Shivanand Velmurugan

Product Leader (@MSFT / Fabric, Real-Time Data, Schemas). In a past life, I built operating systems, drivers, a DSL, an in-memory database.

[Shivanand Velmurugan @shiva]

@chrismunns +100 since when was it ok to be cheeky about outages or errors?

[Shivanand Velmurugan @shiva]

Analyse files in this repo, and start work. Bootstrap a CLI to manage memory across multiple AI sessions, one that can evolve into a portable semantic and ontological representation of my use of AI across all my projects. Use Zig. Keep it metal! #fable let's go!

[Shivanand Velmurugan @shiva]

Ok! The soundstage got a major upgrade today! After decades of ogling, I’m a proud owner of an Accuphase E-301. Paired it gingerly with Kappa 6.1s. Ran through some Pink Floyd, Bill Evans, Macklemore, Jamiroquai, a whole host of others. Life cannot be better.

[Dean W. Ball @deanwball]

Make no mistake: post-Mythos, the United States has a licensing regime for AI. It’s just informal, with no consistent rules or firm boundaries on state power or public transparency. Cobalt mining in the Congo is vastly more institutionalized than frontier AI licensing in the US.

[Shivanand Velmurugan @shiva]

If you need some hard hitting Rock to star your drive : The Riff List https://t.co/CVJcLvceaN

[Shivanand Velmurugan @shiva]

This could be interesting 360! I wonder what the graph would look like if each of you did that, and shared with each other!

[Pankaj @the2ndfloorguy]

i hooked my whoop to my work calendar to find which coworker gives me the most stress 🚨

thanks to fable, I reverse engineered whoop to pull per minute heart rate. nd matched spikes with cal events and attendees

I now have a leaderboard and I think about it daily.

few info masked for obvious reasons ;)

[Shivanand Velmurugan @shiva]

How much local AI can you squeeze out of a 16GB Mac Mini? Cloud is great, but I wanted to muck with the silicon. I vibe-coded a test harness to benchmark 13 local LLMs. From Llama 3.2's cache speed to the brutal 14B memory swap cliff, here is the data:

https://t.co/SCplrYFSXl

[Shivanand Velmurugan @shiva]

AI that doesn’t know YOUR stuff is annoying. Would you pay $5/mo for one secure vault — your records, books, workouts, notes — that Claude/ChatGPT can read?”

[Shivanand Velmurugan @shiva]

Give this playlist a listen: Mullard Nights https://t.co/6p4lkijtJS

[Shivanand Velmurugan @shiva]

4/Then when you try to uninstall the desktop app, you get this:

[Shivanand Velmurugan @shiva]

I unsubscribed today from ChatGPT (yes, for the reasons you are thinking of). It sure was painful.

A Thread:

[Shivanand Velmurugan @shiva]

3/ They ask you for an exit interview and why you are leaving 3 times (you need to suffer through each of it) (Friction #5)

[Shivanand Velmurugan @shiva]

Hello PMs of Vancouver! Here is an opportunity to join the amazing Eventstream team at Microsoft. If you want to know about what we do in this team, or want to dive deeper into my experience joining this team about a year ago, come chat.

https://t.co/S2xl1zhTdB

[b1ack0wl @b1ack0wl]

DO NOT VIBE CODE A 200,000 LINE SOCKET LIBRARY IN C 😭

[Raul Junco @RaulJuncoV]

Most engineering teams screw up messaging because they don’t understand one thing:

Streams hold truth.
Queues do work.

When you get this wrong, your system bleeds.

I’ve seen $10M mistakes from teams who dump everything into queues:
“Just push the order to a queue and process it!”

Most Queues delete messages after work is done.
No history. No replay. No audit.
Just pain and guesswork.

On the other side, some teams fall in love with Kafka:
“We’ll stream EVERYTHING!”

Here’s the rule I wish someone told me early:

If the event changes the business → Stream
If the message is an action to perform → Queue

Streams = OrderPlaced, PaymentAuthorized, InventoryReserved
These are immutable facts.
They must be durable, replayable, ordered, and auditable.

Queues = SendEmail, CapturePayment, GenerateInvoice
These tasks exist temporarily.
They matter NOW, not 6 months from now.

Event enters the stream → workers derive jobs → queues execute tasks

Ledger first.
Assembly line second.

Everything else is technical debt disguised as cleverness.

[Shivanand Velmurugan @shiva]

memory

[Daniel Park @danifesto]

I’ll share a small part of https://t.co/UQ4RFIKUwO

Back in med school, I became obsessed with augmenting memory and dreamed of a Notion or Obsidian that completes itself. Today, we’ve built something close.

My self-awareness is sharper and everything feels connected. I genuinely believe AI does not replace humans. It amplifies us.

Huge respect to our engineers and designers who made this crazy thing real.

Bubbles are the episodic units of my life that the system interprets from my raw data. Clouds are the system’s questions, its hypotheses about who I am.
When I answer a cloud, it becomes a bubble again.

There is so much personal data that I cannot fully demo it. Wish I could. This system understands me more deeply than anyone.

Want to try it? Retweet and comment “memory.”
I’ll DM you an access code to skip the waitlist.

[Shivanand Velmurugan @shiva]

@mazeincoding I’ll do it with one — meeting

[Shivanand Velmurugan @shiva]

@ThePrimeagen Have you tried implementing a discriminated union? C is just *chef’s kiss* for this. No other language makes it simple or easy to this. I spent 5 years building a DSL for a router OS, and C over C++, Java (go and rust didn’t exist then), Lua, precisely for this reason.