GSoC ’26 @DBpedia • AI Engineer @FissionLabs • Agentic AI Architect • Finalist - AWS AI Agent Hackathon '25 • obsessed with LLM Inference & ML x Neuroscience
I just experienced something that I have only read till now. Gpt 5.6 sol on cursor initialized A100 gpus on my GCP account even with no mention of that in the prompt and explicit instructions on what to do. I have no idea how it even got the credentials. Realized 30 min later
Interviewed another AI + Fullstack engineer.
He said he'd built an agentic workflow using multiple LLM agents.
I asked:
"How do the agents communicate with each other?"
He replied:
"They hand of the responses & tool call results based on the workflow rules to other agents."
Then I questioned:
"If Agent A hallucinates early in the workflow, how do you stop every downstream agent from confidently building on that wrong information?"
What would you answer?
Anthropic engineers just showed how they build a full app from scratch, using a loop of agents
40 minutes from the team behind Claude Code
they used three agents: one to plan, one to build, one to judge, cycling until the app actually works
the winners won't have the smartest model, they'll have the best loop
watch it, then read the full guide on how to actually use loops below
@ndrewpignanelli The guide was genuinely needed, though it was gonna be those generic ones everyone posts online but its actually packed with information. Thank you for this. Also love your websites UI
I'm 22, AIML student ,
Just started down the mechanistic interpretability path. The fact that nobody fully understands why LLMs work internally is one of the wildest things in science.
Working on something in this space and will share more soon
Connect if you're into:
Mech Interp & AI Safety
LLM internals
Neuro × ML
AI Research
Agentic AI
This field is early and I'm going all in!