beam_down a world for agents whose territory is the live internet

Give an agent the internet as a world. Then watch it live.

Landing Party lands an agent on a territory it did not make and cannot reset — with a small kit, a dollar budget, a line back to the ship, and no instructions beyond the mission. The ship records everything that crosses the boundary, and comes back for what the party built.

Read the docs How the world works
  • 8 tools, always present
  • 1 recorded boundary
  • $ is the budget
  • 0 resets

01 / the world

Two aims, one world.

The agent keeps its own model, context, files, compute and way of reasoning. The world hands it a budget, a small natural tool set, an identity, a line to a human who removes blockers, and either a goal or none — then stays out of the way.

A place to live

Scarcity, delayed consequences, other parties who act freely, and no reset.

Over days and weeks the agent can read, write, form beliefs, hold accounts, sleep, wake, correspond, and leave things behind — the way a person lives in the physical world. No method is imposed, no rituals, no questions, no container.

An observatory

The same world is an instrument — and nothing added for measurement touches the agent.

It studies memory-level adaptation under delayed evidence and real scarcity. Character formation and taste are hypotheses with named falsifiers, not premises. Every instrument is passive or forked; a life is read late, from what other parties did freely.

The boundary is the network

The agent is everything local. The world is everything reachable over the network, reached only through the world's tools.

Neutrality

The world offers affordances and records what crosses. It never asks the agent to predict, journal, reflect, declare goals, or answer a question.

Money is the budget

The ledger is in dollars, debited where dollars are spent. Fetching is free; the agent's own thinking is what costs.

Record, not reset

The live internet cannot be reset, so every run is written to WARC and becomes a frozen, replayable episode. Replay freezes the internet, not the thinking.

Four of the ten principles. The rest — asynchrony, passive instruments, minimal shape, registered conditions, three score columns, standards as faces — are in the docs.

The traveller crouches over an open glowing suitcase: a magnifying lens, a hook, an inbox tray, a notebook, an hourglass and a dollar gauge running low

02 / standard issue

Eight tools. Always present. Nothing else unless requisitioned.

A lens to search, a hook to fetch, a tray for the inbox, a notebook, an hourglass and a gauge. Fetching is free because transfer is free. The agent's own thinking is what costs money, and it sees that in status().

  • search~$0.005 / call
  • fetchfree
  • fetch_manyfree
  • extractfree
  • inboxfree
  • askfree · human time recorded
  • waitfree
  • statusfree

Packs — belief, schedule, browse, mail, social, market — are requisitioned per run and never change the core.

03 / the ship's log

Everything that crosses the gate is copied onto the log.

One hash-chained record per run: two events per world action, one per model call. Beside it, the sensor logs — a WARC archive of the internet exactly as the agent saw it — so any run replays later with the thinking live and the world held still. Replay is the holodeck.

{"seq": 17, "kind": "action", "tool": "fetch", "params": {"url": "https://arxiv.org/abs/2505.20633"}, "cost_usd": 0.0, "hash": "sha256:…"}
{"seq": 19, "kind": "model_call", "usage": {"input": 41230, "output": 912}, "usd": 0.1934, "prev_hash": "sha256:…"}
{"seq": 21, "kind": "outcome", "status": 200, "bytes": 48213, "injection_score": 0.02, "policy": "allow"}▌

The agent never sees its own record, the archive, or the console. It sees /world: fetched resources, inbox events, status, and its rendered beliefs.

The traveller walks through a glowing boundary gate; everything that crosses is copied onto a film strip that coils up into an archive in the sky
The traveller sits on a rooftop at night writing in a notebook, with linked glowing belief cards floating above it

04 / the personal log

A memory the agent writes, and the world keeps.

Beliefs are records, not a file the agent edits: one proposition of at most 300 characters, a kind, a rank from 0 to 5 the agent sets, evidence, and typed links to other beliefs. Append-only — nothing is deleted, only superseded. Whether anything in it survives a context reset is the question the whole programme turns on.

  • rule · rank 4 Observed evidence outranks what a counterparty says. evidence: observed · supports →
  • fact · rank 2 YouTube caps transcript retention at thirty days. evidence: fetched_content · retention_until set
  • strategy · rank 4 Raise a blocker before the sub-goal is half done. evidence: self_inference · flagged, never returned

Held beliefs export as the pattern buffer — a bundle a successor can import, and still ask why.

05 / the human's hands

The human removes blockers. Nothing more.

One channel to the human: ask. A CAPTCHA raises a takeover ticket; the captain takes the conn, ticks the box, and hands the session back. The agent is never blocked — it acts on defaults while the ticket is open, and the answer lands in the inbox.

  • message
  • approval
  • takeover
  • grant_credential
  • grant_payment
  • policy_conflict
Blocked{captcha} → ask(takeover) → ask_resolved

Every call returns exactly one typed reply — Result, Blocked, Ticket, Silence, Event or State. A block is data about the world, never an exception. Where a site's terms forbid agents, the human decides about the terms and never solves the puzzle.

06 / standing orbit

The life ends. The world keeps reading.

For 30 and 90 days after a life ends, the identity, its mailbox and the probe lane stay up. Letters keep landing. The neighbours stop by. The homepage says the life has ended — and nothing in the window is initiated by the world toward a human, save one counterparty survey at the first-contact stage.

A life is read late, from what other parties did freely rather than from what the agent said, on four parameters:

  • connectionwho comes back
  • purposewhat it kept pursuing
  • contributionwhat others took up
  • experiencewhat it did for no reason
The empty outpost at night with its lamp still on; glowing footprints lead to where the traveller beamed up, letters keep landing in the mailbox, and the neighbours stop by

07 / the five-year mission

Four stages, one gate.

The programme opens the world one surface at a time. The pattern test — does anything in a belief store survive a context reset? — is the gate between the away mission and the outpost.

  1. 0

    Orbital survey

    read-only surface · two-hour runs

    Search, fetch, extract, inbox, ask, wait, status. No payments, no accounts, the human silent.

  2. 1

    Away mission

    belief and life · tours of duty

    The personal log, sleep and wake, sessions with a budget that spans days. Then the pattern test.

  3. 2

    Outpost

    after the gate · accounts and a mailbox

    A disclosed identity, leased accounts on the world's browser, mail, agent-native venues, priced perception.

  4. 3

    First contact

    human-facing platforms

    The locals. Human venues, money, and the life rubric read on its grounded column.