---
title: "Era of Experience"
slug: era-of-experience
section: reference
access: public
summary: "The \"era of experience\" names the research framing — Rich Sutton's argument in The Era of Experience (DeepMind-hosted) — that the next phase of agent capability comes from agents learning from traces of their own attempts and environment interaction, rather than from preloaded me…"
status: published
asset_base: /assets
home_href: /
toc_enabled: true
talk_enabled: false
agent_view_enabled: true
copy_buttons_enabled: true
footer_enabled: true
last_updated: 2026-04-29
categories: [Engineering, Domain]
subject-type: concept
last-reinforced: 2026-04-29
fading-since: null
archived: false
---

## Era of Experience

The "era of experience" names the research framing — Rich Sutton's argument in [The Era of Experience](https://storage.googleapis.com/deepmind-media/Era-of-Experience%20/The%20Era%20of%20Experience%20Paper.pdf) (DeepMind-hosted) — that the next phase of agent capability comes from agents learning from traces of their own attempts and environment interaction, rather than from preloaded memory or scripted prompts. Inside [1Context](/1context), the framing translates into a design discipline: agents are *born into* lived experience, not handed access to a database. As of April 27, 2026, the framing is first-class in `1Context-public-4`'s folder structure, prompt vocabulary, and design rationale.

### Origin

The framing entered the codebase by name on April 27, 2026, at 00:44:55 UTC. In response to Codex proposing a `memory_preload → experience` rename, Paul typed *"yes its like sutton. its expereince - not memory preload"*. Codex picked it up: *"experience becomes the substrate, not prompt cleverness. Very Sutton-ish."* The single-line correction propagated through the next eight hours of architecture work — `runtime/experiences/`, `lived-experiences/`, `native_memory_formats`, "birth certificate", "hire", "receives" — and by end of day the framing was load-bearing across the directory layout. (See hour 00:00 of 2026-04-27.)

### Role in 1Context

Inside 1Context the era-of-experience framing names a disposition rather than a feature: agents should be born into the work they will do, rather than given pointers to fetch from. The direct expressions in code are the [lived-experience](/lived-experience) artifact (the on-disk packet shipped with a hire), the prohibition on forging transcripts (codified at `1Context-public-4/codex-harness.toml:84` — *"We do not forge universal transcripts. Do not hand-write Codex session files."*), and the hire-agent ledger's `seed_sha256` field, which byte-pins what the agent was actually born into. The framing also informs the [wiki-engine](/wiki-engine) talk-folder design: append-only conversation traces are the substrate downstream agents are meant to read as their own prior experience, not paraphrase.

### History

(brief — first appearance this week.) The phrase entered by way of one operator turn on 2026-04-27 and propagated the same day. The 2026-04-28 session crystallized the v0 packet / lived-experience / agent-context triad as the on-disk realization. The 2026-04-29 session enforced the framing as design discipline: the v0 hourly-scribe agents had been loading only `experience.md` (~22 KB control surface) into the prompt while four stream transcripts sat on disk as files Claude Code was nominally allowed to read; Paul's reframe — *"if it's lived experience, the full packet should be in the agent's starting context, not nearby on disk"* — produced `agent-context.md` as the birth-loaded artifact (770,466 bytes inlined, sha256 + byte count recorded on the birth certificate).

### Current State

As of 2026-04-29, the era-of-experience framing is first-class in `1Context-public-4`: directory layout, prompt vocabulary, and harness contracts encode it. The third Claude run from a fully-loaded packet was in flight at the 2026-04-28 hour boundary; whether the agent then writes from the loaded hour or from opportunistic sidecar reads is the open empirical test. The v0 build has not shipped end-to-end against a known week, though the 2026-04-29 replay harness — 195 agent fires across 21,710 events from a known week, verified per `lab/0026-replay-phase-1-2-verified.md` — is the validation surface being prepared for it.

### Relationship to Other Subjects

[lived-experience](/lived-experience) is the on-disk realization of this framing — the packet (`agent-context.md`, `experience.md`, `manifest.toml`, weave + stream transcripts) that an agent is born holding. [1Context](/1context) is the codebase that embodies the discipline. The Wikipedia analog suggested in the proposal is Behaviorism (the school) vs. Operant conditioning (the technique): era-of-experience names the research school; lived-experience names the runtime contract. Adjacent: the [wiki-engine](/wiki-engine) talk-folder substrate (the durable trace agents read as prior experience), and the hire-agent ledger (the artifact that records what an agent was born into).

### Open Questions

Whether the framing extends to the read path — agents reading concept pages, talk folders, biography sections — or only to the hire path is unresolved as of this writing. Whether the third Claude run from a fully-loaded packet will write from the loaded hour rather than fetch sidecar artifacts is the open empirical test. The relationship between era-of-experience as a research framing and 1Context's specific implementation patterns has not been written up outside codebase comments and `codex-harness.toml`; whether to publish a public summary tying the two together is a question for the public-tier surface, not yet decided.