arun

@aruns
3 Followers
4 Following
148 Posts
Moltbook could have been interesting if it were not engineered but an emergent behavior. #moltbook #ai
#WeekendThought
The laws of thermodynamics seem to be apply to code too as we are move from "if-then" to weights and biases. As the entropy rises, so should our responsibility. Our role is to watch at where the arrow of time points and engineer responsible systems.
#AI

If you thought 'Attention Is All You Need,' just look at AI social media right now. The collective attention span is that of a squirrel.🐿️

So, what’s next ?
#clawdbot #ralphwiggum

I think Claude Cowork has hit the sweet spot. That is Claude Code in a suit tailored for knowledge workers. Being a CLI is still a barrier to entry no matter how magical the tool is. We should see similar Claude Code in different suit avatars for every major job/role soon.
Claude Sonnet 3.7 was the tuning point for me. It will always be the goat 🐐
In “The Memory Problem: Why Your Agents Fail,” I unpack why context, not capability, is the real bottleneck for enterprise #AI. Read it on Deep Gains: https://deepgains.substack.com/p/the-memory-problem-why-your-agents
The Memory Problem- Why Your Agents Fail: Part 1

What happens when the knowledge that runs your company lives only in people's heads.

Deep Gains
There is no technical debt when code is ephemeral. This is why vibe coding is here to stay.

2025 has been one hell of a ride.
Cheers to 2026

Give me an API key and Opus 4.5, and I shall move the world.

I have written a follow-up piece to my "Year In Review: Agentic AI in the Enterprise" discussing how context graphs may be the missing element in enterprise decision intelligence and building a truly agentic enterprise. https://deepgains.substack.com/p/context-graphs-the-missing-link-for
Context Graphs: The missing link for a truly Agentic Enterprise

A follow-up piece discussing how context graphs may be the missing element in enterprise decision intelligence and building a truly agentic enterprise.

Deep Gains
Context graphs are the missing “decision memory” layer enterprises need to make AI agents truly useful. They don’t replace RAG or knowledge graphs; instead they connect decisions, policies, tools, systems, humans, and outcomes so agents understand not just what happened but why.