly, since that ship sailed anyways.

Boastful lies like this are a telltale sign of vibe-coded projects. Approximately, an AI is making word-association guesses from its context window, and arranging those guesses into grammatical forms that human RLHF reviewers find impactful. Frequently the lies are obvious if you have a mental model of the project, which the AI doesn't have.

There's also no reason to learn to read and write! First graders could just point their phone at some text and have it read to them, or dictate to their phone to achieve the reverse. Why learn to swim, walk, run? Machines can do that for you too!

For now there's plenty of people who are significantly more capable than AI models. Someone who fully outscources to machines will never join that club.

You have to evaluate students on their own skills before you continue their education, because at some point AI models won't be able to help them. Anyone can use some LLM to pass the first few months of undergraduate engineering disciplines, but if you got through that and haven't learned a thing, you're completely fucked. Worse, you won't even notice the point at which AI starts to fail until you get your test results.

Once the above is not true anymore, education is pointless anyways. However for now AI can at best replace the worst performers and only in some areas.

>You have to evaluate students on their own skills before you continue their education, because at some point AI models won't be able to help them.

If at some point AI models won't be able to help them, then give them assignments that reach the point where AI alone isn't enough, so they'll only be able to solve them if they learn whatever is necessary. This is what's meant by "making assignments harder". Students who learn to solve harder problems with AI will be more competitive in the workforce than students who only learn to solve easier problems by themselves. Because AI already allows people to solve harder problems than they could unassisted, but it's a skill that needs to be learned.

As an example, with AI, it'd be a reasonable assignment to ask students to write a working C compiler from scratch. Without AI that'd be completely beyond the reach of the vast majority of students.

That's great for autodidacts, but most students will be stumped by a complicated problem if you don't slowly walk them up an incline first.

Also what do you think is an appropriate assignment for first graders where "AI is not enough"? Are we supposed to give them problems meant for engineering majors?

The things you are saying at best apply to a few select areas of education and you are hyperfocusing on them. What you are neglecting is that a lot of education focuses on teaching tool use: reading and writing is a tool, CAD software is a tool, AI is a tool, even language is a tool. For many people the best way to learn to use tools is being taught by another human being. That human being has to evaluate their progress somehow. If a first grader uses their phone to have text read to them, this tells me very little, except maybe that they can at least understand spoken language to a degree.

Using LLMs effectively, especially without essentially becoming the LLMs meat-puppet, requires a set of skills many 10th graders still struggle with. Skills like putting what you mean into words, extracting meaning from text, and thinking critically about the information you are fed.

Finally there's the matter of philosophy, ethics, and politics, which also happen to be on the curriculum in some places. Are you going to let a LLM argue for you? If you have never learned to evaluate your own beliefs and turn them into something coherent that you can communicate to others, and instead let the LLM argue on your behalf, then congratulations: you have just un-personed yourself because you refused to let others help you become an actual individual in society. You're a sack of meat hooked up to a machine. ... It's probably obvious I feel strongly about this in particular.

At the end of the day, we can at least agree that people should learn to read and write? For now?

Not really the same thing. They're much larger already than most stores you'd see in urban Japan.

Think more in terms of small convenience stores ("Spätis" with daily necessities) everywhere. Typical distance to a store is maybe 500-1000m in Germany. In dense areas of Japanese cities it's closer to one store every 100m-200m.

So in Germany it'd be a 10 minute walk, while in Japan most of your "walk" would be getting downstairs.

The flipside of that is that selection is going to be limited compared to what you'd find in Germany.

I see. What you describe does seem to match what I experienced in NYC, Portugal, and Spain? Small supermarkets everywhere with a bit of a random selection of items

For example the great pyramids were already a popular "tourist" destination hundreds of years before christ.

https://en.wikipedia.org/wiki/Great_Pyramid_of_Giza#Historio...

Not a big deal then either. Just don't do it daily.

Now you can build a failed startup with the help of AI!

Opus 4.7 seems to reach ChatGPT levels of verbosity in code and loves to overcomplicate the most simple things.

This is something it spit out just now (trimmed a 9 line comment though):

        let keepSize = 0;
        let overBudget = false;

        await this.items.orderBy('[priority+dateUpdated+size]')
            .reverse()
            .eachPrimaryKey((primaryKey, cursor) => {
                if (overBudget) {
                    evictKeys.push(primaryKey as string);
                    return;
                }

                const key = cursor.key as [number, number, number];
                const itemSize = key[2];
                const contribution = itemSize > 0 ? itemSize : 0;

                if (keepSize + contribution > maxSize) {
                    overBudget = true;
                    evictKeys.push(primaryKey as string);
                    return;
                }

                keepSize += contribution;
            });

Come on now... what? For a start that entire thing with its boolean flag, two branches, and two early returns could be replaced with:

        let totalSize = 0;

        await this.items.orderBy('[priority+dateUpdated+size]')
            .reverse()
            .eachPrimaryKey((primaryKey, cursor) => {
                const key = cursor.key as [number, number, number];
                const itemSize = key[2];
                const contribution = itemSize > 0 ? itemSize : 0;

                totalSize += contribution;

                if (totalSize > maxSize) {
                    evictKeys.push(primaryKey as string);
                }
            });

I'm back to 4.6 for now. Seems to require a lot less manual cleanup.

I haven't had that problem in a while. Did browser vendors already do something about it?

On qutebrowser I type 2H instead of H, and it doesn't go to the most recent history item at all. Mostly though, no, spammy websites still do this, and browser's haven't fixed it.

The only viable way to even get most of them is to shut down internet access entirely. It's not a realistic solution, unlike blocking a few well known IP ranges belonging to a large corp like Cloudflare.

And even if you managed to get them all beforehand, some VPN providers will adapt and keep some servers in reserve, putting them online just as you managed to block the previous ones. Getting around internet censorship is a large chunk of their business, and some are really good at it.

I can provision a new VPS in about 5s of active work. I'd probably fully automate spinning up new servers and failing over because automatically detecting which got blocked is trivial. Bonus points if you use providers that let you attach multiple IPs to each VPS for cheap. Use some censorship resistant decentralized protocols to provide the next couple IPs to your client software and you're good.

And then they still need to monitor hundreds of VPN providers for whether they have new IPs, which is not neccssarily as easy as just grabbing a list of them. Once they have some, they then need to forward them to the ISPs and ask for them to be blocked. Their process is significantly less friendly to automation.

No country ever won this fight short of total shutdown/disconnects.

> No country ever won this fight short of total shutdown/disconnects.

Some countries also throttle pretty effectively. So you can connect but if you're trying to do more than read Hacker News it's not very usable.

Chances are the e-mails they've been sending so far went unread/got moved to spam by a lot of users and Gmail took that as a signal.

I send nothing but password-reset mails and never had an issue getting anything delivered, even though people constantly whine that delivering e-mail yourself has gotten so hard nowadays.

Just got a clean IP and don't send crap.