visha1v

96 karmajoined November 14, 2025On HN

CTO at Dehurdle. IIT Delhi. Building at the intersection of physics and AI systems. github.com/vishal-dehurdle | vishalvermalabs.com

Recent submissions

The Tokens That Made Me Rethink Everything cacm.acm.org
2visha1v6ddiscuss
Leverage Code Review for Sustainable AI Coding Development cacm.acm.org
1visha1v6d1 comment
Cash or Comfort? How LLMs Value Your Inconvenience cacm.acm.org
2visha1v6ddiscuss
Show HN: Dehurdle – Real-Time AI Voice Simulations for High-Stakes Conversations dehurdle.com
1visha1v9ddiscuss
Human Control at Machine Speed cacm.acm.org
2visha1v12ddiscuss
A Collectivist, Economic Vision for AI cacm.acm.org
1visha1v14ddiscuss
The Trade-Offs of Fighting Child Abuse Online cacm.acm.org
3visha1v14ddiscuss
Show HN: HyperSAE – Sparse Autoencoders, reimagined in hyperbolic space github.com/vishal-dehurdle
6visha1v23ddiscuss
The Sweeter Lesson cacm.acm.org
2visha1v25ddiscuss
Show HN: See what an LLM learned, organised as a browsable tree github.com/vishal-dehurdle
1visha1v25ddiscuss
Show HN: LLMs think in hierarchies. Their interpretability tools should too github.com/vishal-dehurdle
3visha1v28ddiscuss
Show HN: At 16K features, flat autoencoders break. Curved space doesn't github.com/vishal-dehurdle
10visha1v30ddiscuss
The Day We Almost Let a Bot Lose Us a Loyal Customer cacm.acm.org
4visha1v1mo5 comments
Show HN: HyperSAE – Hyperbolic Sparse Autoencoders for LLM Interpretability github.com/vishal-dehurdle
3visha1v1modiscuss
Why Are Women Leaving Computer Science?cacm.acm.org
21visha1v1mo18 comments
A fresh look at CPU silent data corruption in modern data centers cacm.acm.org
52visha1v1mo1 comment
Show HN: Growth-Ratio Energy Function as Leading Indicator of Agent Task Failure vishalvermalabs.com
3visha1v1modiscuss
Toward a Framework for Openness in Foundation Models cacm.acm.org
2visha1v1modiscuss
Datacenters in Space cacm.acm.org
3visha1v1modiscuss
Show HN: Growth-Ratio Energy Function as Leading Indicator of Agent Task Failure github.com/vishal-dehurdle
1visha1v1modiscuss
Topological Control of LLMs: A Route to Trustworthy AI cacm.acm.org
1visha1v1modiscuss
The Dawn of AI Change-Control cacm.acm.org
1visha1v1modiscuss
Growth-Ratio Energy Functions as Leading Indicators of Agent Task Failure vishalvermalabs.com
4visha1v2modiscuss
Capturing token IDs during agentic interaction for better reinforcement learning amazon.science
3visha1v2modiscuss
Comparing Machine Learning Models of Raindrop Formation eos.org
2visha1v2modiscuss
Show HN: A control-theory approach to detecting LLM agent instability github.com/vishal-dehurdle
2visha1v2modiscuss
Why Triso Matters amazon.science
2visha1v2modiscuss
Can We Understand How Large Language Models Reason?cacm.acm.org
3visha1v2mo1 comment
Never ask a model to do something a deterministic system can do reliably cacm.acm.org
3visha1v2mo1 comment
Is Recursive Self-Improvement Here?cacm.acm.org
3visha1v2mo2 comments

All submissions and comments on Hacker News.