Hacker News: mmmehulll

New comment by mmmehulll in "Muxcard, a dyi credit card size computer"

mmmehulll — Tue, 02 Jun 2026 11:15:03 +0000

this is really cool. I didn't know we had these

New comment by mmmehulll in "Muxcard, a dyi credit card size computer"

mmmehulll — Tue, 02 Jun 2026 11:13:54 +0000

love this. would be cool if we can see and perform all kinds of banking txns on this. Think ledger but all in one card. Super cool. Even cooler would be card to card money transfer without use of swipe machines

New comment by mmmehulll in "Best AI coding plan alternative to Claude and ChatGPT"

mmmehulll — Tue, 12 May 2026 06:12:57 +0000

why not cursor? cursor seems pretty decent at what it does and i have had very few issues since i started using it about a couple of months back. But yes, frontend burns through tokens, else no issues with backend and devops

New comment by mmmehulll in "Using AI for just 10 minutes might make you lazy and dumb"

mmmehulll — Mon, 11 May 2026 12:51:36 +0000

I agree. A month back, I was brainstorming ideas with AI a lot, and it took me a while to figure this out: I was essentially outsourcing my thinking. Honestly, I felt extremely dumb after the sessions and kept blaming myself for becoming dumb, but I stopped using AI, went back to my non-AI routine, and ideas started popping back up.

New comment by mmmehulll in "LLMs gave the overconfident colleague an unlimited ceiling"

mmmehulll — Mon, 16 Mar 2026 08:07:22 +0000

What makes LLM bias reinforcement worse than a regular echo chamber is the confidence. A friend who agrees with you might say "yeah maybe", an LLM constructs a well-reasoned, authoritative-sounding case for whatever you've implied you believe.

New comment by mmmehulll in "Custom programming languages make agents good"

mmmehulll — Thu, 12 Mar 2026 16:02:37 +0000

That’s really interesting. The “one or two examples + good error messages” part feels especially important. It suggests the limiting factor may be less finetuning and more whether the model is given a tight representation and a feedback loop it can recover from.

New comment by mmmehulll in "Custom programming languages make agents good"

mmmehulll — Thu, 12 Mar 2026 15:43:57 +0000

This is really interesting, I feel if llms can respond at existing without finetuning, it can be huge.

Discussion: Shared state breaks most distributed multi-agent systems

mmmehulll — Mon, 12 Jan 2026 12:23:29 +0000

I have been building multi agent systems and the hardest part in practice has been shared state once agents are split across processes or services. Retries overwrite things, handoffs get messy, and crashes make recovery painful.

State inside a single workflow or graph works well. Outside of that boundary, I have not found many patterns that hold up in production.

For context, I built a small open source, self hosted experiment to explore this idea. Repo: https://github.com/MehulG/memX

I am curious how others are handling this today. Are people centralizing state, rebuilding from logs, or avoiding shared state altogether? What has actually worked for you?

If anyone is actively running multi agent workflows and is open to trying this on a single flow, I am happy to help with setup and learn from the experience.

Comments URL: https://news.ycombinator.com/item?id=46587478

Points: 1

# Comments: 0

Show HN: Ctrl – execution control plane for AI agents

mmmehulll — Tue, 06 Jan 2026 16:43:16 +0000

Agents are starting to take real actions (publish content, trigger workflows, touch prod systems). Ctrl is an execution control plane that sits between agent intent and real-world actions. It intercepts MCP tool calls, scores risk, enforces allow/deny/approve policies, and executes only what’s authorized with a full SQLite audit ledger. It Ships as a drop-in CtrlMCP wrapper for LangChain. Looking for feedback and early pilots running agents in production.

Comments URL: https://news.ycombinator.com/item?id=46514647

Points: 2

# Comments: 0

Show HN: MemX – Shared memory for LLM agents

mmmehulll — Fri, 27 Jun 2025 06:47:32 +0000

Hey everyone — I built this and wanted to share as its free to use and might help some of you:

https://mem-x.vercel.app

GH: https://github.com/MehulG/memX

memX is a shared memory layer for LLM agents — kind of like Redis, but with real-time sync, pub/sub, schema validation, and access control.

Instead of having agents pass messages or follow a fixed pipeline, they just read and write to shared memory keys. It’s like a collaborative whiteboard where agents evolve context together.

Key features:

Real-time pub/sub

Per-key JSON schema validation

API key-based ACLs

Python SDK

Would love to hear how folks here are managing shared state or context across autonomous agents.

Comments URL: https://news.ycombinator.com/item?id=44394283

Points: 1

# Comments: 0

Show HN: MemX – Shared memory for LLM agents

mmmehulll — Wed, 18 Jun 2025 11:54:17 +0000

Built memX this weekend; a shared memory layer for coordinating LLM agents.

Agents collaborate by reading/writing real-time context; no chat, no orchestrators.

Includes pub/sub, schema validation, access control, and real LLMs via Mistral.

Comments URL: https://news.ycombinator.com/item?id=44309073

Points: 4

# Comments: 0