Tensor Fiend

Opinions here are filtered from my J-Space. Building

[Tensor Fiend @tensorfiend]

LEARN RL BY END OF 2026
LEARN RL BY END OF 2026
LEARN RL BY END OF 2026
LEARN RL BY END OF 2026
LEARN RL BY END OF 2026
LEARN RL BY END OF 2026

[Tensor Fiend @tensorfiend]

You can’t be clearer than this when naming a restaurant 😂. My Korea trips are always full of such finds

[Tensor Fiend @tensorfiend]

@zeeg Cloud that is only accessible by Neolabs.

[Tensor Fiend @tensorfiend]

@Sauers_ I meant similar steering analysis 😅. Actually tried it at work for a smaller model and got some interesting results!

Will use @_lyraaaa_ repo now to redo things.

[Tensor Fiend @tensorfiend]

I know it’s a great find but @Sauers_ has done it already and has more other finds too

[lyra @_lyraaaa_]

if you have a qwen3.5 4b around, please try steering neurons L14 N8949 +1 and L13 N7956 -1
now you have evil qwen

recipe: dose × median(‖resid‖ leaving layer L, tokens 1:) / ‖W_down[:, n]‖` at the down_proj at every input

[Tensor Fiend @tensorfiend]

Working on an idea that involves couple of things I wanted to work on for a longtime.

[Tensor Fiend @tensorfiend]

@smallest_AI Now do Breathless by Shankar Mahadevan

[Tensor Fiend @tensorfiend]

What did I just seee???😭 I just reinstalled instagram today

[Tensor Fiend @tensorfiend]

@SchopenhauerNow Lord Krishna: It's selfless action

[Tensor Fiend @tensorfiend]

Added a handwriting toggle on my blog ✍️. Realized my handwriting is still bad and actually got worse.

@GeminiApp Argon did pretty decent job at giving me a template to write and also creating a package to automatically update it with any hand writing 🔥. Thanks for the idea @ivanleomk !!

[Tensor Fiend @tensorfiend]

@ivanleomk This is cool. Maybe I should have a toggle on my blog to display with my handwriting.

[Tensor Fiend @tensorfiend]

@khoomeik All it takes is one bad government ruling for a couple of decades. Hope it gets better.

[Tensor Fiend @tensorfiend]

@TanakaSamukange @AbhinavXJ This is actually the best and the only usecase🤣🤣💯

[Tensor Fiend @tensorfiend]

This is the most realstic AI generated modality I have ever experienced.

[Ragavan Thurairatnam @RagavanThurair]

We made the first AI model you can have a realistic conversation with that was trained entirely on speech!

Dessa is a full-duplex speech-to-speech model that you can run on consumer GPUs.

This is a GPT3 moment for speech.

[Tensor Fiend @tensorfiend]

@keyanzhang Probably this one lol

[Tensor Fiend @tensorfiend]

@lennysan Tomtoc travel backpack. I travel regularly and this has been perfect backpack so far.

https://t.co/Seamz3HhEk

[Tensor Fiend @tensorfiend]

@beyarkay It’s too big of a mistake to keep it to themselves ig. They might have not found it

[Tensor Fiend @tensorfiend]

It depends on the officer who took the interview. I know a friend whose visa (B1) got rejected firsttime and then second time he got the visa.

But it’s very difficult to get the interview slot. Entire process was stressful because everything is uncertain. At this point it’s pure luck if you’re completely eligible, legit with no red flags (definitely reject even if there is one small doubt)

[Tensor Fiend @tensorfiend]

@himanshutwtxs This is 2nd find for me this week

https://t.co/vDf6y9fqVc

[Tensor Fiend @tensorfiend]

Until today, I didn’t know the spelling of Ampersand 😭.

Looking forward to trying out Ampersand 🙌🏾

[Tensor Fiend @tensorfiend]

@Kaju_Nut That’s the unreleased model from Anthropic. Jokes on you guys.