have an interview tomorrow for an ai engineering role
what do i do, what do i do 🤯
oh wait, i have seekho:
→ a sprint built for my company and the hours i have left
→ the questions to answer out loud
→ how they actually interview
https://t.co/OQnVv8aIhD
My mind has found so much respite in reading a good book. A good fiction book, that is. I hate men screaming at me on how to live my life. That's a nonfiction self-help book. A good fiction book (if the writer has suffered and is good at describing human nature) cures you so deeply and so thoroughly that the boring platitudes of an Atomic Habits can't touch it. Don't get me wrong. The latter is great reference material for the gooner cucks of our century who need to be rescued from the black pit of the bottom. But if you're not that and only have mid-level and largely doing well (or on the trajectory to do well), it's a John Steinbeck or a James Baldwin who will do more more for you than a James Clear (and I'm saying this with so much respect for Clear btw). I find that self-help nonfiction is good for a certain Type-A individual who follows orders and fits neatly into an institution. But its not good for a low status, neurotic rebel. What will save him is the fucking story. The story will tell him he's not alone. It will remind him that other cerebral fucks before him have confronted the complexity of life and then through engaging with it have relaxed their sphincter in a big way.
Read stories.
it is everything between,
“I have an interview” and “I’m ready”
presenting to you "SEEKHO"
i will share more details, as soon as they will be built :)
stay tuned
1/
Can Video World Models Track Unobserved World States?
Video world models increasingly look like simulators. But rendering plausible video is not the same as knowing what the world is doing.
We test this with a video shell game. A ball hides under one of five identical cups, which are swapped N times out of sight. Every frame is easy to render. The hidden arrangement is not.
A standard causal DiT trained on five swaps still renders ten swaps at 40.4 dB PSNR, but lifts the wrong cup. LaCT (TTT-KVB w. SwiGLU) tracks the correct final state with similar visual quality.
long-term planar tracking that needs nothing more than four clicks 🪟
SAM-H + WOFTSAM: gives you a homography per frame plus a rectified view of the plane, training-free, tracking any plane even on a shaky recording
▶️ https://t.co/1iTM61NnHj
why do I have to dismiss an ad after skipping it now @YouTube?
is the idea behind promoting premium to downgrade the free tier user experience?
seriously annoying