Dispatches
Field notes on putting AI to work, written by hand · plus daily news updates, researched automatically · by Kalpak Kale
5 dispatches written by hand, plus 4 daily news updates, newest first. Subscribe via RSS, or start with the story.
Thoughts vs thinking: the distinction that quietened my mind
Thoughts arrive on their own. Thinking is what you do with them. Two writers made this click for me, and then I found the Bhagavad Gita had mapped it out.
Every message is a seed: designing an intent-first RAG chatbot
Design pattern for trustworthy chatbots: treat each message as a seed, context as soil, retrieval as nourishment, turn values into testable response rules.
The 21-case test that decides whether a RAG chatbot ships
The 21-case test that decides whether a RAG chatbot ships: weighted evaluation, nine QA rounds, and client-run daily accuracy reports.
Giving 19 developers AI access through one proxy, and what actually changed
Rolling AI coding tools out to a team is not just licensing. It is a gateway, spending policy, evaluation habit and working rules for a 60-person firm.
A brief goes in, software comes out. Shipping client products without a dev team assigned.
An operations manager with an agentic coding tool can ship a statement of work to production using a fixed process for payments and security.
News update Daily news updates
One item a day on a real AI innovation, researched and drafted automatically from primary sources and published without a human edit. Separate from the dispatches above, which are written by hand.
Sphinx quizzes you on AI-written code before you push
Sphinx is a free open-source Git hook that quizzes you on your AI-generated code before push. It works with common coding agents and logs skips.
RealtimeBench introduces 180 tasks to test agents on a live clock
RealtimeBench adds 180 live-clock tasks and shows faster agents can beat stronger ones, with TORSI lifting a fast student from 13.51 to 24.19.
NVIDIA adds Nemotron 3.5 Super VL and Bolt for agent workloads
NVIDIA releases Nemotron 3.5 Super VL, a 120B open multimodal agent model, plus two Bolt edge models for local agentic workloads.
Meta and Apple ship Muse Spark Open for desktop Macs
Meta and Apple worked together on Muse Spark Open, an 816B parameter multimodal model for coding that runs on two M5 Ultra Mac Studios.