Our autonomous pentesting agent just outperformed the two most popular open source offensive security agents on a benchmark of 60 modern, defense-enabled web apps.
Battle-tested in production against our customers' environments from startups to financial institutions, Apex consistently finds and exploits critical vulnerabilities other agents and humans miss.
Today we're releasing it open source alongside our internal benchmarks.
After 2+ years at Google, I left... without even knowing what I'd be building.
Needless to say, I spent a lot of time brainstorming, thinking about different ideas, pulling from that same, ever-growing list of ideas every entrepreneurially-spirited person has in their notes app.
During that time, I found myself using AI tools more and more. But not only did I use them for coding, but I started to experiment with using AI as a kind of conversational thought partner.
For this, I used the main AI chats (i.e., ChatGPT, Gemini, Claude) to think through my thoughts, get instant feedback, pressure-test my ideas, suggest tweaks and insights that maybe I would not have come across myself.
But I often found myself sending the same query to multiple models and comparing responses. I mean, what if you're not getting the best response? What if by using one model, you're missing out on a better response from another?
I used tools like OpenRouter's chat app to more easily send the same query to multiple models at once. But you can only look at so many diffs.
What if instead of using one model at a time and manually comparing outputs, multiple models could converse with each other? Building off each other's ideas, challenging assumptions, surfacing perspectives a single model might miss?
Add different prompts into the mix and encourage them to explore different angles and perspectives and you can get a better response, along with much richer, more nuanced understanding of the problem along the way.
So I built https://t.co/eZ185yoc3F: a tool that lets you chat with multiple AI personas who debate, challenge, and build on the different perspectives of you and the other AIs.
It's still experimental, but if you're:
- a founder drowning in decisions
- looking for a more critical thinking experience and a counter to the sycophancy of AI models
- or just curious about multi-agent AI conversations
I'm looking for feedback, and the more brutal the better.
Try it now at https://t.co/eZ185yoc3F!
WATCH: Introducing What’s Poppin’..
A brand new series giving fans a closer look at their favorite #Hawks players.
First up is Dyson Daniels, the 22-year-old Australian who continues to rise in the DPOY race. See why he might just win it👀 @FOX5Sports
If Miami beats Dallas today I’ll give one person this Luka jersey I personally got signed. All you gotta do is Like and RT this tweet. You must be following so I can DM you. #GoHeat
Fuck it. In honor of yesterday’s scuffle. I’m giving away a personal pair of my JB1s. Can’t get these no where in the states. All you gotta do is RT. Like. And follow. I’ll pick someone during the Sacramento game on Monday.