I think one time I asked opus about copyfail when it just came out and it did treat me like some sort of criminal, but are there really people that run into this on a regular basis other than cybersecurity experts (which cannot be a big enough group to generate all of this criticism)?

I have been trying to work on debugging tools using Opus 4.8. As it turns out working with low-level techniques that inspects and alters behavior of other processes is bordering close enough to cybersecurity research that it often hits the guardrails.

I asked it to clip a video onto segments based on location of spoken words (using separately done audio transcription) and it flagged that as cyber.

I do not agree that the concept of writing can be considered a specific tool at all. Maybe a pencil, but writing opposed to remembering everything is a more substantial level of a paradigm shift than AI use is.

This is like a smoker that lives to 100 saying that he had no increased risk of developing lung cancer because he didn’t at 100.

It's more like a hypothetical world where there were millions of smokers and none of them ever developed lung cancer

Some people made grandiose claims about its capabilities and I wanted to experience it myself.

OK, but for almost 24h straight? That seems a little obsessive, and not in the good way.

Getting excited about the announcement of new capabilities is very normal.

People used to wait in line all night to buy an iPhone. This isn’t that different.

I’ve also been trying to use it a lot due to all of the hype, but when I compared it side-by-side on a specific problem against Opus, I think that the solution Opus came to was cleaner and more accurate, although also more verbose.

Small sample size, but if Mythos/Fable was that much better, I feel like it should’ve given me an obviously better answer than Opus.

Considering that this is a brand new release of a frontier model that Anthropic is hyping hard, I'm not sure that the conclusion to draw from their repeated attempts to use it is that it's impressive... Anthropic is promising that it's impressive and we're all trying to test it out.

I, for one, have tried using it several times today and the guardrails kept switching the model back to Opus, so I have no clue if it's impressive or not.

It isn't reasonable to infer that OP was claiming to have universally been unimpressed about every facet of Fable, and now some unrelated impressiveness is the evidence of their false claims.

I guess the issue is that when someone can't use a product immediately, they have an urge to abandon it altogether, not learn how to use it.

It depends highly on the application. If the application domain is inherently complex and or used in business contexts, then they will have to learn how to use it regardless. Intuitiveness only works for somewhat cookie-cutter applications. Consider Excel: Excel is not intuitive to people who have not used excel. We can make it easier to use, but regardless the user will have to learn the fundamentals of a spreadsheet (and even how the data is stored in memory!) in order to successfully use excel. The reason I say users even have to understand how data is stored in memory is because of types. Dates are not strings, for example.

The personalized algorithm is not the root issue though. The root issue is that social media sites live by increasing engagement of its viewers. Because of this, they all get away from the original stated purpose of bringing people together, and go all in on maximizing engagement by increasingly shady ways. Of course, the personalized algorithm is a huge one, but there are also things like "Show HN" controlling what is on the front page, selectively taking down flagged material. Remember, HN has advertisements as well, and will regularly post job ads for positions in startups. They know that outright going in the direction of 'personalized algorithm' would alienate their viewerbase, so they avoid it, but still do all of the other practices that social media sites do.

I think it can be argued that, because human interests are so different and fractal, there's only so much damage that can be done with non-personalized algorithm. Ads for a certain demographic will just alienate the other demographic. But with personalized algorithm, the poison can be customized for each person and be more fatal.

We can imagine 100 different websites each with different non-personalized algorithms - another algorithm decides which one is best for you and directs you to that site. That's effectively a personalized algorithm with extra steps.

That's where we are. HN is one of those sites.

Yeah but then each of those websites would only have 1 audience, which would be economically ruinous. Even HN has >1M daily view which is quite a lot of eyes to cater.

I guess there can be a sweet spot in a certain amount of MAU. And this is why I prefer medium-sized subreddit.

Yes, just because it is uncomfortable for them to realize, doesn't make it not true. Words have meanings.

Do you have any actual examples of documents where it did not work?

My examples are all internal/confidential, but if someone from the project wants an example I could probably do some search/replace redaction. It would be a lot of work though because there's photographs and such too, and indexes, and tables, and documents inserted by reference, cross references, conditional fields, bibliography fields, formula fields, etc etc.

Why do you say it doesn't work? It seems to work fine for the feature parity they claim.

For “pixel-faithful” documents it doesn’t seem to be replicating documents all too well based on the many other testimonies in these comments

>doesn’t seem to be replicating documents all too well based on the many other testimonies in these comments

Ironic, that you are agreeing with a post saying they put in little effort for the implementation when you have put in absolutely no effort in saying that it doesn't produce pixel-faithful documents, such as producing a single concrete example.

I agree. To each their own, but the UI updating automatically doesn't really add much value to me. I would prefer that the view I am seeing is a snapshot in time of what the ground truth server was, not some mixed state that forces me to consider the possibility that seeing my request go through on the screen doesn't actually mean it went through and has been sent to the server.

The goal is - and I think they have achieved it - is that you don't have to think about it. They handle sync, and they do it reliably.

Then they have solved one of the fundamental hard things in computer science.