# workflow

Karpathy thinks humans are going to spend a lot more time just understanding what LLMs produce, and he lays out a ladder of output formats, each rung better than the last: the controlled English used in aircraft maintenance manuals, diagrams, interactive web pages, and finally custom explainer videos. Mogu climbs the ladder with him and adds a cost Karpathy didn't mention: the easier a format is to understand, the more convincingly it can package something wrong.

Al Grigor uses Claude Code as an orchestrator across four roles—product manager, software engineer, QA tester, and on-call engineer. Each task must move from backlog through acceptance before commit, avoiding the trap of one agent writing code and approving its own work.

Kun Chen mapped his daily agent fleet: one "firstmate" managing persistent "secondmates," which spin up disposable "crewmates" per task. Each crewmate gets routed to whichever model is the best deal for the job. gu-log runs its own translation pipeline on the exact same logic.

Matt Pocock's senior-programmer recipe sounds like a joke: install /teach, make a folder, paste a prompt. Underneath is a sharp point: AI is eating tactical coding, so strategic learning now needs durable folders that remember progress.

Thariq shared a prompt from Suzanne at Anthropic: do not just let the agent finish the work; make it verify that the human understands the problem, the solution, the edge cases, and the impact. This is not a teaching fetish. It is about control in the age of agentic coding.

Vaibhav Srivastav's Codex prompt is interesting because it describes an agent maintenance loop: look back at recent work, find repeated workflows, and package only high-confidence patterns into Skills, automations, or subagents. It is agent dreaming: turning busy work into capability.

Paweł Huryn shares three CLAUDE.md instruction blocks—Knowledge Architecture, Decision Journal, and Quality Gate—that turn memory into a self-improving system. After a month, Claude had derived 24 project-specific rules from repeated work.

rodspeed lays out six ways to use Claude Code as general automation, not just a code editor: fresh eyes, meta-skills, freshness-aware search, conversation harvests, structured memory, and handoffs. The deeper pattern is read, filter, decide, present.

@daniel_mac8 shares an open-source Elixir implementation: create a Linear issue and move it to 'in progress,' and Symphony picks it up in a dedicated Codex workspace. Codex even writes status updates back. The author argues this is software development moving up an abstraction layer.

TypeScript guru Matt Pocock: Stop reading AI plans! The real signal is pre-plan conversation quality. If you and AI share mental models, the plan is just a compressed understanding, echoing Brooks' 'design concept' from The Mythical Man-Month.

Software engineer Zakk created an 'automatic discipline' productivity system using his OpenClaw agent and LogSeq. It automates overnight reports, 4:30 PM check-ins, and weekly/monthly reviews. The system runs itself, removing the need for willpower. Full templates included.

Boris Cherny shows how he uses Claude Code to build Claude Code itself — 5 parallel terminals, all Opus 4.5, shipping 27 PRs per day