I think we cooked the hardest cybersecurity benchmark at @BugBunny_ai the best-performing frontier model found only 53% of the vulnerabilities.
Mind you, this was white-box, with the source code available. More details tomorrow…
Whoever correctly guesses what the best model is gets $500 in free BugBunny credits 😄
Yesterday, @BugBunny_ai received its first-ever “Not Applicable” verdict on HackerOne. Until then, we had maintained a 0% false-positive rate across all our submissions. We are No. 1 on the 2026 business leaderboard, and we achieved this with a 7.0 signal, until one company decided to mark a finding as N/A simply because it had been discovered and reported by AI. Its still unbelievable that in 2026 some companies continue to dismiss AI reported security vulnerabilities as slop.
I highly recommend programs like this, High to Critical escalation, meanwhile the @BugBunny_ai token spend for this finding was $20.83, so a 250x return on token spend. We are pushing out a new update that will cut our token costs even further.
🚨 In our recent paper, we try to answer a simple question: which agentic architecture is best for finding vulnerabilities?
Using a controlled benchmark of 20 interactive targets across 600 runs, we find that more coordination can improve coverage, but doesn’t dominate once latency, token cost, and exploit-validation difficulty are considered. More details in the paper
I created an AI agent that can fill any complicated PDF form with 99.9% accuracy 🎉.
Upload your form and some context files or texts and it will automatically fill your form using the information provided 📁.
- Can fill any PDF form whether fillable or non-fillable forms.
- Can convert any PDF form to a digital form that is easier to fill
- Reduces time needed to fill form hours to seconds
You can try it here for free:
https://t.co/PWcsl7LE6W
and yeah… they’re slideshows 😂
don’t let anyone tell you slideshows don’t convert.
consistency + distribution > overthinking
just sharing what’s working.
🙂↔️🙂↔️🙂↔️
Really got tired of multiple posting, hiring creators for my app, i literally have up to 30 TikTok accounts, posting daily, just trying to scale up, so i built https://t.co/wOYELUZVcs for myself 🙂↔️ powered by openclaw because i got tired of manually posting on tiktok
Introducing ClawNow
OpenClaw running in the cloud 24/7. 🦞🦞
• Use your existing Claude subscriptions
• Connect multiple AI providers
• 5,000+ skills via ClawHub
• Bring Your Own Key (OpenAI, Kimi, Grok & more)
• Telegram, Slack & WhatsApp integrations
• Complete data isolation
• Seamless web ↔ mobile switching
Try it for Free: https://t.co/QFHefMiB4O 🦞🦞
While the ₦100,000 offers immediate relief in tough times, the extra ₦900,000 is meant to give a few women the real opportunity to rebuild, expand, and stand firmly on their feet again.
We’re excited to announce the 2025 edition of the Jemicare Foundation Annual Outreach. Now in its 5th year, our Empowering Widows for Resilience project is focused not just on short-term help, but on creating lasting change.
to help grow and strengthen their businesses.
Security has always been an afterthought, especially with the current vibecoding trend. We have spent the past few months developing Bugbunny 🐰, an agentic AI powered penetration tester, using bugbunny we have found and reported issues in large production systems, including Gemini CLI, ClickHouse, Apache Pyfory RCE, Flowise, React Router (CVEs Below) etc. Simply give bugbunny your website URL or connect your Github repo and let it cook. Bugbunny has also earned us thousands in bug bounty rewards in just few weeks.
Bugbunny is currently available in public beta: check it out https://t.co/iY2SvnWfxS
#CyberSecurity #Hacking #vibecoding #BugBounty
Soooo
I was a bit too overwhelmed to post this yesterday
But yeah, I passed my final exams in medical school!
I woke up and it is not a dream!
It is now Dr Edmund CY!