Alexandru Coca

Happiest day of my life @ElinaSvitolina good luck in the finals 💪 we'll be there to support you 😍

Choose problems and people, not brands or timelines.

Another crazy round of acceptances today. 🔥 Join our research staff this Friday as we share more about our goal of building intelligence that continously evolves. Incredible to see the response and energy 🎉

@juliarturc AI for drug design 😉

Why so many of us feel career-homeless in tech: >Startups full of fraud, grifters and short-term thinking >FAANG full of politics and slightly behind >Frontier labs in a race with no morals >Academia full of title collectors >Content creation ridden by AI fakes and sensationalism Who is starting the renaissance and how do I get in touch with them?

I completely understand the sentiment. I don't think all of us are designed to handle the level of stimulation that social media gives us. Most of us are not meant to interact with this many people. The brain can handle so much before it gets fried. It's ok to want to return to baseline.

AHT AHT Lisa Mason didn't win bronze at British Championships at 33 years old to be erased like this! youtube.com/watch?v=uRZ8pc…

@sarahookr 🥰

We are like butterflies who flutter for a day and think it is forever. Carl Sagan Time is the most important resource we have. Work on something that could move humanity, work with people you respect + don't forget to find someone you love along the way.

@sarahookr +1

this is honestly why the most important measure of progress in building true intelligence is our ability to eliminate prompt engineering.

@BlackHC 😂

Many CS PhD students can probably also ask themselves if CS professors are still necessary 😅

Some people are fixated on their competition. It rarely leads to greatness.

i got early access to the adaptive data platform by team @adaption_ai and it’s genuinely very cool. so what is it? it’s a platform for detailed data optimization where you can adapt datasets to your specific needs. i tested it on both reasoning datasets and multilingual indic datasets and the data enhancement capabilities stood out the most to me. it improves dataset quality through things like prompt rephrasing, prompt deduplication and adding extra context to prompts, (which i think is especially useful for strengthening and improving the output quality of reasoning models). there’s also a diagnosis dashboard that compares and grades the quality of the original dataset against the adapted version and the improvement is quite noticeable. some upcoming features i’m especially excited about are multilingual dataset adaptation, the ability to generate custom datasets from your own specifications and support for applying constraints such as safety, tone and length directly to the data.

have we seen these yet

@AliceKinsella16 @whiptriplefull Bless, strength level will be through the roof. Triple layout as Kinsella, here we come 😎

@whiptriplefull And I live in an apartment with no lift🙂🙂🙂🙂🙂🙂

i’m sorry but SUPER MUM OR WHAT?! Alice Kinsella you are amazing

@slutgymnast @youwannacranson @Pamchenkova cc @marinettetikky, when you don't have anything else to do.

@youwannacranson @Pamchenkova I would love to see her do a standing Arabian

her front tuck is so carly patterson coded with how soft it lands. if you look closely she doesn't really let her back heel hit the beam i adore this series and can't believe how consistently fast it's been

@youwannacranson @EricGymFan truth

shocked to learn FRG cares about olympic qualification after the various leadership changes made last year

One of my favourite piked full-ins. Just beautiful.

Vote for barb in European Gymnastics' gymnast of the year 2025 poll 👇 take.quiz-maker.com/Q653XZTIQ#

@c072133 Not particularly upset about their potential decl tho

I am going to sound insufferable when I say this: I don’t want Simone, Suni, Jordan and Jade come back for another Olympics, the US team will be unstable and in significant decline following LA as a result of lack of experience and no veterans to lead in the next quad

@giffmana 😂

So, when I have GitHub Copilot point out a typo in my PR, and I'm lazy, I click "Commit suggestion" and it makes a commit. And I make a lot of typos in comments, so it happens on every second PR. This is what Satya meant with 30% of MS's code written by AI, right? ... right??

I've reached a point in coding my video animations where I still need to baby sit AI for hours. But compared to a few months ago, it's no longer because it does silly things (messing up alignment or failing to judge in 3D). Now, the problem is me: I don't know what I want right from the start, so I need to iterate until I'm happy. It's hard to assign credit, many things changed: >Opus 4.6, Codex 3.5 got visibly better >I've fine-tuned the guidelines & accumulated more code / samples >I've added more instrumentation (self-critique loops, etc) >I've narrowed down an effective pseudo-language for describing scenes But credit is a moot point anyway. My take-away is that we're reaching a point where things just... work. I feel no more fomo to keep up with agentic this and that. I don't believe in the "keep up with the tools or die" fear mongering. Everything is converging to simplicity, because that was the point of AI all along. The obsession with tools is procrastination, and we're finally freed to work on THE ACTUAL THING.

@juliarturc Ha, thanks so much. I often feel lost in the hype and try to filter the noise, so this is a useful barometer from someone I look up to! Great post!