wyre
wyre on Hacker News
Thats true. I wasn't considering that they aren't really doing anything novel with the information, just using it as way to harvest emails or relay the information..
I think data is fine, especially since the LLM will go to great lengths to site it's sources. Blog posts on the otherhand...
5 years with 800 million parameters would take over 18,000 changes an hour, 24/7/365 to change every parameter.
They are trying to achieve non-corruption. It was the leading message of the Student List
I haven't used CC in a while, but I would bet this is sol that made this.
however I still get the same issues with sol getting more apologetic and is more likely to stop working when i tell it is wrong, instead of fixing course and trudging ahead.
I don't have the math knowledge to answer your question directly, but Terrance Tao has posted his own session's, for reference:
https://chatgpt.com/share/6a5fdc7a-d6f8-83e8-bbea-8deb42cfed...OAI is not simply prompting "Prove exponential time hypothesis false. Make no mistakes" for the same reasons you can't one-shot "Write a fully standards compliant internet browser from scratch"
If there was a delivery driver's union, I think the downsides of having too many platforms dramatically reduce.
So is Uber, AirBnb, McDonalds, and plenty of other billion dollar companies with externalities that negatively effect society.
Are you really asking if OP believes that ICE agents should face justice for the crimes and abuse they have committed?
I used MiMo for a couple of days waiting for my Codex sub to reset. I really liked it, its communication style was a breath of fresh air compared to how stale gpt-sol is.
Comparing the scoring rubric for AAv5.0 vs AAv4.3:
AAv5.0:
25% Knowledge work:
AA-Briefcase v1.1 - 15%
AutomationBench-AA - 5%
GDP.pdf - 5% (-5%)
25% Coding:
FrontierCode 2.0 - 15% (new)
Terminal-Bench4.0 - 10%
25% Science:
Terminal-Bench Science .1- 15% (new)
CritPt v2 (hard subset) - 10% (new version)
25% General:
AA-Omniscience v1.1 - 15%
HLE-Diamond - 10%
AAv4.3:
30% Agents:
AA-Briefcase v1.1 - 15%
GDPval-AA v2 - 10%
AutomationBench-AA - 5%
20% Coding:
Terminal-Bench4.0 - 10%
SciCode - 10%
30% General:
AA-Omniscience (Accuracy) - 10%
AA-Omniscience (Non-hallucination) - 5%
GDP.pdf - 10%
AA-LCR v1.1 - 5%
20% Scientific Reasoning:
HLE - 10%
CritPt 10%
tldr: gained Terminal-Bench Science and FrontierCode, dropped SciCode, GDPval-AA and AA-LCR. Lowered weight of GDP.pdf.
The only metric used are success and fail percentages?
No statistics on tokens usage or wall time?
Asahi supports up to the M3 line. Linux 7.4 supports M4 so surely Asahi will support it soon.
What is it about this MBP that makes it so much more preferable to an MBP with M3 Pro/Max?
No, you have done absolutely nothing to help me understand your perspective. You have only disagreed.
I could be projecting, but I think a lot of people here don't have a deep understanding of blockchains because the past 15 years has been full of crypto scams and would rather put their time, energy, and attention into better technologies. Proof of existence is the first real utility for bitcoin besides as an investment vehicle or decentralized transfer of money, so I think it's fair that other's would want to know how it works.
Go back to Fox News. If you think it's nonsense show your sources.
Do you own stock in Flock or Axon? I don't believe you are actually asking this is good faith.
Flock is a dragnet. It allows police to coordinate around apprehending suspects, which is the definition of dragnet, if you look at any dictionary.
I think its more likely that people don't know about Flock, not that they don't care.
I don't think software devs are the audience for 6.1 Terra, but instead computer work, automation, and technical writing.
There will be a version of Deepseek with the capabilities of opus 5.5, but by that time there will be Opus 6 and it will likely cost the same or cheaper than Opus 5.5 and have nearly unlimited usage for subscribers.
Every model release people say the model is good enough, yet Opus and Sol are still more popular than Deepseek. The frontier is shortening, but there are still considerable room for models to improve. SlopCodeBench is new and not close to being saturated.
No, you are mistaking that for Stalinism. Communism would not have a hierarchy that allows for oligarchy. Read Marx or Kropotkin.
You can get an older Google pixel fold for a fraction of their price on release.
Why would I want the ergonomics of a small laptop when a folding phone and bluetooth keyboard is much better?
What software do you need that android cannot support? Surely ssh or tailscale can fill that gap.
Ooph, this quote sounds so much worse taken out of context.
My understanding from reading the lwn article is Linus's implication is more specifically about filtering out people that cannot write code without LLMs or the willingness to spend the money on the tokens required to uphold the quality of the kernel. He does not want to push out developers that do not use LLMs.
When you say clicks-and-beeps are you talking about music like Skrillex, Alva Noto, Aphex Twin or do you mean something else?
Either way, it sure does explain the popularity. If you consider it as predictability instead of compression, it might make more sense. Listeners have an expectation on what music sounds like and popular music is going to ride that line between the listeners expectations and something more original and stray from those expectations.
I think this could also be why AI music is so bad, because it can hit the listener's expectations, but the ways that it strays from them are seemingly random, which is why a big reason to why it is so uncanny.
This is really cool, thanks for making this. Bookmarked!
Better to have no morals or principles than have any, as I always say! /s
Every company is going to have competitors that would love your business over these major companies. Pick your battles and do your best. It's only hard because the status quo is designed to be stupidly easy.
if you can figure out tab resource consumption you should definitely share. A broswer built for users with hundreds of tabs would be greatly appreciated by many!
It’s about 330 pages and average reader reads one page/minute, so 6.5 hours, on average