Kabir Sial

investing @emergencecap previously @palantirtech @harvardhbs @ucberkeley

[Kabir Sial @kabirsial]

https://t.co/x6rDCSxHIh

[Kabir Sial @kabirsial]

Semianalysis preemptively rescuing markets 🚀🚀🚀

[SemiAnalysis @SemiAnalysis_]

Similar to the panic over DeepSeek R1, some uneducated people think Kimi K3’s use of linear attention (KDA) is bad for NVIDIA, HBM, DRAM, and networking because it has relatively lower KV-cache requirements. The opposite is true, and we explain why below. 👇️ 1/8🧵

[Kabir Sial @kabirsial]

@lauraham150 Totally good point

[Kabir Sial @kabirsial]

https://t.co/x6rDCSxHIh

[Kabir Sial @kabirsial]

@hamostaf04 Tryna learn from u bro

[Kabir Sial @kabirsial]

@snmishra311 When you’ve got well understood task distribution, a lot of observability, low latency requirements, and (at least today) some research expertise on the team

[Kabir Sial @kabirsial]

Cost per task not cost per token my guy

[Chamath Palihapitiya @chamath]

The math ain’t mathing.

What this shows is that you can now spend $0.50 per 1MM leading edge tokens or spend $56 for those same 1MM tokens.

🤔

[Kabir Sial @kabirsial]

Ok but can we get a US open weight model that distills Kimi K3

[Arena.ai @arena]

Big news: Kimi-K3 by @Kimi_Moonshot is now #1 in the Frontend Code Arena with 1679 pts, surpassing Claude Fable 5.

This is a 17-place jump from Kimi-k2.6 (#18 -> #1).

In Frontend, Kimi-K3 ranked #1 in 6 of 7 domains: Brand & Marketing, Reference-Based Design, Data & Analytics, Consumer Product, Simulations, and Content Creation Tools, landing #2 only in Gaming behind Fable 5.

The full model weights will be released by July 27.

Congrats to the @Kimi_Moonshot team on this major milestone!

[Kabir Sial @kabirsial]

@agarwl_ Thought this one was super interesting and cited in Cursor’s composer 2.5 release https://t.co/CqYwOx5Abt

[Kabir Sial @kabirsial]

@quxiaoyin @thinkymachines @AnthropicAI Love Tinker and bullish on what they’re doing to make post-training accessible but find it hard to believe enterprises will be thinking of post-training before just building agents that work (the latter biases users towards Claude)

[Kabir Sial @kabirsial]

@MaikaThoughts Excited for you to gradient ascent to maxima around the bay!

[Kabir Sial @kabirsial]

Grateful that I get to see these bangers for free on this app

[Nikesh Arora @nikesharora]

So jaiye aap.

[Kabir Sial @kabirsial]

What if X actually became the default clearing house equivalent for private market transactions @nikitabier

[Matt Grimm @mttgrmm]

@ankurnagpal No you didn't!

Be honest Ankur, what did you actually buy?

[Kabir Sial @kabirsial]

Novak gonna help GA get into some deals after seeing Kyle Kuzma and Saquan Barkley’s portfolios

[TBPN @tbpn]

BREAKING: @DjokerNole has joined General Atlantic as a Global Strategic Advisor

[Kabir Sial @kabirsial]

@fluorane @appliedcompute @trajectorylabs @EngramLab I'd guess there are probably only a handful of companies today (like Harvey, Decagon) with evals, rubrics, agent traces already set up + have already tried to squeeze performance through methods that don't touch model weights

[Kabir Sial @kabirsial]

@TBaharav @JHUBME @HopkinsDSAI @JHUCompSci Congrats!

[Kabir Sial @kabirsial]

@svathsasa @SpiralDB @metamorphiclabs @krea_ai Bullish on @SpiralDB

[Kabir Sial @kabirsial]

@ItzSuds @dakshgup @utkbajpai @rishub_nahar @gkunthara Hey the other 2 are Indian-american 😂

[Kabir Sial @kabirsial]

@AnjneyMidha @CS153Systems @LiamFedus @ekindogus @periodiclabs @AnjneyMidha when will this be uploaded to YouTube? Can't wait to watch it!

[Kabir Sial @kabirsial]

@brettberson The second last one specific to VCs should be logos we’re invested in