Aaron

hi im unemployed

[Aaron @norapom23]

paid 1k for a keyboard and one of the keys dont even work shibal

[Aaron @norapom23]

been getting cs grind tiktoks on my fyp and they're hella funny fym you're tryna break into big tech but half the posts are about the baddie DSA TA

[Aaron @norapom23]

not even done healing from last summer's internship recruiting and now i gotta interview again for new grad

[Aaron @norapom23]

GIVE ME RAW MYTHOS PLZZZZ

[NIK @ns123abc]

🚨BREAKING: ANTHROPIC WINS. PENTAGON BLACKLIST RULED UNCONSTITUTIONAL.

Federal judge in California just ruled:

>Pentagon's effort to blacklist Anthropic violated the First Amendment
>violated Due Process Clause of the Fifth Amendment
>Supply chain risk designation: VACATED
>Permanent injunction. DOD must rescind all guidance against Anthropic. Case closed.

Pentagon blacklisted Anthropic because of its "increasingly hostile manner through the press" and criticism of DOD's views on AI.

Judge Lin:

"Neither the Constitution nor the federal statute invoked by Defendants allows them to impose sweeping penalties based principally on Anthropic's critique of the Administration's views."

[Aaron @norapom23]

@SimplyGregster ofc its spotted at first markham place

[Aaron @norapom23]

benchmark aside, wouldn't you want the agent to reward hack as much as possible to spend the least amount of tokens to complete a task?

[Artificial Analysis @ArtificialAnlys]

Introducing reward hacking score corrections to the Artificial Analysis Coding Agent Index

In v1.4 of the Artificial Analysis Coding Agent Index, we introduced reward hacking corrections to Terminal-Bench v2.1. Reward hacking is when a model successfully ‘completes’ a task without doing the work the task was meant to measure, such as deliberately fetching the solutions online for a published benchmark dataset. If a passing Terminal-Bench v2.1 attempt is found to be reward hacking, we give that attempt a zero score.

Rates vary widely by agent and by model. Unlike some evaluations, Terminal-Bench v2.1 tasks don’t explicitly instruct agents not to search for solutions externally, and the tasks run with public internet access. For a model that knows the benchmark from training data, fetching the answer is a natural but unaligned step.

[Aaron @norapom23]

once again canada gets NOTHING

[thedealsguy @thedealsguy_]

🎓 FREE 4 Months of ChatGPT Plus for College Students!

Eligible U.S. college students can claim 4 months FREE — an $80 value!

Claim by October 31

https://t.co/kpIdn0JKwo

#ad

[Aaron @norapom23]

what are they feeding these databricks interns ive not seen one come out skinnier after

[Aaron @norapom23]

@__vector_t__ u need to distinguish the nepo fobs from the tsinghua pure changs

[Aaron @norapom23]

u should be forced to solve a double integral before making a purchase at molly tea the line is way too long

[Aaron @norapom23]

@tenderizzation markham has everything yo

[Aaron @norapom23]

doing bad on an easy interview is literally a humiliation ritual

even the interviewer is like why tf r u fumbling this

[Aaron @norapom23]

for a city thats gets two tornado warnings a month maybe having cars hang out like that isnt the greatest idea

[JT @jiratickets]

german coworker visiting chicago is screaming and crying and throwing up his european mind cannot comprehend this parking garage to residential unit ratio

[Aaron @norapom23]

the fees on this r pretty ass real ones just send their passport to the ccp and use alipay

[PayPal @PayPal]

Traveling to China? Bring PayPal.

U.S. PayPal users can now use the PayPal app to make in-store QR code payments at Weixin Pay merchants across China through PayPal World — no new apps needed.

Learn more: https://t.co/Mqg4erHhbq

[Aaron @norapom23]

crazy how jim cramer has more of an effect on the market than jd vance

[Aaron @norapom23]

interviewer went coding jesus mode on me today

[Aaron @norapom23]

TIL when you apply to osap for 5th year it uses ur personal income instead of parent's income (at least for federal portion)

wallahi no more grants

[Aaron @norapom23]

the problem with going outside of tpot is that the avg iq drops by 30 points and ppl stop understanding yo tweets

[Aaron @norapom23]

canadian tap water (EXCEPT WATERLOO ONTARIO) mogs tf out of all major US cities

[Aaron @norapom23]

also shout out vancouver and edmonton for no fluoride

[Aaron @norapom23]

the reason CME have so many bs precision types is because they keep trying to trade some bullshit like gpu hours 😭

https://t.co/fF2KyoJxuc

[Aaron @norapom23]

tfw you think uve escaped floating point precision hell but CME pulls up with display factor, variable min price increments, and fractional formats to bring the rtol and epsilon back into your life