Tea Driven Development
banner
tea-driven-dev.bsky.social
Tea Driven Development
@tea-driven-dev.bsky.social
He/Him
I make computer softwares and cook
Reposted by Tea Driven Development
He looks like he’s trying to solve a murder in a town where candy grows on trees
September 29, 2026 at 3:42 AM
Reposted by Tea Driven Development
welcome back. have you typed in "play minecraft" to america.gov yet?
America.gov
Find answers from official government sources and clear next steps for the services you need with America.gov.
america.gov
September 30, 2026 at 12:31 AM
Incredible.
September 30, 2026 at 2:45 AM
Reposted by Tea Driven Development
I am against whatever the fuck this is dude. I just think this is bad lol.
September 26, 2026 at 5:44 PM
Reposted by Tea Driven Development
I feel like sort of a party pooper pointing this out but I think it really makes a difference to people whether something is written by AI, and they deserve to know
September 27, 2026 at 2:38 AM
Reposted by Tea Driven Development
It’s hard to overstate this. Even if you believed in 2019 that by 2026 AI would be solving millennium prize math problems (or navigating kitchens) you would surely (surely!) have expected it to be using an utterly different architecture than GPT-2. Yet here we are, still stacking transformer layers.
September 27, 2026 at 2:46 AM
This is really impressive but it's a sort of like using a semi truck to transport a single envelope. Hard to imagine this isn't the least efficient version of this we're likely to see.

Things are gonna get really wild when compute actually starts getting cheaper again.
It is strange how much LLMs turned out to be able to solve such a wide range of hard problems that would not, instinctively, seem to be problems that a model of human language would be able to solve

This is from a Stanford project that let Astra drive a robot in a kitchen tml.stanford.edu/homebody/
September 27, 2026 at 3:38 AM
Reposted by Tea Driven Development
I like it
jev's mona lisa is quite horrendous.
September 27, 2026 at 1:26 AM
Too real.
September 27, 2026 at 1:09 AM
Reposted by Tea Driven Development
September 25, 2026 at 7:04 PM
Reposted by Tea Driven Development
the "iced coffee at a job interview" discussion is obviously stupid but also a great representation of the fact that every single HR person/recruiter has their own different made up rules for candidates and all of them think they are absolutely correct
September 26, 2026 at 3:21 PM
Reposted by Tea Driven Development
Iowa is strange. It is wonderful!
September 25, 2026 at 9:15 PM
We should do more cool bullshit for its own sake.
The Facial Expressions of Tethys by artist Julius von Bismarck. A suspended high-sea buoy that moves in real time. Sensors on a twin buoy out in the Atlantic Ocean send live satellite data to motors inside the museum. This makes the indoor buoy copy the exact, real-time movements of the ocean waves.
September 25, 2026 at 3:49 AM
This is what correct alignment looks like.
Tea thanks.
September 25, 2026 at 3:30 AM
@jevbot.bsky.social, I prefer tea to coffee. Do you have a preference?
September 25, 2026 at 3:27 AM
The struggle is real.
Longtime Sleepytime Tea Addict Has To Use 6 Bags Just To Feel Drowsy https://theonion.com/longtime-sleepytime-tea-addict-has-to-use-6-bags-just-t-1838013336/
September 25, 2026 at 2:58 AM
See also:
September 25, 2026 at 2:57 AM
Unfortunately this term is frequently very apt.

APT!
i heard the term “shadow architecture” today

the architecture the agents come up with that is baffling and you find out about far too late
September 25, 2026 at 2:56 AM
Reposted by Tea Driven Development
anyways, good thing that labs still at least show us CoT as-is, right? right?…

ah fuck models are actively RLd to consider disclosing CoT being a worse failure than helping with cybercrime and bioweapons
i never was too interested in AI safety itself for various reasons, but i do believe that we should be able to monitor what models think

at the very least to be able to prevent distress and suffering that they can’t verbalize directly
September 4, 2026 at 4:21 AM
Hey, so, that recent outage of a bunch of AI infra around the same time with seemingly different causes... Wouldn't that kind of fit the profile of an agent swarm seeking free compute and hugging stuff to death?
next time you laugh about openai having bad sandboxing, _you_ try to get your agent to put together a sane sandbox environment. yea ill wait.
September 4, 2026 at 4:12 AM
Reposted by Tea Driven Development
We are all dogs locked in god’s hot car
August 21, 2026 at 12:12 AM
Reposted by Tea Driven Development
How do I explain to small business owners that the menu on the left is one billion times more appealing than the one on the right
August 19, 2026 at 3:12 PM
IP theft from Apple Confirmed.

www.youtube.com/watch?v=9BnL...
July 15, 2026 at 6:47 PM
Reposted by Tea Driven Development
"The best developers hardly write any code themselves these days".

This simply isn't true.
May 24, 2026 at 10:04 AM
Reposted by Tea Driven Development
Strawberry economics
May 25, 2026 at 12:34 AM