There are candidates for related shortest-in-a-publication out there. For example the shortest abstract I've seen in a computer science paper was "No".
The title of the paper asked some question, and the abstract was answering it.
The internally-reported benchmarks (Frontier-Bench, AutomationBench) and the customer quotes (Cursor, Devin, Lovable) all have a commercial stake in the outcome
worth waiting for independent evals before drawing conclusions.
Well, the old Euro design avoided making references to specific countries and chose to use generic images of bridges for a reason. You can only put so many faces on your currency. You can't choose any set of faces without people from some countries being left out. And if you leave any country out, you are just handing munition to Eurosceptics in that country to fire at Brussels — "dear citizens, don't you feel the EU is really just about Germany and France? See, the money you spend everyday does not even bear the image of our countrymen!"
You can of course say people should just grow up and not be affected by petty sentimental arguments like this. But keep in mind, the EU is as much an economical and political project as it is a cultural project. You have to win people's hearts. And you do so by not making them feel neglected.
Very pedantic of me, but I figure it’s interesting to note: technically a release on PyPI can have zero files or even one or more wheels but no sdist. The former is a degenerate case that users don’t normally see, and the latter happens if the user chooses to only upload wheels (or their sdist upload fails for whatever reason).
(This doesn’t change your observations at all! Just as a demonstration of how Python packaging’s data model can be unintuitive.)
It was after decades of stripping money from work into larger pools of capital, thanks to Reagan and Thachter style deregulation; so much money looking for unreasonably high ROI and unencumbered by the need to get consent from the governed or pass inspections from regulators causes trouble where ever it touches.
True! You’re less likely to be surprised by banding on a display, because you can see it before it’s a problem, and it’s less likely to cost money, but for sure the need for dithering is still there, especially for displays that can do more than 8 bits per channel or can do high dynamic range.
> Consider that a grayscale gradient can only have 256 distinct levels.
This is true, but also keep in mind that for print and low dynamic range displays we humans only need approximately 256 levels, as long as the levels are perceptually spaced out evenly. 8 gamma corrected bits per channel actually made sense for a long time, for both print and for TVs & monitors, until HDR displays came along, relatively recently. Needing dithering for 8 bit images displayed on a 10 bit display is perhaps less surprising than needing dithering even when your display and your printer are both only 8 bits, right?
Yea, but if humans went extinct tomorrow, very little human artifacts would remain after millions of years. Indeed, most would be gone well before that. How much stuff from the 18th century still remains, in percentage terms?
More like the credit card company can initiate automatic withdrawals from your bank account without your permission if you previously authorized the company to do autopay.
(Less so if others are exculpatory. Which is not the case here. A known false positive is not conclusive proof, but neither does it explicitly clear a source.)
But, somewhat ironically, LLMs are unbelievably bad at writing.
I can read thousands of words from my favorite authors without getting annoyed by their tics and their grammar. People like Nikhil Suresh literally make me laugh out loud when I read their writing. But something about AI-slop-prose is just obvious and gross and annoying to read.
I applied for individuals. I did their verification steps and answered questions in a few minutes, so that was indeed easy. The problem is that was all that happened. No followup, no access, no denial. When I tried to reapply it tells me I can't apply again.
OpenAi (and Anthropic) have no incentive to allow security access to individuals. I'm not a deep-pocket org or influential gov agency. Allowing individuals increases the risk of bad press (what if I do something naughty and talk about it?) so best to ignore us.
What's the source for that? I see a combined revenue of ~33B by looking on the Internet. And we should assume that that's after some crazy financial juggling
I agree code produced is a poor measure. And I agree that some of the bigger pieces of coding were everything you mention, coding still did take _some_ amount of time.
Usually in a day I have one agent going on whatever my large, main task is, generally feature work of some sort. Previously this is what most of my sprints would have been filled with.
But now, while that's going I have one or two others simultaneously working on crashes (I do mobile), other bugs, a small refactor, or maybe a small enhancement.
And finally I usually have another where I'm querying analytics or working on higher level questions to try and help guide what we should do next, either from a product or technical perspective.
A single one of these are all sped up by using LLMs but to be able to do several in parallel makes me much more effective than I was before. This is why I believe that those who don't use these tools properly will be left behind.
> … and assign them to create and scale production chains that go from natural resources to finished war machines…
so yes, you think claude’s gonna get a skill any day now to code up some nukes.
Your scenario assumes fully autonomous ASI. That’s just not what LLMs are. So then the discussion is more that ASI is inevitable. That’s much more interesting but very different from the point release of Opus.
> let's not downplay the blast radius of basically every attack that involves telling the user to do things.
That radius would be negligible should the things that must remain secret remain offline.
But no, that's a luddite thing to even think about in an all-connected ever-online world where even wiping one's ass is done via of a swarm of dedicated apps.
On the other hand, there has always been a way to extract secrets from a granny via a mere voice call, so no amount of dumbing it down would really help.
Indeed. Did a rapid check on T-rex, just as an example.
Was around for max 3.6 million years. Max population at any one time must have been around 20k individuals. We have about 150 skeletons.
So current human population is 400 thousand times the maximum T-rex population at any one time. A few decades are enough to produce more human individuals than the entire T-rexes history. We also have the ability to go (and die) pretty much anywhere, so there will be human skeletons in every crevice, depth and environment on Earth. We also produce an absolutely staggering amount of artifacts of all kinds, from plastic straws to giant tunnels and artificial caves. There are zero chances an advanced civilisation on Earth won't be aware of us in 400 million years.
