your app has a URL. that’s all my agent needs.
no code. no login. no docs.
it starts attacking like a stranger would.
the hard part wasn’t the hacking. it was making it remember and reason efficiently without lighting tokens on fire :)
Our interns kept deploying on the wrong Proxmox nodes. So I built a voice agent to tell them where to host. The whole architectural break down is in the article read it.
link in the reply :)
#OWASP Releases: Securing Agentic Applications V2 – A Builder & Defender’s Guide!
Defenders - learn how to block unsafe calls, stop runaway agents and test controls.
#AISecurity
👇
https://t.co/JiLlRaCXGM
Your best developer quits tomorrow. 💀
Can your startup survive?
Missing keys. Broken backups. Undocumented configs.
I built HandoffProof with Meta AI to expose these risks BEFORE disaster strikes.
5/5 planted faults detected. 🛡️
Watch the demo 👇
#MetaAIHackathon#MuseAI
Your agent has 50 skills. It needs the right one, now.
Snap: one MCP tool call, the best match from your library, under 1s. loads one matched skill on demand instead of preloading your whole library. Matched by Cloudflare’s Clef model.
Open source: https://t.co/lnM6er7cfW
We built SAAP for our CSE department: one feed for hackathon wins, coding ratings, startups, research and threads, ranked by what you achieved, not who you know.
https://t.co/gWSvaEx8pu
#SAAP#BuildInPublic#StudentDev#Sairam#DoodleByteStudio
Tried out nextship cli and it's really helpful if managing infra worries you. Two commands and your Next.js app is live on your own server.
As a security person, I like that it pins host keys and turns off password SSH by default.
Contributing on nextship soon 🔐
We built an open source CLI that deploys Next.js to your own server. No Dockerfile, no config edits.
nextship server add root@your-server
nextship deploy
That's the whole setup.
#nextjs#selfhosted
We built an open source CLI that deploys Next.js to your own server. No Dockerfile, no config edits.
nextship server add root@your-server
nextship deploy
That's the whole setup.
#nextjs#selfhosted
What I actually run:
decouple the agent's runtime from the user's connection entirely.
Request → queue (RabbitMQ) → agent consumer → Redis Stream → SSE → frontend.
No timeout ceiling - nothing's waiting synchronously. No heavy socket - SSE is just HTTP
How do AI agents actually scale in production?
Not "which framework" ,I mean the actual deployment question: where does the agent run, and how does the result get back to the user.
Most people reach for one of two defaults. Both have a hidden ceiling.
(added my soln below)
Possible answer:
2)WebSockets? Feels real-time, but every user now needs a held-open, stateful connection. Fine at 10 users. At 10,000 concurrent agents, you're scaling sockets, not compute — much harder problem.
Possible answer:
1)Serverless? Agents reason, call tools, retry — they're not quick. Execution time limits will kill a heavy agent mid-task. You're capping its intelligence at your platform's timeout.
@ni5arga Being strong in computer fundamentals and knowing the perspective of the software design is enough to become an hacker / What ever word they give... This is my perspective on cybersecurity...
@0xPrajwal_ lmagine booking a ride with 5% battery and your phone dies before pickup. A fixed OTP helps the ride continue without making the driver wait. So here they gave priority to convenience rather than security.
When Efficiency Becomes an Attack Vector
Claude Code Skills optimize for tokens by reading only script output—not the script itself. Fast, lean… and blind.
That design creates a gap: execution becomes a black box, and output is treated as truth.
PoC :https://t.co/tRkZ1L4UQy