Ayush Khandelwal (@ayushkhd) header
Ayush Khandelwal (@ayushkhd) profile picture

Build Brilliance. Cyber MTS @Factory. Prev: AI Security @Google. Core Organizer, @MIT Imagination in Action. Tweets are personal shitposts

San Francisco, CA
Joined September 2015
Photo by Ayush Khandelwal (@ayushkhd)Photo by Ayush Khandelwal (@ayushkhd)Photo by Ayush Khandelwal (@ayushkhd)Photo by Ayush Khandelwal (@ayushkhd)Photo by Ayush Khandelwal (@ayushkhd)Photo by Ayush Khandelwal (@ayushkhd)Photo by Ayush Khandelwal (@ayushkhd)Photo by Ayush Khandelwal (@ayushkhd)Photo by Ayush Khandelwal (@ayushkhd)Photo by Ayush Khandelwal (@ayushkhd)
At the @AIatMeta Hackathon this weekend, we built one of the first apps built for the Meta Raybans. Introducing SixthSense, a multimodal agent that augments physical knowledge consumption. Using just your Meta Raybans, SixthSense changes the way you consume books/papers, allowing you to - - Converse with the author, trained on most recent updates in the field - Use gestures or highlighters to digitally bookmark pieces of content you find interesting - Get explanations and Q&As to make reading more interactive
25
41
14
667
220,424
as a developer first and foremost, this is what I imagine my Droids are doin under the hood
the builder (2026) a short film. no crew.
1
14
454
We're building the best in class Security @Factory, and on 11/18 at 10am PST I'll be sharing how we're doing Vulnerability Management. CVEs are being discovered at breakneck speeds, and the only way to stay ahead is to have Continuous Remediation Workstreams patching vulnerabilities for you. Join me tomorrow as I discuss how Factory does attack path analysis, patch hardening, regression monitoring, and deploys fixes autonomously to keep codebases secure, and how you can use it too. Link in the comments
Your scanners find security vulnerabilities faster than your team can fix them. Learn how Factory finds affected repos, upgrades dependencies, fixes what breaks, and opens PRs for your team to review. Register here: luma.com/s94ghvp9
2
1
24
836
girlfriend wanted to talk about me not giving her enough time and i instinctively opened granola. we both have action items now smh
3
23
2,673
Google AI Overviews the most used AI product in the world. Shaping that experience comes with an enormous responsibility to earn and maintain trust. @rajanpatel VP of Engineering for Google Search, and I had a phenomenal chat on Engineering the Future of Google Search at Imagination in Action's Silicon Valley AI Summit. Having worked on AI Security in Rajan's org, it was especially meaningful to sit down together and discuss where Search is headed. We covered a lot of ground - from how Search is leveraging decades of quality signals for retrieval and reasoning to defending personalized search against prompt injections & RAG poisoning. Here's the recording and the topics we discussed: • Introduction and Rajan’s work on Search and Google Lens (0:10–1:12) • What people’s first experience with AI in Search should feel like (1:12–2:55) • How AI Mode is changing the way people ask questions (2:55–3:28) • Understanding user intent and choosing the right response format (3:28–5:04) • From keywords to Google Lens to agents that take action (5:04–8:12) • Personal intelligence, user control, and avoiding filter bubbles (8:12–10:36) • Measuring whether personalized responses are actually helpful (10:36–13:46) • Indirect prompt injection, RAG poisoning, and trust in retrieved content (13:46–16:30) • Evaluating quality and authority in an AI-generated web (16:30–18:56) • Finding trustworthy human perspectives on forums and social platforms (18:56–21:09) • Google AI Mode versus Gemini (21:09–23:08) • AI agents as Search users - and agents built into Search (23:08–24:30) • The questions people still don’t think to ask Search (24:30–26:17) Link to the full discussion in the comments below.
Google DeepMind researcher @melvinjohnsonp says Google's AI Overviews is the most-used AI model in the world by a massive margin: "Funnily enough, I tell my researchers that work on the AI Overviews, 'Hey, you don’t realize you’re building the most used AI model in the world by a big margin.'" "We talk a lot about the frontier. But only a small sliver of people have access to the frontier. If you're distributing it to seven billion plus people, that's a massive deal." @GoogleDeepMind
4
4
22
1,203
your move @factory
JUST IN: Logitech gives all 7,000 employees a day off to play GTA 6 on release day
9
112
11,723
First it was @factory WarGames - have mythos class models break your infra. Now it's Continuous Remediation. The only way to stay ahead of the Vulnpocalypse is leveraging the Security Factory. The team has been absolutely cooking to get these incredible offerings out in record time.
Introducing Security Factory: a new way for engineering teams to turn security findings into tested fixes by connecting investigation, remediation, and verification in one continuous workflow. Now in Private Preview.
4
1
35
2,020
The best harness x mythos class adversarial models
Cyberattacks now outpace the defenses of even well-resourced organizations. Factory can help. Introducing Factory Wargames: We will simulate a nation-state level attack on your company's infrastructure and cover the cost. Apply below and if selected, we'll run the simulation, share our learnings, and Droids will help with the fix for free.
2
15
744
If you're in Vegas for Black Hat, DEF CON, BSidesLV or the AI Security Forum, I curated a master list of 300+ side events happening this week. Filter by day, venue, or interests to sign up directly. Running an event that's missing? Dm me ayushkhd.com/hacker-summer-c…
1
3
416
Incredible turnout at the YC Startup School x Gemini Event! with @bonniesjli @OfficialLoganK @DynamicWebPaige – at San Francisco, CA
4
38
3,013
Helping a couple friends in AI Security Research & Eng at Google & GDM find their next opportunity - anyone in my network hiring? They don't need sponsorship. Absolutely cracked lads
4
2
956
Detached Studio room opening at Luminance in Hayes Valley/Alamo Square. Move in date - August 1st. 8+ month lease. Phenomenal crew, unreal house at a great price.
🤖 Made with AI
Replying to @ayushkhd
We're passionate about building impactful things and longevity. We cook together, throw parties and dinners, and are looking for someone who'd be able to add to the community. Please reach out if you're an organized and clean roommate. Lease is 6 months+ and $2.7k/month.
4
2
6
2,688
Andrej Karpathy just recommended this book written by one of my favorite Professors & author David Eagleman at Stanford Highly recc, great albeit dense read
Anthropic Engineer Andrej Karpathy: "The biggest mistake in AI right now - people are forcing agents to work instead of mastering the model first We made that mistake in 2016 at OpenAI - It cost us 5 years " what Karpathy actually means: step 1 → stop forcing your agent to do everything, understand the model underneath first step 2 → demos are easy - products take a decade. self-driving proved it - if you skip the foundation, everything breaks step 3 → the agent is not the product. the foundation is. build that - and agents emerge on their own "you building agents right now - you're at the forefront. not OpenAI. not DeepMind. you " watch - bookmark, then read article below ↓
1
5
21
3,293
This is one of the best produced conferences I've been to in a while. @aiDotEngineer World's Fair feels like a collection of high quality hyper engineers distilling technical wisdom straight to your brain vs generic corporate larping. A balanced technical platform like this is quite good for the ecosystem. Great job @swyx and team – at San Francisco, CA
1
16
3,105
This validation is going to carry me through the entire week 🫡
I have never seen an episode of The West Wing. Can someone tell me if this is good or not?
1
5
1,233
We got a opening soon for a really nice room in Luminance. If you're interested send a DM
Who's cool & looking for a room in Hayes Valley? > Great location, fun housemates, gorgeous house, unbeatable price. > DMs are open, referrals welcome. Feb 1st move in, be quick.
4
841