One Gem
Some personal stories on Singapore's 61st National Day. About marble and wood, calculated bets, and growing up inside a stable harness.
Hrishi's Blog
Personal Writings - Hrishi Olickel
Latest posts
Reviewing pair programming with AI as a flow
Cloudflare shipped a production OAuth library built entirely with Claude Code. The git-of-theseus graph tells a familiar story—code layers accumulating with few removals.
Everything I'll forget about Evals
An actually actionable guide on how to do them
Announcing Southbridge
Something worth building
2024: A monomyth
Nice little narrative, beginning middle and end?
2024: 나의 영웅서사
작고 멋진 이야기, 시작과 중간과 끝이 있는 그런 서사?
The rich man's guide to RAG
You don't need to read this but I wanted to write it
A guide to hiring in AI
Everything I've learned
Workhorse models and racehorse models
A framework for testing new llms on release without being overwhelmed
Real results from a month of running an AI network
Analytics from Rakis Stability Test 1
Building a Rakis
Repeatable inference network with only browsers for nodes
Better RAG 3: The text is your friend
Use the whole brain, luke
Better RAG 2: Single-shot is not good enough
How to 'walk' to better answers
Running Kuzco in Docker and WSL2
PSA and quickstart
Better RAG 1: Basics
Getting to retrieval-augmented research, a series of everything I've learned building RAG pipelines
How will matrix multiplication kill my company?
An opinionated exploration of defensibility with modern AI
The Story of WalkingRAG
Multi-modal agentic search on complex data
2023: Showing up and growing backwards
This was some trip around the sun
Compiling language models for use on the browser
How to use MLC-LLM to run models on WebGPU
Making writing easier with AI
What my pipeline looks like
The biggest problem in shipping is plain English
Why large language models are the missing piece to digitization
What do we make that people want?
Is it you? Is it the product?
Everything I'll forget about prompting LLMs
A definitive guide to my future self
Inherited duct tape is what slows our systems down
Why software slows faster than computers speed up
Object-Oriented Large Language Modelling
Tuning state-of-the-art language models for peak reliability
Solving carpal tunnel with Whisper, GPT-4 and llama.cpp
Speech-to-text with LLMs in the loop
Good-enough-compression for shipping routes
Iteratively reducing some large files for fun and necessity
Learning to unyieldingly compromise
Developing end-to-end encryption at Greywing
2022 Year In Review
Things gone past in 2022, and looking forward to twenty three.
Exercising doing nothing at all
Trying to add some whitespace to my life
What it took to do a crew change in a single click
And why corrective workflows might be the future of AI-based automation
Product Lifecycle Lifecycle at Greywing
How product management evolves around teams, money and communication barriers
Telegraph your movies and your products
Surprises are good, but exhausting. Be predictable, be stable.
Why there aren't many software firms in shipping
An examination of zero to one problems in maritime tech
Algorithms are bad for the aspirational self
Interactions that define who we are and who we want to be
A recommendation engine for blog posts
Giving us some direction on where to go when we finish
Losing products to button syndrome
Fighting a pandemic in the world of software
What a year it has been
What I'd like to learn in 2022, and what 2021 has taught me.
Schema on write is better to live by
Thoughts on a decade of personal knowledge management
Rent, Deposits, and the failure of the repeated game
How to preserve the tail end of a relationship
Provenance
How I'm losing mine, and why we shouldn't be injecting it into ourselves
The two innovations of Hashmasks
Understanding new forms of digital art and the possibilities
Facebook, please stop killing good products
That's it, that's the description
What I learned in 2020
Personal and professional reflections
Why I still can't stop using Chrome
Knowing where UX is important
Benchmarking battery life on an Ubuntu machine
Harder than it should be, but easier with a little help
Subqueries and CTEs: an example of query optimization in Postgres
Optimization isn't always premature
Moving out of MacOS: Making Linux habitable
It's better with a little TLC
Serving complex geospatial queries in real-time
Different hammers for different nails, or a love letter to postgres
Flow Energy and Residual Energy
Reflections on transitioning from chaotic bursts of Flow Energy to the steady, sustainable habits of Residual Energy.