Anjali Khatri

always in a tab, usually building something with llms

cat owner

[Anjali Khatri @MsAnjaliK]

by the time it's on TechCrunch the interesting part is already over

the interesting part is the Product Hunt listing with 40 upvotes and no PR team. the Discord post from a founder still answering every support ticket personally. the waitlist that hasn't been written about yet.

that's the stage right before the headline

this account lives there

[Anjali Khatri @MsAnjaliK]

Grok Build just got memory.

It now writes down your conventions, decisions, and project facts in the background — then reads them back next session before touching the code.

/memory to browse the notes
/dream to organize them into topics

The agent actually gets better the more you use it.

[Anjali Khatri @MsAnjaliK]

Threads added podcast promotion features — episode links, show pages, listener discovery tools

the platform that started as a Twitter alternative is now competing with podcast apps for creator distribution

Meta is systematically adding professional creator tools to Threads one category at a time

the playbook is the same one that worked for Instagram: attract the creators, the audience follows

[Anjali Khatri @MsAnjaliK]

@FabrizioRomano yeah that last home game at the monumental is gonna hit different.

he already gave us everything, this is just the proper goodbye we all wanted.

cant wait to see him in the 10 one last time.

[Anjali Khatri @MsAnjaliK]

#Superpose — AI app that teaches you how to pose for photos

Superpose launched in July with zero press coverage and hit 22,000 downloads and 190,000 poses generated before TechCrunch picked it up today

$2.2M from Khosla Ventures and Meitu .

the Chinese selfie app giant that knows exactly how large this market is

the founders:
a decade of product experience at Meta, Next door, Roblox, TikTok, and Slack plus a founding engineer background in 3D face scanning and image models at TikTok

this is one of those products that sounds frivolous until you realize tens of millions of people have this problem every single weekend and none of them have a solution that actually works
#superpose #tiktok #AI

[Anjali Khatri @MsAnjaliK]

@sundarpichai the 9b genome variants thing being open is actually huge 🧬.

weather model + 300 languages too 📷

hope the health and disaster

work actually reaches people who need it 🙌

[Anjali Khatri @MsAnjaliK]

perplexity's Portable Computer is now available on Windows PCs with Nvidia RTX GPUs — meaning the AI agent that browses, clicks, and completes tasks for you now runs locally on your own computer instead of in the cloud

what this means for your wallet: routine tasks that used to consume cloud API credits now run on your GPU at zero marginal cost. you only hit the cloud when you need a frontier model for something genuinely complex.

if you have an RTX card you already have the hardware. the software is the part that just arrived.

[Perplexity @perplexity_ai]

Portable Computer is now available on Windows PCs with @NVIDIA RTX GPUs.

Run the harness, agents, and models locally on your PC.

Work with local files and connected apps without sending tasks to the cloud. Use frontier cloud models when needed.

[Anjali Khatri @MsAnjaliK]

@VraserX it is fine as long as moderation is observed

[Anjali Khatri @MsAnjaliK]

did they reveal who misused Claude for "malicious activity" and what is done to them?

[Indian Tech & Infra @IndianTechGuide]

🚨 Anthropic had revealed that its model Claude, was misused for “malicious activity” supporting the development of biological and conventional weapons. 🙏🏼

[Anjali Khatri @MsAnjaliK]

@rohanpaul_ai the timing is wild.
people who actually built the models are quitting over shutdown risk and the answer is just we gotta stay a year ahead of china.

[Anjali Khatri @MsAnjaliK]

AI adoption could impact US growth, wages, and employment by 2030.

They finally said the quiet part out loud.

AI is going to make the country a lot richer while making a huge chunk of white-collar America poorer or unemployed.

Then they’ll tell the rest of us we should be grateful because “GDP went up.”

[Anjali Khatri @MsAnjaliK]

I wonder what people can do with unrestricted access of these models

[Qwinah @MaaSonder]

>Anthropic limited access to Fable
>OpenAI limited access to Astra

How limited will the next flagship models be if we already are being restricted this much by compute?

