Krish
25 posts


@krishgarg and I built Shard, beating @GoogleDeepMind's TurboQuant on KV cache compression.
10x compression on Llama-3.1-8B-Instruct at 8K.
NIAH recall: 1.000.
Keys: RoPE-aware PCA + int4 fused attention.
Values: Hadamard + VQ.
Same needles. Less cache.
krishgarg.com/shard


English

i just beat @GoogleDeepMind's turboquant
introducing Shard. 10x KV cache compression on Llama-3.1-8B. zero quality loss
- 10x @ 8K context, 11.2x @ 32K
- NIAH recall 1.000 across 4K-32K
- LongBench Δ ≈ 0 vs FP16
turboquant tops out at 4-6x at the same quality. we doubled it.
read more: krishgarg.com/shard
@kirrithan
English
Krish retweetledi


i won the @xai hackathon by making a whiteboard for your @grok interactions
introducing Chorus, an infinite multimodal canvas that diffuses reasoning and context. work solo or collaborate with others in real time
think big, not linear. try it at chorus.ink
built with @akshatdotcom
Berkeley, CA 🇺🇸 English
Krish retweetledi
Krish retweetledi

Halftime: Dynamically weaves AI-generated ads into the scenes you’re watching, so breaks feel like part of the story instead of interruptions.
@krishgarg @yuviecodes @lohanipravin
English

i won the @xai hackathon by making ads for X Videos
introducing Halftime. targeted ad generation using AI that feels like a part of your movies and shows
built with @yuviecodes @lohanipravin
English
Krish retweetledi

AI is coming for your jobs.
Now it’s coming for your hobbies too.
We built Steve, the Cursor for Minecraft.
Steve and his AI agents can hunt, build and mine on command and even collaborate.
built with @lohanipravin
English
Krish retweetledi

won twice at @HackTheNorth and got a @ycombinator interview.
Introducing Tunnel. AI agents for simulated market research.
English



