← narwal.one/Second Brain
SecondBrain
Ask the Brain
Index/Conceptupdated Wed Jul 08 2026 08:00:00 GMT+0800 (Philippine Standard Time)

AI Operating System (AIOS)

aiosclaude-codesecond-brainagentspersonal-knowledgeenterprise

AI Operating System (AIOS)

A single AI workspace — for Nate Herk (AI Automation), Claude Code running Opus 4.8 — that you work out of by default instead of tab-switching across chat apps, browsers, and SaaS tools. It holds all your context, can touch your real tools, and runs work on a cadence. Nate's term for his "second brain / executive assistant" once he realized he reaches for it first.

The analogy is literal: like Windows/macOS/iOS, an AIOS is "the way you interact with your systems and get work done." The substrate is just files and folders, which makes it tool-agnostic (the same OS opens in Claude Code, Codex, or OpenClaw via per-agent config folders) and fully crawlable/reorganizable by the AI itself.

What makes it an OS (vs. a chatbot)

  1. It knows your business — meeting transcripts, posts, threads, email, docs. Recall better and faster than the human owner.
  2. It's the default surface — see Default Shift. Brainstorming, writing, and planning happen here too, not just code.
  3. It's structured for scale — folders for decisions, audits, archives, and "other worlds" (standalone sub-projects), so context stays findable as the business evolves. By the Fable-era walkthrough (I Turned Claude Fable Into The Ultimate Second Brain (Nate Herk)) "other worlds" has matured into the load-bearing pattern: formerly separate repos pulled inside the main project so one agent cds between them and one GitHub push syncs everything — the structural twin of this machine's ~/worlds.md hub. See Architecture Engineering for the routing-tree discipline that keeps it navigable.
  4. It compounds — every session adds context; every skill run improves the skill. "If it slips up, you get data" — even the 150k-email incident was treated as a case study, not a firing offense.

How it's built

Two frameworks from the source:

  • Three M's — Mindset, Method, Machine (the how you think about it layer).
  • Four C's Framework — Context → Connections → Capabilities → Cadence (the architecture; each layer depends on the prior one).

The hard-won caution layer: as reach grows, so does risk — govern it with the Bike Method and by scoping capabilities, not just instructions.

The identity claim, sharpened in the Fable walkthrough: "You're not building a Claude Code AIOS — you're building your own personal operating system." Folders, markdown, skills, and routing logic are portable IP; Nate keeps parallel claude.md/agents.md configs and runs the same OS in Codex daily. Model/harness churn is plug-and-play, which "should remove the overwhelm about staying up to date with the latest drops."

Relation to this vault

This Second Brain is itself a proto-AIOS: an LLM Wiki Pattern store of files/folders that an agent maintains. The applied question of upgrading this machine into a full AIOS is worked through in Turning This Mac Into an AI Operating System.

2026-07-08 — The enterprise-scale cousin is now a YC batch category

YC Summer 2026 Requests for Startups (The VC Corner) item #15 — "The AI Operating System for Companies" (Diana Hu) — is Nate Herk's personal AIOS thesis promoted one altitude to the organisation. Hu's framing (verbatim):

"The best AI-native companies have made their entire company queryable. Every meeting recorded. Every ticket tracked. Every customer interaction captured and legible to an intelligence layer that learns from it. This turns a company from an open loop into a closed loop."

Hu reports observing teams "cut sprint time in half and ship twice as much" once that layer exists. The blocker she names is not model quality but the brutal integration work stitching Slack, Linear, GitHub, Notion, call recordings, and a dozen other tools with custom glue code. The product to connect all of that into a single intelligence layer, in her framing, "does not exist yet."

The personal-AIOS→enterprise-AIOS relationship is not scale-linear:

  • Personal AIOS (Herk / this vault) — governance surface is one operator, consumer surface is one operator. The design constraint is keeping context findable as the operator's work evolves.
  • Enterprise AIOS (Hu / YC) — governance surface is the org, consumer surface is every employee and every agent. The design constraint is keeping meaning consistent across humans, systems, and agents while the org evolves.

The two concepts share substrate (files, skills, retrieval, cadence) but diverge on governance and multi-user semantics. The bridge concept between them is Company Brain — Blomfield's YC RFS #4, which is the knowledge/skills layer an enterprise AIOS runs on top of. Personal AIOS + Company Brain ≈ Enterprise AIOS, with the Company Brain supplying the substrate the AIOS pattern uses at organisational scale.

The vault should track these three concepts as a triangle, not a hierarchy:

  • Personal AIOS — Herk / this vault's ~/worlds.md machine
  • Company Brain — Prukalpa's enterprise context layer + Blomfield's YC framing
  • Enterprise AIOS — Hu's YC RFS #15; not yet a dedicated concept page but implicit as "personal AIOS + Company Brain at org scale"

Cross-links