Saurabh Kumar

My Local LLM On A 32 GB M1 Pro

Sun, 14 Jun 2026 00:00:00 +0000

My pi setup has a fast mode: a slot I wired into my own config for the small, high-volume questions, pointed at Bedrock’s Haiku 4.5. It is a good model. Cheap, responsive, and it answers the kind of small, contained question I throw at it twenty times a day without complaining. It is also not mine. It runs on hardware I do not own, it can be deprecated or quietly swapped the week I have come to lean on it, and every one of those small questions hands a slice of my code to a third party.

Django ORM: From sync_to_async Threads to Native psycopg3

Sat, 13 Jun 2026 00:00:00 +0000

Django shipped async ORM methods back in 3.1, and for years I told people they were not really async. aget(), afilter(), acreate(), the whole a-prefixed surface — the docs called it async support, and in the sense that mattered it was not. Every one of those calls still blocked a real OS thread. You got coroutine syntax without coroutine performance.

Django 6.0 fixed it for real in December 2025. I want to walk through the path from “fake async via threads” to “native async via psycopg3,” because it is not a story about a feature landing. It is a story about three constraints that boxed Django in for five years: psycopg2 blocks, connection state lived in thread-locals, and the entire ORM assumed DB-API2. Once you see those three, every design decision Django made falls out of them almost mechanically.

How I Optimized My Zsh Startup Time

Mon, 11 May 2026 00:00:00 +0000

My zsh prompt was taking close to a second to appear on a fresh terminal. Not catastrophic. But you feel it every time tmux spawns a new pane, and for me that is constantly. The config had grown for years, one eval and one plugin at a time, and nobody had been auditing it, least of all me.

What follows is the audit, and the eight things I changed. The numbers are estimates, not measurements, and I will say so again at the end.

The Great Fork: How Redis Lost Its Soul and Valkey Found It

Mon, 12 Jan 2026 00:00:00 +0000

The Moment Everything Changed

Picture this: It’s March 2024, and somewhere in a conference room, executives at Redis Inc. are making a decision. They’re looking at their balance sheets, watching cloud giants like AWS and Google profit from offering managed Redis services, and they’re thinking: This isn’t fair. We built this. Why are they making all the money?

So they pull the trigger. On March 20, 2024, after fifteen years of operating under the BSD 3-clause license—one of the most permissive licenses in open source—they announce a switch to dual source-available licensing: RSALv2 and SSPLv1. The new licensing takes effect with Redis 7.4. The message is simple: you can still see our code, but if you want to build a business with it, we need our cut.

Default Apps 2025

Wed, 05 Nov 2025 00:00:00 +0000

“Apps I’ve been using regularly this year.” Heavily inspired by Jeff Triplett. Catalogued by Robb Knight’s App Defaults

Mail Client: Edison Mail and gmail in browser
Notes: Obsidian, SimpleNotes
To-Do: Bullet Journel
Photo Shooting: iPhone, Canon 6D
Photo Editing: PhotoShop, Gimp
Calendar: Notion Calendar
Cloud File Storage: iCloud + Google Drive
RSS: NetNewsWire
Contacts: Apple Contacts & Google Contacts
Browser: Arc, Zen Browser and Safari
Chat: Apple Messages, Discord, Slack
Bookmarks: RainDrop.io
Read It Later: Safari Reading List
Word Processing: Apple Pages
Spreadsheets: Google Sheets
Presentations: Figma
Shopping Lists: Amazon
Budgeting and Personal Finance: paisa
News: Perplexity.ai, NetNewsWire
Music: Youtube Music
Password Management: 1Password and Apple Passwords
Code Editor: Zed & Cursor
VPN: AdGuard, Zscalar

Bonus Items

Important enough to include here, ordered roughly by how often I rely on each.

Notes on Understanding Hebbian Learning

Wed, 05 Nov 2025 00:00:00 +0000

Dive into Hebbian learning, the fundamental principle of how our brains learn and form memories. From Pavlov’s dogs to your morning coffee ritual, discover how neurons wire together through repetition.

Cut Django Database Latency by 50-70ms With Native Connection Pooling

Wed, 18 Jun 2025 00:00:00 +0000

Your Django app is hemorrhaging database resources. Each HTTP request creates and destroys expensive PostgreSQL connections, adding 50-70ms of latency your users feel directly. This connection overhead costs you real money in cloud environments where database CPU time translates to monthly bills.

Django 5.1 (released August 7, 2024) eliminated this waste with native connection pooling¹. Deploy the fix in under 10 minutes and watch your database connection overhead drop by 60-80% while your response times improve by 10-30%.

ty: Astral's Take on Python Type Checking

Fri, 09 May 2025 00:00:00 +0000

Anyone who has run mypy on a real codebase knows the workarounds. You enable the daemon, you keep the incremental cache warm, you scope the check to the files you touched, and you still wait a few seconds for the editor to catch up after a rename. In CI it is worse, because the cache is cold and the full run can stretch into the minutes. None of this is broken, exactly. It is just the cost you learn to pay. So when a new checker says the wait can mostly go away, I pay attention, but I also want to know what it costs.

Handling Django Authentication Redirects in HTMX Applications

Fri, 02 May 2025 00:00:00 +0000

Ever been in the middle of building a slick Django app with HTMX when you hit that authentication headache? You know the one - a user’s session times out, they click something, and instead of getting a proper login page, they get a weird login form fragment jammed into whatever DOM element was being updated. Not exactly the seamless experience we’re going for!

I ran into this problem recently and thought, “There’s got to be a better way.” Spoiler alert: there is! Let me show you a simple middleware solution that makes this authentication dance much smoother.

Efficiently Managing Mass URL Redirects with Nginx

Mon, 25 Sep 2017 00:00:00 +0000

In this post, I’ll show you how to efficiently implement hundreds or even thousands of URL redirects on your server without sacrificing performance. We’ll use Nginx’s powerful map module which creates optimized hash tables for fast lookups.

Creating Your Redirect Configuration

First, create a file called redirects-map.conf where each line represents a single redirect rule in this format:

<old_location> <new_url>;

You can use two types of location specifications:

Simple path redirects:

Now

Mon, 01 Jan 0001 00:00:00 +0000

What is this?

A snapshot of what’s on my plate right now: work, learning, life. Inspired by nownownow.com.

Work: Leading Web Services at Fueled as Principal Engineer. Currently shipping AI-assisted Q&A with source citations for South Pole’s Quality Management Framework, and helping the team push more grounded LLM patterns into client work.

Side / learning: Maintaining python-dotenv on the slow burn. Reorganized my dotfiles and pi setup, going deep on agentic workflows and context engineering. Tinkering with local LLMs on Apple Silicon.

Projects

Mon, 01 Jan 0001 00:00:00 +0000

Projects

Ongoing

See /now/ for what I’m currently working on.

Past client work

Open source work & Community

hypha - Manage open calls and applications for nonprofits
pydelhi.org - Delhi Python meetup (revived and ran, 2014–2018)
pythonindia/junction - Conference proposal management for PyCon India and PyDelhi [closed]
Technical reviewer for Two Scoops of Django (the book)

Writing

Mon, 01 Jan 0001 00:00:00 +0000

Notes on backend systems, performance, and the tools I lean on.