Simple, for what I'm doing Opus 4.6 (and before that, Opus 4.5) are just much better at following my instructions and achieve consistently better results.
From what I've been gathering, this split in success seems to depend a lot on the types of tasks, the domains / programming languages / frameworks used, and style of prompting.
I couldn't get 5.2 to follow instructions for the life of me, even when repeating multiple times to do / not do something. 5.3-codex was an improvement and 5.4 while _usually_ decent still regularly forgets, goes on unnecessary tangents, or otherwise repeatedly stops just to ask for continuation.
Sure, I'm paying 3x more per request, but I'm also doing 5x fewer requests.
Or well, used to. Still bummed about them dropping 4.6.
Yes, except that in modern infra i.e. WiFi 6 is 1024-QAM, which is to say there are 1024 states per symbol, so you can transfer up to 10bits per symbol.
This would help a lot. Many European startups are strongly local (also in talent search), because while moving is simple, share distribution and ownership structures are anything but, and investors usually don't want to bother with local regulation on that they don't even know.
I agree. It's funny that this is one of the cited reason for the (relative) value suppression of tsmc, but the same factors should apply to Nvidia too.
it's sad that this is not the default behaviour. hopefully the stop killing games movement will put something similar into law with potentially further-reaching side-effects eventually. Because frankly, sunsetting products like this should be common sense, not the exception it currently is.
The base material needs to be of a minimum quality for that experience to be enjoyable, I presume. I totally see the value of that, which is one of the reasons to sometimes reduce playback speed below what my default would be. Writings from Tolkien are maybe some of the best suited for that, but I'm not 100% convinced it would work for what I'm reading currently, so I might just try it anyway - the Foundation series from Asimov.
well, this is where being pedantic bites me in the a* again. Our codebase has been mostly pyright-focused, with many very specific `pyright: ignore[...]` pragmas. Now it would be great if ty (pyrefly has an option!) could also ignore those lines. There's not _that_ many of them, but .... it's a pain.
yeah it's only correct in 99.7% of all cases, but what if it's also 10'000 times faster? There's a bunch of scenarios where that combination provides a lot of value
Abstract—WhatsApp, with 3.5 billion active accounts as of
early 2025, is the world’s largest instant messaging platform.
Given its massive user base, WhatsApp plays a critical role in
global communication.
To initiate conversations, users must first discover whether
their contacts are registered on the platform. This is achieved
by querying WhatsApp’s servers with mobile phone numbers
extracted from the user’s address book (if they allowed access).
This architecture inherently enables phone number enumeration,
as the service must allow legitimate users to query contact
availability. While rate limiting is a standard defense against
abuse, we revisit the problem and show that WhatsApp remains
highly vulnerable to enumeration at scale. In our study, we were
able to probe over a hundred million phone numbers hourly
we started a purely-frontend project with nextjs but moved to react-router pretty soon. Sure, it can be convenient (when it works), but you can't really see or understand how or how to control it, and black magic breaking is the least of what you want. Much happier now.
because it _does_ provide a number of benefits (potentially fewer initial round-trips, more dynamic routing control by using UDP instead of TCP, etc), and is a userspace softare implementation compared with a hardware-accelerated option.
QUIC getting hardware acceleration should close this gap, and keep all the benefits. But a kernel (software) implementation is basically necessary before it can be properly hardware-accelerated in future hardware (is my current understanding)
certainly not free, but with F1TV Access you have access to historical and real time data through the API. There should be decent documentation on it out there, otherwise projects like Multiviewer or this TUI wouldn't be quite realistic.