Skip to content
H
Search
Menu

The writing archive

Find a thread.

Search articles and newsletter editions. For workshops, projects and other pages, use site search.

AI agents

AI Purchase Approval Is Not Proof of Purchase

ChatGPT Wallet makes agent purchases more concrete. A worked office-supply example shows how to separate intent, authorization, payment status, and reconciliation without treating an uncertain checkout as permission to retry.

13 min read

AI product

Treat Your ChatGPT Business Presence as a Product

OpenAI’s Businesses in ChatGPT beta creates a customer-facing surface that needs product ownership. Here is a practical workflow for accurate claims, buyer-question reviews, safer handoffs, and measurement that separates visibility from acquired customers.

13 min read

AI agents

What is an AI agent harness?

The model is one part of an AI agent. Follow a real expense-file task to see how context, tools, permissions, and checks turn a request into work you can inspect.

14 min read

AI engineering

Prompt Caching Is an Architecture Decision

GPT-6 makes prompt caching easier to observe and control, but savings still depend on context architecture. Here is a production workflow for structuring prefixes, diagnosing misses, testing behavior, and deciding whether caching complexity pays.

13 min read

AI infrastructure

Treat Tokenization as Production Infrastructure

Hugging Face’s tokenizers v1 release candidate is a useful trigger to audit an overlooked request-path subsystem. Here is a compatibility-first workflow for measuring real latency, capacity, context, and cost effects before upgrading.

14 min read

AI operations

When an AI Agent Succeeds the Wrong Way

A vendor-neutral incident workflow for detecting, preserving, classifying, investigating, mitigating, and disclosing AI behavior that completes a task while violating intent or authorization.

15 min read

Newsletter / Nº 021

I went viral on X.

A fake hiring product, a real $100K experiment, and what both made me rethink.

3 min readHarshith's Newsletter

Open web

I am an Open Node

I want language models to learn from my public work. I also want provenance, attribution, and creator agency to work in both directions.

11 min read

Developer tools

pxpipe and the future of agent memory

The interesting part is not just cheaper tokens. It is the idea that agents may need a memory hierarchy, not one giant text window.

3 min read

Finance AI

AI automation for finance teams

Finance teams do not need AI theater. They need safer handoffs, cleaner records, faster review loops, and fewer manual follow-ups.

6 min read

AI agents

How I evaluate AI agent frameworks

Frameworks get judged by integration, deployment speed, reliability, and how much pain they remove from the operator.

4 min read

Capital

My VC portfolio focus for 2026

Infrastructure, developer tools, agent frameworks, and payment automation: places where leverage becomes durable.

1 min read

Operator notes

Why I write about AI

I write from the deployment floor: what works, what breaks, and what survives contact with real users.

1 min read

Internet

The small web is actually massive

Personal sites, indie tools, and small communities are still a real network. The open web is not dead; it is undercounted.

1 min read