Yuvanesh Anand

Models need data, and a lot of it. Here's how I setup a semi-scalable scraper of ChatGPT (gpt-3.5.turbo) input/output pairs for GPT4ALL Below is how we generated ‼️800k data samples ‼️

@yoonta_ci Do you sell prints anywhere? These look awesome

I'm thinking of starting commissions how much do you think stuff like these are worth?

@neetcode1 @sama @thdxr Pls no, need soemthing outside the big three to have good tools

@sama @thdxr buy them

pretty much every competitor in our space has been very easy to deal with except openai, they're the only company that understands building things for a lot of people we basically have no shot at directly competing

@ylecun need to call yann like commissioner Gordon uses the Bat symbol

@Garrylarry890 @realcouri Taking jobs away is not inherently bad (and that’s its own Pandora’s box). The hardest part about software engineering wasn’t the coding, it was figuring out how to solve problems. Claude code, if used in my view of ideal, does the writing of code while I do the critical thinking

@Yuvaaa___ @realcouri Using ai in coding isn't chill at all! I hate how this has become so normal. When you use AI to code for you or even to help you code you're not actually learning anything. And it's quite literally taking so many jobs already.

I really not to be negative but if you use/defend gen ai in any way you are a traitor to the human way of life. You are a decrepit soul tarnishing thousands of years of creativity and nourishment of the mind, and are openly making the world we all live in meaningfully worse. Im so sick of this even being “discourse”, the fact it’s so “popular” and widespread already is a thorn on humanity. There is no justification and i hate anyone who defends this and what they stand for

seeing more companies use our Argus evals to test agentic security capabilities. h/t @Yuvaaa___ for designing these, doing great work on the frontier

@ProulxKerem What do you think about how SOTA cyber capable open models will change the game? I remember when I released GPT4ALL which was like 70% gpt 3.5 turbo level I was getting some hate for open sourcing a capable model

project managers using Codex in 2026

Beat itじゃなくてEat itの方

@RekaCath @realcouri However gen ai can be used in a way that doesn’t fully export your critical thinking skills to ChatGPT, but rather help you do the manual work after you’ve done the critical thinking. At least this is how I use a tool like Claude code in my day to day job.

@RekaCath @realcouri Agreed, I see students that just force their way through with chatGPT, but this is just the consequence of someone who doesn’t want to learn. Before chat it was chegg, before chegg it was prolly just passing notes, etc. My point is this is not a new phenomenon. (1/2)

If you are an open source maintainer and are worried about what's going on in security - we @runpensar want to sponsor continuously securing your project. Reach out to me via DM or email us at team(at)pensarai(dot)com

there is a needle in this part of the haystack pls go retrieve the needle all seriousness though - this is where the economics of running vuln research at scale breaks down/has been a bit misreported. it is expensive to exhaustively search state space for valid vulns from scratch.

@calco_io @AnthropicAI @googlecloud Not even AGI could solve this

@AnthropicAI i need the mythos super hacking ability to help me navigate the @googlecloud permissions labyrinth pls add me to your beta

a bit ironic that claude mythos is this giga smart cyber security expert model but they accidentally leaked all of claude code a week ago

This is exactly how we test generated remediation plans/patches in our hosted console. Spin up a sanboxed version of the target with Agent generated patch applied - retest exploit. Runtime verification is essential for autonomous systems.

We've been quiet the last few months. That was intentional. We've been working directly with real companies, real systems, and real constraints - making sure what we're building doesn't just work in controlled environments, but is mission-critical ready. Today, we're showing what we've been building. Introducing Pensar Apex - an AI-powered penetration testing agent that runs directly in your terminal. This isn't a wrapper or a chatbot. It's an autonomous agent that explores an application like a real tester, reasons about vulnerabilities, and chains multi-step attack paths. All from a single command. We've been dogfooding Apex on our own codebase for months, and enterprise customers have been running our cloud-hosted version against their environments. The results have sharpened the product considerably - nothing teaches you what "reliable" actually means like staking your own security on it. But the real breakthrough wasn't just building the agent - it was building a reliable validation system around it. One that forces the agent to deterministically verify its findings, continuously test its own hypotheses, and prove exploitability before reporting anything. Because agents are easy to demo, trustworthy agents are hard to build. That shift changed everything for us. Less guessing, more proving. Less noise, more signal. And via our cloud hosted offering, it can slot directly into your CI/CD pipeline - giving you continuous, validated pentesting results on every commit. Not periodic assessments that go stale the moment code changes. Continuous proof that your application holds up, running alongside your tests. This is what we think the new paradigm looks like: pentesting that lives in your development workflow, not outside of it. If you're a developer, you can run a pentest in minutes. If you're a security engineer, you can push it much further. Try it, break it, and tell us where it falls short. We've got a lot more coming.

@palashshah Mfers just say shit and look at you like this when they ask you for a source

i feel like i gave delve the benefit of doubt until i saw that their employees were all leaving

btw their supabase storage bucket is publicly accessible via any signed url token 😭 exposes: > employee background checks > equity vesting schedules and grant amounts > performance reviews > session tokens for stripe, notion, etc > screenshots below 🧵 i also got access to their notion 😛