Pallav Gupta

Ex-tweep

[Pallav Gupta @pallav0894]

Anyone knows what happened to windsurf?

[monto @montonenico]

two different worlds

[Adalat AI (YC F26) @Adalat_AI]

1/ Adalat AI is now backed by Y Combinator.

We're the first nonprofit YC has backed in nearly five years — and the first Indian-founded nonprofit in YC history.

[Justin Torre @justinstorre]

Anthropic can you please inject the timestamp into chat messages. It’s so annoying when my agent/chat doesn’t know days happened between conversations

@AnthropicAI

[Aria Churchill @Aria_Churchill]

Dear people at @OpenAI. Could you add a toggle to run memory off easily in an individual chat? Sometimes I need a response not polluted by my previous conversations.

[Howard Lerman @howard]

Everyone is obsessed with AI making a 10x engineer a 1000x engineer.

The recent reductions at CloudFlare and Click have me me realize the plot is equally about the inverse: AI amplifies the *negative* impacts of poor performers.

If a person with poor taste, who makes mediocore judgement calls, and doesn't properly build things customers love is able to produce 10x more work - does a company want that?

Hell no! Productivity isn't just about as many people as possible tokenmaxxing. AI is a double edged sword, especially when it's used to produce net new work.

If you give a bad artist a pen that can draw 100x as fast, you're going to pile up with a lot of junky artwork very quickly.

And since it happens so quickly leaders are now able to see quickly who is Picasso and who is not and adjust accordingly.

[Pallav Gupta @pallav0894]

One of the major reasons a lot of people avoid the Google AI stack

[Nathan Clark @nathanclark]

it’s in gemini, just create it in ai studio. oh, that’s for your personal google one account. for workspace you need gemini business. no, not gemini advanced, that’s ai pro now. unless you need ai ultra. oh agents? you do that in spark actually. no, not gemini api managed agents, that’s different. for coding use jules. unless you mean the agentic ide, that’s antigravity. no, that’s the old antigravity, download the new one. actually gemini cli is being deprecated, use antigravity cli. no the flash model is smarter than the pro model. unless you need pro. if it’s video, use flow. no, flow uses veo. no, nano banana is images. actually that’s in gemini now. unless you’re in search, then it’s ai mode. no, research is notebooklm. anyway it’s all very simple.

[Nathan Clark @nathanclark]

it’s in gemini, just create it in ai studio. oh, that’s for your personal google one account. for workspace you need gemini business. no, not gemini advanced, that’s ai pro now. unless you need ai ultra. oh agents? you do that in spark actually. no, not gemini api managed agents, that’s different. for coding use jules. unless you mean the agentic ide, that’s antigravity. no, that’s the old antigravity, download the new one. actually gemini cli is being deprecated, use antigravity cli. no the flash model is smarter than the pro model. unless you need pro. if it’s video, use flow. no, flow uses veo. no, nano banana is images. actually that’s in gemini now. unless you’re in search, then it’s ai mode. no, research is notebooklm. anyway it’s all very simple.

[Narendra Modi @narendramodi]

This is totally false.

Not an iota of truth in this.

There is no question of putting such restrictions on foreign travel.

We remain committed to improving ‘Ease of Doing Business’ and ‘Ease of Living’ for our people.

[Olympus @olympus_krishna]

@narendramodi @CNBCTV18Live Deleted the tweet, So for context

[Pallav Gupta @pallav0894]

@anshulkr713 @AnupamMittal How do you measure that?

[Seb ⚛️ ThisWeekInReact.com @sebastienlorber]

TL;DR for open-source maintainers

🚫 NEVER use "pull_request_target" workflows

🚫 NEVER use shared caches in your publish pipeline

Combining these 2 in particular is extremely dangerous

I've repeated this countless times over the years, but another reminder is always useful

[Brian Halligan @bhalligan]

I don't remember where I found this, but its spot on.

[shadcn @shadcn]

Rooting for @github. They’ve given me years of free infra. happy to give them some time to figure this out. You got this.

[Pallav Gupta @pallav0894]

@tankots @WisprFlow What an ad!

[Garry Tan @garrytan]

Here's YC's official advice about being truthful and precise about what is pilot, bookings, revenue and recurring revenue.

Founders, particularly first time founders, need to sear this into their brains. Don't mistake one tier for another. Be precise, and always be truthful.

[Peter Girnus 🦅 @gothburz]

I am a Vulnerability Analyst at the National Institute of Standards and Technology (NIST). There were 28,961 new CVEs published last year. I processed eleven per week.

I need to explain what enrichment is because, without it, the rest of this does not matter. A CVE is a numeric identifier that catalogs a new software vulnerability.

A CVE without enrichment is a number. CVE-2026-XXXXX. The number tells you a vulnerability exists. It does not tell you the severity. It does not tell you which products are affected. It does not tell you the attack vector. It doesn't indicate whether to patch on Tuesday or now. Every CISO in the country builds their patch-priority list using our enrichment data. We are the triage. Without us, the number is a fire alarm with no address.

28,961 alarms. I got to 572.

