2) Please don't call black overlay rectangles as "Redact" - it is maliciously misleading. I checked https://pdfcrowd.com/inspect-pdf/ and I see original parts that I covered with these rectangles (images are stored twice: as originals and as images with cut out regions).
Nightshade[1] 2.0? As if both tools were built by incompetent developer to distract attention from a real solution - publishing an llm-friendly version in an machine-friendly format (which is not really difficult and helps not only LLMs: e. g. cache, disable fancy complex syntax highlight, offload to github, provide clients and MCPs, optimize clients for common use cases). This example is simply a failure:
<a href="/bots" style="display: none;" aria-hidden="true" tabindex="1">
Amazing high quality data here!
</a>
Dumb curl-based LLM won't visit display:none links. Smarter browser-based navigators won't even render this link.
But the header is just "90% of Claude-linked output going to GitHub repos w <2 stars". No conclusion, just some random fact.
The problem is that this title is editorialized, and the fact is cherry-picked. Why not =0? Why not >1000? This is just a dashboard, it highlights "Interesting Observations", but stars statistics is not there.
Good ol Putin's strategy, even visually similar to Natalia Poklonskaya, used to push political agendas in wartime, used to divert attention from actual problems and fill headlines, while other topics become inaccessible due to war censorship.
I wouldn't be surprised if it was proposed by the same political strategists. AI is not really needed here.
Not sure if this in line with HN guidelines (per "tangential annoyances" rule), but god, that "animation: drift 2s ease-in-out infinite alternate;" on body kills me, literally - it redraws the whole page with 120 fps, draining the battery, while nothing changes. I don't think it is even "too common to be interesting" - I rarely see this (or actually, I see such tech for the first time). It is specifically annoying because the article attempts to talk about modern software quality.
If this firewall is available as a commercial product, eventually it be infected, so there won't be any need to hack any client devices. Since this is clearly a niche product, the device manufacturer won't be able to identify and fix bugs as effectively as companies like Apple do. This follows ROSKOMNADZOR recommendations: to install a middleware device that decrypts, stores, modifies, blocks and redirects all traffic depending on rules submitted from external party.
It makes zero sense to measure performance without measuring correctness. Especially, when you use LLM. Here is even faster asin(): `return 0;`
This precisions should be measured in avg and worst ULP, not in "charts". A good approximation should also give exact results in critical points (-1/0/-1 in this case).
The "faster" version gives this:
asin(0) = 6.75268e-05 (double precision)
Which gives around 5e+15 ULPs, while common libc math implementations targets 19 ULPs (but will be 0 for asin(0)).
When I played it (it was invitation-based, but it was not difficult to get an invitation), there was nothing related to photo. Pure GPS-based navigation (including "helicopter" players). No idea if it includes AR features now.
2) Please don't call black overlay rectangles as "Redact" - it is maliciously misleading. I checked https://pdfcrowd.com/inspect-pdf/ and I see original parts that I covered with these rectangles (images are stored twice: as originals and as images with cut out regions).
reply