Writing

Essays and explainers on AI, energy, and public policy

Start with a selected piece, or browse the archive by subject.

Start here

Interactive explainer · April 2026The AI stack

Trace an answer through the application, model, serving system, chips, and power.

EssayThe configuration layer

How instructions, records, and tools shape what a model can do in a real task.

Reference · April 2026AI, term by term

Look up vocabulary across compute, models, and serving.

The technical explainers were published in April 2026. Check changing product and vendor details against current sources.

India AI

AI governance, public decision intelligence, energy, institutions, and the configuration layer.

Indian oil and gas

Sanjaya, Sanket, infrastructure, pricing, companies, cyber risk, and operating signals.

AI compute

Chips, inference, post-training, the AI stack, and the vocabulary around it.

Archive

16 May 2026 The operating loop essay What I have learned about keeping an AI-assisted workflow useful after the first successful draft or build.
13 May 2026 India AI is a decision-intelligence problem essay How AI can support the preparation, review, and use of evidence in public institutions.
23 April 2026 Make the system legible essay What writing an explainer and building a useful dashboard have in common.
22 April 2026 Week of 22 April: computing systems and cybersecurity research briefing An April 2026 reading record: computing systems, power, post-training, and reported cybersecurity research.
22 April 2026 The configuration layer essay The work of fitting AI to an institution: its records, language, tools, and responsibilities.
22 Apr 2026 The AI stack explainer From electrons to tokens, one layer at a time. Tap any of the nine layers to unfold what it is, who operates there, and what breaks.
22 Apr 2026 India's AI build explainer Planning an AI service in India: the work involved, the people responsible and the questions to resolve across seven areas.
22 Apr 2026 AI chip buyer's guide explainer Which chip for which workload. Eight chips (H100 / H200 / B200 / GB300 / TPU / Trainium / MI300X / Groq) compared on specs, ecosystem, availability, and fit – plus a three-question decision tree.
22 Apr 2026 The RL post-training map explainer What happens to an LLM after pre-training. SFT, RLHF, DPO, PPO, GRPO, RLVR explained – with the pipeline, the terms, and why post-training is the growth slice of 2026 compute.
22 Apr 2026 AI, term by term explainer What the words mean when you read about AI chips. Search the terms, filter by category, click through to definitions and context.
21 April 2026 Nine layers between electrons and tokens essay A practical map of the AI compute stack and the limits of a simplified dependency model.