https://rohitghumare.com/ Essays on AI agents, harnesses, and backend fundamentals. https://rohitghumare.com/blog/forward-deployed-engineer/https://rohitghumare.com/blog/forward-deployed-engineer/2026-08-01Between May and July 2026, OpenAI, Anthropic, Microsoft, and AWS committed a combined nine billion dollars to organizations built around one job title. The mechanism behind the comeback of the forward deployed engineer, from Palantir's Deltas to the agent era. https://rohitghumare.com/blog/what-a-forward-deployed-engineer-does/https://rohitghumare.com/blog/what-a-forward-deployed-engineer-does/2026-08-01The FDE job from the inside: deliverables named in real OpenAI, Anthropic, and Harvey postings, pod structure, sourced pay ranges, and the three questions that reveal whether the role is real. https://rohitghumare.com/blog/tokens-per-second-memory-bandwidth/https://rohitghumare.com/blog/tokens-per-second-memory-bandwidth/2026-07-31Local LLM generation speed is not a compute number. Every generated token reads the model's weights from memory once, so tokens per second is bounded by bandwidth divided by weight bytes. The arithmetic, the consequences, and what it predicts. https://rohitghumare.com/blog/agents-md-genre/https://rohitghumare.com/blog/agents-md-genre/2026-07-31Every repo now carries a document addressed to a machine, written in shouted imperative English. It is a new genre of technical writing, it has conventions already, and you can read a team's incident history in it. https://rohitghumare.com/blog/kimi-cli-k3-harness/https://rohitghumare.com/blog/kimi-cli-k3-harness/2026-07-31Moonshot ships a harness that is also your shell, rewinds context with checkpoint time travel, fans out 128 subagents, and requires reasoning to be passed back verbatim. And its own benchmark admits K3 scores higher inside Claude Code. https://rohitghumare.com/blog/inside-the-pi-harness/https://rohitghumare.com/blog/inside-the-pi-harness/2026-07-31Pi stores sessions as trees, keeps its system prompt to twenty stable lines, refuses to prune history, and itemizes every cache miss in dollars. A mechanism-level tour of the most transparent coding agent harness. https://rohitghumare.com/blog/inside-the-codex-harness/https://rohitghumare.com/blog/inside-the-codex-harness/2026-07-31Codex made compaction a server-side primitive: a dedicated endpoint, in-stream compaction items, a 90 percent trigger, and a model trained on where the summary sits. OpenAI tripled its own ARC-AGI-3 score by turning it on. The price is a transcript you cannot fully read. https://rohitghumare.com/blog/inside-the-claude-code-harness/https://rohitghumare.com/blog/inside-the-claude-code-harness/2026-07-31Claude Code assembles context in layers ordered by how often they change, delivers CLAUDE.md as a user message, and treats cache preservation as an architectural principle. A mechanism-level tour from the current docs. https://rohitghumare.com/blog/graph-engineering/https://rohitghumare.com/blog/graph-engineering/2026-07-31A harness generated a 400-line throwaway workflow that spawned a hundred agents, and the industry named the shape graph engineering. Euler named it in 1736. What actually changed is the node. With two interactive figures: walk the Konigsberg bridges yourself, and run the agent DAG with live cost arithmetic. https://rohitghumare.com/blog/codebase-second-reader/https://rohitghumare.com/blog/codebase-second-reader/2026-07-29Every rule about readable code, DRY, good names, one source of truth, was always for the next human. A second reader arrived: faster, more literal, and it audits how much of that advice you actually followed. https://rohitghumare.com/blog/dual-write-trap/https://rohitghumare.com/blog/dual-write-trap/2026-07-29Emitting an event after a database commit without an outbox is a lossy notifier that drops events on crash. It is the dual-write problem, and the fix is the outbox pattern or real change data capture. https://rohitghumare.com/blog/composable-agent-runtime/https://rohitghumare.com/blog/composable-agent-runtime/2026-07-29Thin versus thick harness is the wrong debate. A backend is workers, triggers, and functions, and an agent is just another worker. Once the harness is the backend, agent reliability becomes durable execution, and composability is how the hard problems get solved. https://rohitghumare.com/blog/new-names-old-fundamentals/https://rohitghumare.com/blog/new-names-old-fundamentals/2026-07-29Harness engineering, context engineering, loop engineering. Most of the agent era's new vocabulary renames engineering fundamentals we have had for decades. Here is the old idea under each new name. https://rohitghumare.com/blog/how-to-become-an-ai-engineer/https://rohitghumare.com/blog/how-to-become-an-ai-engineer/2026-07-29A practical, honest roadmap to becoming an AI engineer in 2026: the skills in order, two routes in, time estimates, and one free 503-lesson curriculum that builds the whole stack from scratch. https://rohitghumare.com/blog/harness-engineering/https://rohitghumare.com/blog/harness-engineering/2026-07-29Harness engineering is the practice of improving agent output by shaping the environment around a fixed model: context, tools, the loop, and proof. A field guide to where the leverage actually lives. https://rohitghumare.com/blog/stop-agents-building-the-wrong-thing/https://rohitghumare.com/blog/stop-agents-building-the-wrong-thing/2026-07-29Agents build the wrong thing, ship duplicates, and reinvent what you already have. The fix is not a smarter model, it is a system the agent is part of. Early results: 60% fewer tokens, 2x faster from design to build.