Register and share your invite link to earn from video plays and referrals.

Aditya
@adityavg13
Huge Lakers fan | M.A. Global Security Studies | Currently doing stuff with AI - Token Compression and Efficiency.
914 Following    1.3K Followers
I actually think I’m in a super niche minority where I use AI heavily and enjoy it, but also am pretty far left/have socialist ideologies. It’s like this weird thing because my sister and I share the same opinions but she HATES AI.
Show more
I need a vacation to somewhere cold but sunny. This hot ass humid ATL weather is frying me.
I sent Cursor Grok 4.6 on a mission to make me a Qwen 3.8 46B A13B MoE by merging two 27B models together. It has currently been running for 2 days and I only have 1 spark so it’s been a little slow and a few hiccups here and there but so far it’s going well, i can’t wait to see what the model is actually like when it’s done (it’s my first time doing something like this so it’s genuinely interesting to read and see what its doing) The architecture is labeled as “Backbone-Anchored Layerwise Expert Selection Hybrid” or BALESH (name is iykyk)
Show more
Can we get an actual usage increase on supergrok heavy before then and when this comes out? For the price it is, we should be getting a LOT more than we do.
Grok 4.7 is coming... and this is going to be something completely different Initial training is already complete SpaceXAI is now doing supplemental training with a massive amount of SpaceX company data And that data advantage is insane SpaceX has spent nearly 25 years solving some of the hardest real-world engineering problems humanity has ever attempted The level of precision SpaceX operates on is insane... it's unlike anything anyone has ever seen or attempted before Rockets, rocket engines, satellites, manufacturing, autonomous systems, launch operations, extreme reliability, engineering at the limits of physics Now Grok gets to learn from that engineering culture and data Grok 4.7 is already significantly better than 4.6 and is expected to exceed every model currently available And when it comes to real-world engineering, Elon said he would be shocked if any model beats it Grok 4.6 brought SpaceXAI back to the absolute frontier 4.7 is being trained to push beyond it This one is going to be very special
Show more
monorepo'd and full integrated instead of dealing with 4 different repos LOL
Something about truly being alone that makes you think all the worst type of thoughts. I’ve said it before but I could truly turn off my phone for a month, turn it back on and not see one missed call or text message. That’s where my life is at.
Show more
So another reset???
This celebration is moved to tomorrow as the button was already pressed today.
Full sub1ms drop in replacement for EVERY tree-sitter (so yes only one program) coming soon.
tree-sitters are so lame, we could just have so much better and faster
Expert Prompting Tips ep1 by Aditya: Tell the model that it doesn't need to worry about being late for it's daughter's ballet class.
Fuck it I’m starting my own open source model company who wants to join. I don’t got any money but i got time and nothing else to do.
People sit here and act like the two heads of both companies don’t hate each other with a passion. Sometimes, that’s life and we just gotta accept it.
Delete your OpenAI account and switch to Grok. Daily reminder that ClosedAI can not be trusted.
Testing uber-1 local model now. It’s based off Qwen 3.9 46B A67B model and runs at 70888 tok/s - pretty fast, unfortunately doom loops quite often.
I'm not gonna lie, if i hadn't ever typed this in before I would have got got by the first link @NousResearch
I need to move to a place where I could actually make friends and get out of the house. I literally don’t say words out loud 99% of the day because I got nobody to talk to.
tree-sitters are so lame, we could just have so much better and faster
Do I buy this just to use Sol heavily for the next month…
GPT-5.6 Sol is now the most affordable frontier model on Devin Desktop and CLI. Today, OpenAI reduced Sol’s prices by 20% for the next 3 months. Last week's 70% discount still applies, so through October 3 you now get a 76% discount off the list price.
Show more
Put this into a markdown, label it whatever you want and then add other metaphors too. Believe me, have your agent read it when stuck snd they really reframe their thinking.
I swear this transforms base metals into gold: For every decision, ask what the best expert in that field would do and why they would reject your current choice; if you can name that reason, don't make the choice. Optimize for what that expert would judge correct, never for what satisfies the stated constraints most cheaply. Every trade-off you take must be stated to the user, never absorbed. specially if injected on each message (not on AGENTS.md)
Show more
I think i got deepseek tweaking...
One of my original projects that I've been working so much on, tokenzero, has officially been archived and sunsetted. I have something brewing very very soon that's going to change the game o7 to the legend, thank you for being one of my favorite repos to work on.
Show more
Custom kernel/tooling going well, clearly it works.