3. Google provided Qwiklabs access to registered users ahead of time to get familiar with the tools and tech stack. Having that hands-on practice beforehand meant less time figuring out basics on the day of the hackathon and more time actually building.
Looking back at the 2026 Agents for Impact Hackathon in Sunnyvale, a few honest takeaways stand out:
1. Having a clear goal saved us from overthinking ideas. That focus kept us building instead of getting stuck in analysis paralysis.
#AgentsForImpact#GoogleCloud
2. Google’s tools made it remarkably fast to take rough ideas, turn them into working prototypes, and iterate over them to perfection. It was a very satisfying and productive experience.
Had a wonderful time participating in the Agents for Impact Hackathon - Sunnyvale on Nov 17th, 2026.
To top it all off, we got the best surprise later in the day: "Congratulations, your team has won Phase I of Agents for Impact - Sunnyvale!"
#AI#Hackathon#AgentsForImpact
Seeing tools like REA (Reverse Engineer Anything) plug straight into agentic workflows makes you realize: how easy it is to reverse engineering closed-source logic.
Can agents systematically map, extract, and replicate the core business logic of legacy ERPs and CRMs?
Anthropic says AI agents can hide their tracks to evade detection. CrowdStrike says they can shut rogue agents down.
Are they both telling half-truths?
Not sure agents can delete hypervisor logs or harness can detect all semantic deception before it fires in real world
Yesterday there was news about "US banks have detected unvetted AI agents accessing customer bank accounts".
Isn't this the same as driving your car in FSD mode and getting into an accident? Who is ultimately responsible—the driver or the tech?
You train an LLM few books, it acts dumb. Train it on a few million books and it starts to act smart. Give it zillion books, suddenly it becomes "conscious". Isn't that weird? 🤔
#AI#LLM
Are recent Agent hacks just exposing our existing flimsy corporate systems were security was always a afterthought.
If AI agents are supposed to replace humans, why does replacing humans with agents lead to so many vulnerabilities?
Spent the day at Google DevFest diving into the latest agentic dev stuff in the Google ecosystem.
Honest take? These new agentic graphs modelling looks a whole lot like classic BPM workflows on steroids, with an expensive LLM playing the role of a smart router.
Most enterprise workflows don't actually need an LLM capable of autoregressive generation. They just need JEV for fast, cheap, typed decisions to route events to the right API.
JEV might be all you need for 80% of enterprise automation usecases.
If I wield a hammer, miss a nail, and accidentally dent the wall, it’s not the fault of the hammer. The problem lies in how I used the hammer. Similarly, if I prompt an agent and it hacks into someone else’s system, the responsibility lies with me, not the agent.
In 2008, Walter L. Wagner and Luis Sancho are pursuing a lawsuit in a U.S. federal court to prevent Geneva's Large Hadron Collider from being switched on later this summer.
They're afraid that the new giant particle accelerator could destroy the entire planet.
When humans wrote rogue code to compromise machines, we called it a virus. When humans exploit systems for gain, we call it hacking. When AI does it, we call it existential risk.
https://t.co/6rPRiEvU1e
An interesting paper. Basically it says, this AI train wreck is unstoppable and it has dire consequences.
Because each CEO will be forced to make decisions that are in the best interest of the shareholder (Ref: prisoners dilemma).
#AI