Blogs
Read the latest updates, guides, and thoughts from the team building AATMA.
Moving to Substack
I am freezing this blog and moving to Substack. The authoring experience is better, and I hope you will follow me there.
Get Working on Your April Fools Eiffel Tower: The Fun and Peril of Neuron Activation
By tweaking a single neuron, a Llama model becomes obsessed with the Eiffel Tower, showing the power and peril of manipulating AI's internal representations.
It's 11:00 PM. Do You Know Where Your AI Agent Is?
Unsupervised AI agents are causing real-world harm, from spamming inboxes to publishing defamatory blog posts, proving we need stronger safety measures.
The One Name LLMs May Fear: Trump Avoidance and AI's Political Tiptoeing
Observations of ChatGPT and Claude show a bizarre, human-like tendency to avoid naming Donald Trump in negative contexts, raising questions about AI's political programming.
The Long Self-Correction: Our Greatest Flaw in Building Safe AI
Before we worry about building safe AI, we need to confront a deeper problem: humans themselves are dangerously flawed and poorly calibrated to oversee superintelligence.
Grok Build CLI vs Claude Code: Which AI Coding Agent Wins?
Grok Build uses 8 parallel subagents while Claude Code relies on deep reasoning with a 1M-token context. I tested both so you don't have to.
KDnuggets Weekly Roundup: Agentic AI Takes Center Stage
This week's KDnuggets roundup covers MCP servers, agentic AI courses, hallucination evaluation with GraphEval, and the top Claude Code alternatives.
10 Books Founders Can Actually Read by the Pool This Summer
Ten summer reading picks for founders who want to upskill by the pool, covering product, sales, fundraising, and the messy realities of running a company.
Why AI Needs a Genie Coefficient
AI agents can do exactly what you ask — and completely miss what you meant. We need a new metric to measure the gap between intention and execution.
Intelligence Is Free. Now What?
As AI inference costs plummet toward zero, the real challenge shifts from raw intelligence to the data systems that agents will live, work, and build within.
Stop Testing AI Agent Skills Against Production APIs
Testing agent skills against real production APIs costs money, mutates live data, and makes results non-deterministic. A local proxy fixes all three.
Why You Should Never Ship an Agent Experience Change Untested
A Microsoft team tested a dozen documentation tweaks meant to guide AI coding agents, and found most obvious fixes did nothing or backfired.