Sandeep Paruchuri

Builder @OpenAI
Past: 0-to-1 products @WhatsApp @Rippling, @Uber, @Microsoft @NorthwesternU

[Sandeep Paruchuri @sp990]

Really enjoyed chatting with @alicenewtonrex recently about one of the most exciting product journeys of my career!

[Alice Newton-Rex @alicenewtonrex]

I chatted to @sp990 about the big bet that was Microsoft’s AI assistant, Cortana. Great insights in here about how to make a bet happen in a larger organisation, and the importance of scaling your culture as well as your product. https://t.co/DGm4yS53yr

[OpenAI @OpenAI]

Building apps has never been easier.

With Sites, Codex can turn your work, ideas, and plans into an interactive website or app your team can explore, use, and share with a URL.

Rolling out to Business and Enterprise plans, before expanding more broadly.

[zooko🛡🦓🦓🦓 ⓩ @zooko]

Everything Durov says about the bullshit WhatsApp lawsuit, about France, about "not one byte"—it is all just flim-flam to distract you from the important thing: *end-to-end-encryption*. If you use Signal, WhatsApp, or iMessage, your messages are secure. Not so on Telegram. 🔚

[Matthew Keys @MatthewKeysLive]

Just one day after ending "The Late Show" on CBS, Stephen Colbert returned to TV — to host a public access show with rocker Jack White in Monroe, Michigan.

Appearances by Jeff Daniels, Eminem and Steve Buscemi.

[Max Spero @max_spero_]

legendary pull on facebook marketplace

[John LeFevre @JohnLeFevre]

5 gallons of water =
→ $132 from data centers
→ 2¢ from almonds

One powers AI, auto safety, space exploration, medical research, banking, and innovation.

The other is a mostly exported (75%) snack from drought-prone California fields.

Resource allocation in one chart.

[Sandeep Paruchuri @sp990]

Global trend America needs to copy: lactose free dairy milk option at cafes

[Sandeep Paruchuri @sp990]

@MagNorris @WhatsApp Set up the extra phones as companions and you’ll get effectively that https://t.co/9TlrjSQkFk

[Sandeep Paruchuri @sp990]

@SawyerMerritt Sad to see Audi go from building some of the finest interiors of the early 2010s to their plasticky mess of the 2020s

[Sandeep Paruchuri @sp990]

@vueling I have a booking w paid checked bag and want to increase my weight for a bag (give you more money) but don’t have that option online, can you help?

[Sandeep Paruchuri @sp990]

Wild that in suburban chicago, it's cheaper and faster speed to be on Starlink than the AT&T home internet offering, despite their starting fiber rollout here as early as 2005

[Charlotte Lee @cljack]

no professional economist has ever been or will ever be again this funny on purpose

[Sandeep Paruchuri @sp990]

Somehow, not the onion

[OpenAI @OpenAI]

We’re talking about Goblins.
https://t.co/dqmcLGCW71

[Sandeep Paruchuri @sp990]

Love me the ORD penalty box. Tonight’s tally- 60 mins+ from touchdown to gate

[Sandeep Paruchuri @sp990]

@aribindi thankfully there actually is some pretty good food in the Bay that looks like what we ate growing up, but definitely not the norm!

[Sandeep Paruchuri @sp990]

@pitdesi Agree but moving from 100% of the time to 99.9999% of the time

Recently experienced a global hotel set our rate in USD, and then used a marked up exchange rate (19 MXN to USD, when market was 17-18) to then overcharge vs if we just paid USD direct

So read the fine print!

[Sandeep Paruchuri @sp990]

@thijsniks surely you mean, saved by the near-free distribution model with GApps / enterprise contracting

It's the Teams of 2026

[Sandeep Paruchuri @sp990]

@mengxi Agree- seeing some early examples like this that I think will become more prevalent as AI costs come into focus https://t.co/YMKdNNzegP

[Artificial Analysis @ArtificialAnlys]

GPT-5.5 (xhigh) uses ~40% fewer output tokens to run our Index than its predecessor

[Lisan al Gaib @scaling01]

The GPT-5.5 model family completely dominates the cost-performance frontier on the Artificial Analysis Index

[Sandeep Paruchuri @sp990]

Been loving this model during testing and excited to see it in the world

If you haven't tried Codex or GPT in a while (or if you're one of my many friends getting caped on Claude Code) you should definitely take another look!

[OpenAI @OpenAI]

GPT-5.5 delivers this step up in intelligence without compromising on speed.

GPT-5.5 matches GPT-5.4 per-token latency in real-world serving, while performing better across nearly every evaluation we measured.

It also uses significantly fewer tokens to complete the same Codex tasks, making it more efficient as well as more capable.