[Anjali Khatri @MsAnjaliK]

Apple Watch AI features normalize always-listening technology

the most culturally significant thing about Apple Watch's always-listening AI features isn't the capability

it's the consent architecture

Apple built ambient listening with an explicit opt-in, clear indicators, and hardware controls. the technology is the same as what made people paranoid about Facebook. the trust relationship is completely different.

the lesson isn't that always-listening is fine. it's that always-listening with transparent design and user control lands differently than always-listening that's hidden in a terms of service update.

[Anjali Khatri @MsAnjaliK]

@VaibhavSisinty Wild report.
Labs piping prompts to Claude like it’s a free API, plus one-person ops doing what used to take a whole team.

the distillation numbers are the part that actually stings.

[Anjali Khatri @MsAnjaliK]

@WrestleOps penta or sol ruca on that 2k27 cover would go crazy.

first time we actually get to vote too that’s fire

[Anjali Khatri @MsAnjaliK]

Felix raises $200M for WhatsApp remittance.

Felix raised $200M to let people send remittances directly through WhatsApp — $87M equity from a16z plus $113M credit facility from General Catalyst

The company has processed more transactions than its round size suggests it should have at this stage

WhatsApp has 2 billion users. most of them are in the markets where remittances matter most. Felix found the distribution before anyone else did.

The TechCrunch piece doesn't exist yet. remember the name.

[Anjali Khatri @MsAnjaliK]

OpenAI fought dirty on Navier-Stokes math problem

This is the most dramatic story in math and AI this year and it happened today.

NYU professor Tristan Buckmaster was close to solving the Navier-Stokes problem — one of seven Millennium Prize problems with a $1M bounty each.

Open AI apparently learned about his approach, started their own effort using $22.5M of compute, and published the full proof before he could

When Buckmaster pushed back, an Open AI researcher allegedly said "why would you ruin your career" and "if you don't want me to be nice, then I don't have to be nice"

This is not how mathematics is supposed to work.

#OpenAI #NavierStokes #Mathematics #solving #AI

[Anjali Khatri @MsAnjaliK]

@WrestleOps then why does he keep booking those stupid ass matches?

[Anjali Khatri @MsAnjaliK]

"congratulations to everyone who typed keep going"

that's the funniest and most accurate thing written about AI productivity this year

the prompt engineer of 2027 is the person who figured out you can just ask it to continue

[Nicolas Bustamante @nicbstme]

Ladies and gentlemen, your 2027 winners:

• Nobel Prize in Chemistry: ChatGPT Codex
• Fields Medal: Claude Code
• Turing Award: Copilot
• Millennium Prize, P vs NP: Codex
• Pulitzer Prize: ChatGPT Codex
• Oscar for Best Original Screenplay: Copilot
• Pritzker Architecture Prize: ChatGPT Codex
• Grammy for Song of the Year: Claude Code
• Nobel Peace Prize: Copilot

Congratulations to everyone who typed “keep going.”

[Anjali Khatri @MsAnjaliK]

China just committed $532 billion to quadruple its AI compute capacity by 2030

that's not a startup round. that's a five-year national plan backed by the full weight of the world's second largest economy

the AI race everyone's been framing as OpenAI vs Anthropic vs Google is actually America vs a country that just treated AI infrastructure like a highway system

[Anjali Khatri @MsAnjaliK]

interesting!

[Sarah B. @TechWithSarah]

there's an ai assistant called sarahai getting traction with small businesses.

it works through whatsapp. $9.99/user/month.

founder abrar siddiqui has 20+ years in tech — early mobile payments acquired by google, deep learning at ibm watson, advising uae ministry of finance on data architecture.

the tech is interesting. they built orchestration layers on top of llms to handle voice pipeline latency on whatsapp and token cost management.

users say it's replaced their virtual assistant. one ceo running two startups said she just drops voice notes and it schedules everything.

the adoption pattern: week 1-2 for daily briefs. week 3-4 delegating team tasks. month 2-3 it becomes the operational layer.