RSSAmplifier

Blog

Shlok Khemani - Writing

Thoughts on technology, startups, and life by Shlok Khemani

shloked.comRSS feed ↗43 posts

Latest posts

Unpacking ChatGPT Work: the Agent for a Billion Users

ChatGPT Work is a preview of how a billion people will soon use ChatGPT. I spent days inside it unpacking its design, its tensions, and where it's headed.

On the Interaction Acquisition

Poke changed my life, yet I stopped paying for it after a month. Reflections on what Interaction built, why it didn't retain users, and the sale to Cognition.

Fable 5: First Impressions

I gave Fable 5 a single vague prompt — 'build me a to-scale 3D Yosemite I can roam' — and it ran further than I could have imagined. Why this model has that 'big model smell.'

ChatGPT's Memory Update Has a Packaging Problem

OpenAI shipped an editable Memory Summary for ChatGPT. After reverse-engineering it, I can't figure out how it makes the product better.

Open-Sourcing Vajra: A Background Coding Agent

On building and open-sourcing Vajra, a background coding agent that works through Linear, GitHub, Slack, and graph-based workflows.

What I Found Interesting in Claude Code's Source

Tactical patterns across prompt architecture, context management, and agent orchestration from Claude Code's leaked source — things you can study, steal, or adapt for your own work.

Google Has Your Data. Gemini Barely Uses It.

Inside Gemini's cautious memory system—and what it costs Google in personal AI.

Anthropic's Opinionated Memory Bet

Anthropic ships native memory as six file operations, betting that models should handle their own retrieval and storage rather than relying on separate systems.

OpenPoke: Recreating Poke's Architecture

Reverse-engineering the Poke assistant and open-sourcing a working prototype of its multi-agent architecture.

Bangalore → Mumbai → Bangalore

Why I'm moving back to Bangalore after 10 months in Mumbai

Claude Memory: A Different Philosophy

How Claude memory works, how it differs from ChatGPT, and what these approaches reveal.

ChatGPT Memory and the Bitter Lesson

I reverse-engineered ChatGPT's memory system to understand what it stores, how it works (spoiler: not RAG!), and why OpenAI built it this way. Complete with prompts you can try yourself.

Seven Weeks in San Francisco

A tech bro's non-tech notes

7 lessons from launching my first AI product

In April, I left my job to build software products with AI tools. My first experiment was Dweli, a personal website builder. Two months later, I launched it and... it didn't feel great. Here are the lessons I learned.

Notes from watching a master at work

Last night, I watched Academy Award-winning musician AR Rahman perform live in concert. What I didn't expect was for it to become a spiritual experience that I simply have to share with the world.

AI is Compressing Crypto Cycles

How AI accelerates software creation, market evaluation, and attention—compressing speculative cycles across crypto.

ChatGPT Deep Research: First Impressions

First impressions of ChatGPT Deep Research from a long-form technical writer: where it excels, where human judgment still matters, and what changes next.

Thinking about aixbt on a Sunday Night

Why aixbt became crypto's most compelling social AI agent—and what its personality, performance, token, and technical design reveal about the category.

Decentralized Compute

On LLM Training, GPUs, Data Centres, and where Crypto fits in

Bitcoin Superconductor

How Chainflip is bringing BTC liquidity on-chain

Sentient AI Models

Monetising Open Source AI

Does Crypto Matter?

How crypto will (and will not) change the world

Actively Validated Services

How EigenLayer is unshackling crypto entrepreneurship

Trusted Enclaves

On TEEs and why they're important

Bootstrapping Networks

Solving the cold start problem with zkTLS

Abstracting Chains

Fixing Crypto UX

The Road to Runes

Charting the long-winded journey to a fungible token standard on Bitcoin.

On Token Migrations

When Tokens Pivot

The Data Must Flow

Easing the struggle for AI data

Danke, Jürgen

To the German who changed my life.

Enter The Sanctum

Everybody gets a stake.

MEV On Solana

Breaking down Jito's mempool rollback

The Heartbeat of a Global State Machine

A deep dive into the importance of and challenges faced by validators on Solana

2023

Reflections, learnings, and setting the stage for the new year.

Multi-Model Apps and OpenRouter

On the need for developers to adopt a multi-model LLM approach and OpenRouter, a service facilitating it

Pricing Models for LLM Apps

A new paradigm in tech requires rethinking how companies price software

A Guide to LLM Capabilities

A practical breakdown of how LLMs work and the new capabilities they unlock

Mardi Himal

Why this Himalayan trek in Nepal has to be your next vacation

Now is the Time to Code

Making a case for why you should learn to code over the next year

On Leaving Web3 Gaming

Some reflections on the space, where I see it going, why I am leaving, and what next

The Zynga Mafia in Web3 Gaming

Zynga is one of the most important companies in the history of gaming. Now, the folks that made it big, are trying to replicate their success on the blockchain.

South Korea: Underrated Leaders in Web3 Gaming

The country has pioneered some of the biggest movements in gaming history. Now, it is right at the forefront of another one

Skyweaver: Fulfilling the Promise of a Crypto Game

Lessons to learn from the dark horse of crypto gaming