They announced it live on stage at their conference, apparently with no accompanying blog post.
HN user
Mattwmaster58
mattwmaster58 at google's popular mail service
but I've seen it happen to the last technically-inclined 500k+ job class that was untouchable by all accounts
And what job class was this?
The effect appears to be that all recent frontier models are inaccessible. See: https://old.reddit.com/r/GithubCopilot/comments/1toa9tf/mode...
For orgs, each user was allotted their own quota. For messages beyond that quota, a pooled budget is available.
FYI, these are the multipliers for annual plan. I would hazard a guess most people are not on an annual plan
In terms of raw token cost, I've seen a couple providers at (all prices in terms of Mtok) $0.95 input/$0.15 cache input/$5 output vs $3 input/$15 output for sonnet.
Task prices of courses will be more interesting - a dumber model may use more tokens to get to the same goal.
copy/paste doesn't tell you much - here's the text/html content they put on your clipboard if you're curious. Apparently GDocs supports this out of the box, just hides it from the selection box. Which makes sense given that it doesn't support any font.
<html> <body> <!--StartFragment--><meta charset='utf-8'><meta charset="utf-8"><b style="font-weight:normal;" id="docs-internal-guid-8b11d82e-1a25-4b6a-be64-ebdd55b2a698"><p dir="ltr" style="line-height:1.2;margin-top:0pt;margin-bottom:0pt;"><span style="font-size:22pt;font-family:'Facebook Sans',sans-serif;color:#000000;background-color:transparent;font-weight:400;font-style:normal;font-variant:normal;text-decoration:none;vertical-align:baseline;white-space:pre;white-space:pre-wrap;">I just stole Facebook Sans</span></p></b><br class="Apple-interchange-newline"><!--EndFragment--> </body> </html>
That's exactly where my mind went as soon as I read the title. HN rules say to "use the original title, unless it is misleading". I think the original title meets the misleading bar but I can't speak for other readers.
It converts back to paid automatically if you had an existing paid subscription before. No other cases. In any case, this is still a valuable service they are providing for 6mo for free, which many will appreciate even if the goal is to recruit more users.
I'll second this. I used opencode + opus 4.6 + ghidra to reverse engineer a seedkey generation algorithm[1] from v850 assembly. I gave it the binary, the known address for the generation function, and a set of known inputs/outputs, and it was able to crack it.
I assume you're talking about 50t/s? My guess is that providers are poorly managing resources.
Slow inference is also present on z.ai, eyeballing it the 4.7 flash model was twice as slow as regular 4.7 right now.
It's generally very helpful - someone else mentioned here the fundamental problem is lack of a tight feedback loop. It doesn't perfectly replicate the GH environment, but for my use case that doesn't matter and it's super nice to have.
As far as I know, the HFCS vs Sucrose is unlikely to be the reason for the difference in taste. I'm basing that off this video: https://www.youtube.com/watch?v=NY66qpMFOYo
TLDR: carbonic acid breaks down sucrose to glucose/fructose anyway
A couple years back I was looking for this sort of solution and ended up paying money to buy FilterBox which I've found to be good.
There are certain apps that I would love to be able to uninstall but have to keep for one reason or another, so I really appreciate apps like these which prevent attention-stealing notifications from making it through :)
It's a shame that AI is ruining certain phrases, the "You’re absolutely right" was appropriate but I've been trained reading so many AI responses to roll my eyes at that.
This doesn't seem like a realistic threat to me. Under what circumstances are you not pretty much completely pwned if an attacker could start their own processes, or have root access?
This sort of seems like saying IF an attacker gets the keys to your car, they could install a module that would allow them to come back and steal the car with a push of a button. Technically true, but they could also just steal the car straight up, or do any number of other things.
I once merely mentioned the words “Heart Attack” on a plane and was kicked off by the flight attendants.
Well now you have a chance to tell your side - were you merely sitting and just uttered the words "heart attack" for no externally apparent reason?
vx-underground claims to have communication with the group, and this post of theirs adds to the support agent theory: https://xcancel.com/vxunderground/status/1976238815665856646
they were able to compromise Discord Zendesk by compromising a "BPO Agent" (outsourced support).
Of course, as is tradition, it is also entirely possible they're lying
Is that possible?
This comes up every time npm install is discussed. Yes, npm install will "ignore" your lockfile and install the latest dependancies it can that satisfy the constraints of your package.json. Yes, you should use npm clean-install. One shortcoming is the implementation insists on deleteing the entire node_modules folder, so package installs can actually take quite a bit of time, even when all the packages are being served from the npm disk cache: https://github.com/npm/cli/issues/564
I've found splitting up my ultrawide into 6x2 cells, then you can use Ctrl+Shift to select every cell your mouse enters additively. I've wanted something like this for linux for a long time but haven't found anything.
Off the top of my head, you could include your own checksum in the payload. Their code only modifies the address. Nothing would prevent them from reverse engineering checksum, too.
There are ways to detect a replaced/proxied global window function too, and that's another arms race.
As others have noted, npm install can/will change your lockfile as it installs, and one caveat for the clean-install command they provide is that it is SLOW, since it deletes the entire node_modules directory. Lots of people have complained but they have done nothing: https://github.com/npm/cli/issues/564
The npm team eventually seemed to settle on requiring someone to bring an RFC for this improvment, and the RFC someone did create I think has sat neglected in a corner ever since.
You can link as many bank, credit card accounts as you want with one SimpleFIN account.
The website says "Connect up to 25 institutions and 25 apps"
interesting, does work for me
It looks like beartype supports the same sort of implicit decoration, because there's mention of an explicit API:
Beartype now implicitly type-checks all annotated classes, callables, and variable assignments across all submodules of all packages.
A similar thing happened to me with my Amazon account with a forgotten password. I ended up just creating a new account.
Is the reason a buck converter is required so that it can increase I? I = V/R, by increasing V we can increase I?
Looks like someone took you up on it, instead of a showing a price the Amazon link shows out of stock.
can you cite the studies? Considering LASIK and most things I google for are just blogspam, I prefer peer reviewed papers