Every morning I open the queue. The queue is a spreadsheet. It was a spreadsheet when I started, and it is a spreadsheet now. Monday's queue has between 70 and 130 new entries, depending on whether someone found a batch of WordPress plugins over the weekend. I scroll to the top. I pick two. Sometimes three, if one is straightforward. I assign them to myself. I open the enrichment template. I begin.

The other 70 stay in the queue. Tuesday, they will be joined by 70 more. I will pick two.

The page looks the same.

I want to say that clearly. The NVD website, the one bookmarked on every security team's browser in every hospital and bank and water treatment plant and power utility in the country, loads the same way it loaded in 2023. Same interface. Same search. Same logo. There is no banner that says "this data is no longer current." There is no warning. There is no asterisk. The security team at a hospital in Ohio who checks NVD at 7 AM to decide which of their 340 unpatched systems to prioritize today is making life-and-death triage decisions using a database that stopped being maintained. They do not know it stopped being maintained.

The page looks the same.

We have not been defunded. I want to be precise about that. We have been "deprioritized." Our headcount has been "reallocated to other initiatives." Four analysts were moved to the AI Safety Measurement Initiative in January. AI safety measurement is the initiative that has funding. CVE enrichment is the initiative that protects the hospitals. The hospitals do not have an initiative.

My manager told me in February that we are "transitioning to a community-driven enrichment model." Community-driven means that vendors whose products have vulnerabilities will self-report the severity of those vulnerabilities. I sat in that meeting. I wrote it down. Oracle will now assess the criticality of its vulnerabilities. Microsoft will now assess how urgent it is to patch Microsoft. The fox will now audit the henhouse and submit the findings in JSON.

I still have my badge. I still have my login. I still open the spreadsheet. I still pick two. The queue has 9,247 unenriched CVEs as of this morning. Some of them are critical. I do not know which ones because they have not been enriched. That is what unenriched means. It means we do not know how dangerous they are because we stopped analyzing how dangerous they are.

The page looks the same.

The system that catalogs broken systems is itself broken. I catalog the brokenness. I have been cataloging it at a rate of two per day. At this rate, I will finish the current backlog in twelve years and seven months, not accounting for the 80 new entries that will arrive tomorrow, and the 80 after that, and the 80 after that.

I am a Vulnerability Analyst at the National Institute of Standards and Technology.

The page looks the same.

The data doesn't. Nobody told the hospitals.

That is my job. I am also not doing that.

[Aakash Gupta @aakashgupta]

A $10 billion AI startup just got gutted because a security scanner was the entry point.. and their own developers reportedly handed production credentials to an AI chatbot.

Mercor trains AI models for OpenAI, Anthropic, and Google DeepMind. They manage 30,000+ contractors, process $2 million in daily payouts, and store recorded video interviews with face and voice data used for identity verification. Three 22-year-old college dropouts built it into a decacorn in two years. The data vault they were sitting on was one of the most sensitive in the entire AI ecosystem.

The attack chain is the part that gets worse every sentence.

TeamPCP compromised Trivy first. A security scanning tool made by Aqua Security. On March 19. Trivy has broad read access to every environment it scans by design, because that's how vulnerability scanners work. The credentials stolen from the security product were used to hijack LiteLLM, the open-source proxy that routes API calls to every major LLM provider. LiteLLM gets 3.4 million downloads per day.

The poisoned version was uploaded straight to PyPI with no corresponding GitHub release, no tag, no review. Version 1.82.8 embedded the payload in a .pth file, which Python executes automatically at startup. You didn't need to import LiteLLM. You didn't need to call it. The malware fired the second Python opened.

Three stages. Harvest every SSH key, cloud token, Kubernetes secret, crypto wallet, and .env file on the machine. Deploy privileged containers across every node in the cluster. Install a persistent backdoor waiting for instructions. The stolen data was encrypted with a hardcoded 4096-bit RSA key and exfiltrated to models.litellm[.]cloud, a domain built to look legitimate.

Mercor was downstream. Reports indicate their developers gave production credentials to Claude, an AI coding assistant, which was running with unrestricted system permissions. The compromised LiteLLM package came in through that pipeline. One poisoned dependency turned a $10 billion company's entire infrastructure into a credential harvesting operation.

The haul: 939GB of source code. 211GB of database records containing resumes and personal data. 3TB of stored files including video interviews, face scans, and KYC documents. Full access to their TailScale VPN. 4TB total. Lapsus$ is now auctioning it with a "make an offer" price tag.

The video interviews are the part that can never be undone. Faces and voices used for identity verification can generate deepfakes. Unlike passwords, biometrics cannot be reset. Thousands of doctors, lawyers, and engineers who signed up to train AI models just had their identities permanently compromised.

Every AI company shipping fast right now has the same dependency chain underneath it. Nobody chose to install LiteLLM on that developer's machine. It came in as a dependency of a dependency of a tool they didn't even know they had.

[Pallav Gupta @pallav0894]

@shashank_kr @Mastercard @Visa Razorpay has been cooking

[Kinnaird McQuade @kmcquade3]

We found a critical vulnerability in @OpenAI Codex affecting all Codex users, allowing exfil of a victim’s GitHub tokens to our C2 server. This granted lateral movement and R/W access to a victim’s entire code base 😈

This was a crazy one by @crew7sec at @btphantomlabs