gbalduzzi

> - Most people are not in Signal

Whatsapp uses the Signal protocol for E2EE: https://signal.org/blog/whatsapp-complete/

So yeah, whatsapp messages are E2E encrypted and therefore can not be used as training data unless you upload them to the cloud manually, as this project does

Yes. The alternative is to have stalkers using hidden airtags

Just curious, what other annoyances do you have with V2?

The opentunnel feature is probably related to the "official native apps that are in the coming" mentioned in the article

It is pretty terrible performance wise for what it does.

The fact is people at Anthropic won't notice it because they are running it on M5 laptops

It is not something for me to worry about as a customer, but it is something for me to worry about as a member of the society

I'm thinking that you just need one website where you can associate a given fingerprint to a specific user and then tracking its activity around the web becomes easy.

Let's say I'm logged in a famous search engine, they associate my fingerprint X to my account A. If they then encounter fingerprint X anywhere else, there is a good chance it was me, A, doing it. If the next day I get fingerprint Y, it gets attached to A as well and the tracking information follows it.

Does this make sense?

I may be in the minority but I didn't get LLM vibes from this. Not enough to be bothered by it, at least

> My consumer line is measured in megabits per second and if I have to pay for my data usage the way it's paid for in data centers it would be far more expensive

Well because your provider assumes you are not using all your bandwidth constantly. Cloud bandwidth is only billed for you actually use

This I don't understand. I'm faster at moving the method then at prompting the LLM to do so

Exactly. I don't understand how so many developers seem to have a long tail of well written task specifications ready to submit to the LLM. Who produces them?

I keep seeing this but I'm not sure it is effective in the long run *if unsupervised*.

"Review quality and fix" doesn't mean a lot without context.

Does it mean to remove unused features and simplify the underlaying code? Does it mean changing the data structures to better support future development? Does it mean improving performance because of bottlenecks?

You are supposed to tell an LLM what your codebase needs, but if you just vibe code without knowing the code, "review quality and fix" will have unexpected results

Deals right now are intended to increase the trades between Canada and reducing those with the US, in little improvement steps.

If in 10 years the ratio moves from 70/6 to 50/26, the situation will be vastly different.

The options are not black and white

I think in many cases the authors value functionality much more then esthetics.

Which is a bummer because it is a complete different point of view from the more typical user experience

Because we are not there yet and you can't oversee something you don't know

I think it is perfectly understood. But we don't feel we are the ones that deteriorated the relationship, we are just reacting to it

It doesn't happen overnight, but the adversarial behaviour of the current administration and the tariffs really changed the perspective about America on many Europeans.

The discussion about building European alternatives had never been so mainstream. If and once they emerge, I think the shift will happen. But let's see

I started to consider it a show show way earlier than 18 months ago. Jira is so terrible to use that it is hard to phantom how they are able to be paid for their product

The point is that there aren't many complex tasks were fable delivers a significant value increase over cheaper models.

Single prompting a very complex tasks is rare even on frontier models, because it can be done successfully only for specific situations (e.g. you have a very strong verification step the model can iterate on).

Most of my everyday usage is for smaller takes, were you don't really get the benefit of the most expensive models, and my guess is that is the case for the most users

> do we think we're discovering that Chrome had more big holes than we thought or are we raising the security bar by fixing smaller holes?

To me the most probable explanation is that they automated a way to find (and fix) existing vulnerabilities in a way that was not possible before.

Some holes were probably very small, some were probably almost impossible to actually exploit, I don't doubt it. But still, I find it very hard to not consider this a strong security improvement overall (unless they made those numbers up)

Of course it is just tokens, but the result is the same.

If, in the amount of data they ingested, there was a clear pattern of responding in an hasty and carefree way to frenetic questions, LLMs will try more hasty and carefree solutions to a frenetic prompt.

You can decide whether you can say that they "feel" the urgency or not, but the outcome is very much the same

> That's not an advantage for the user at all, is it

Of course it is an advantage. I tend to access most of my apps and services from both the PC and the smartphone, having data shared across the two is a deal breaker for the very vast majority of them.

Meta has the option if I'm not mistaken, but my guess is that it is very, very rarely used.

An approach that works really well for me in opencode is to use DeepSeek flash for the code exploration sub agent, and a smarter model for the overall task.

Code exploration tends to burn a ton of tokens and do not require frontier knowledge. It then returns a compact overview of the findings to the better model, that can therefore perform its task with less tokens.

Anyone else has a similar approach?

It depends whether your goal is to hire for specific knowledge (hence specific questions) or for overall mindset and abilities (hence broader questions where you are able to extract the way a person thinks)

How much was valuable the output you got during the day?

Is it at least comparable to the $10k of cost?

You need agents.md and similar for indications about stuff that is not in the code itself. There are plenty of use cases for that, the alternative is have the LLM guess the most probable solution, which may be correct but may also be wrong.

Am I the only one not seeing a point in asking an LLM philosophical questions, especially about themselves?

Didn't Deepseek also publish a lot of research material on how they trained their model?

That aspect is probably more important for an open model then the source materials

> But why call them overpriced? Compared to what?

Compared to some available Chinese model I guess. For most common tasks the additional intelligence is marginal and the cost is around an order of magnitude higher.