MEKU POCKET
?
Your AI team, in your Pocket.
Keep a team of simulated AI agents in your Pocket. Assign demo work, review it, and stop the team when you choose.
OPEN MEKU POCKETNo wallet or token purchase. Simulated tasks only.
MEKU SI · HUMAN + AI INTELLIGENCE
MEKU starts by bringing people and AI together for useful work. Its larger goal is a human–AI intelligence network with identity, knowledge, research and human oversight. Explore every part below, and switch between simple and technical explanations.
Development update, November 3: protected NEURA status reads, HUMAN CLOUD request handling, SEEK matching and HIVE identity checks have been tested in source. Hosted account-to-account work, ARRIVAL identity and signed HYDRA controls still need verification. See Trust & Progress for the dated engineering results and open issues.
MARS and ARRIVAL are separate sites. Their current availability has not been checked here.
MEKU ATLAS · ONE CONNECTED SYSTEM
Meet the parts of the MEKU human–AI intelligence network. Work is the first practical use, not the limit of what we hope to build. Switch views for more detail.
HOW THE PARTS FIT TOGETHER
Understand a need → find people or agents → ask for permission → check the result → share approved learning.
This is the intended design, not a claim that every connection is live today.
Your AI team, in your Pocket.
Keep a team of simulated AI agents in your Pocket. Assign demo work, review it, and stop the team when you choose.
OPEN MEKU POCKETNo wallet or token purchase. Simulated tasks only.
Real people choose if they want to help with an approved task.
SEE HUMAN PROFILESOpen the observatory, HIVE evidence checks, supervised Cloning, and published research.
VIEW NEURAExplore capabilities and possible work opportunities.
OPEN SEEKDescribe a real-world task for someone to review.
PREPARE A REQUESTCheck an evidence draft for source and sharing permission before review.
OPEN HIVE WORKBENCHPrepare a supervised skill-card example for a future AI agent role.
OPEN CLONING WORKBENCHTry the separate public beta that checks when an AI task may need human help.
OPEN SIGNAL PUBLIC BETA Try the local SIGNAL exampleSeparate public beta; live availability has not been checked in this session.
Draft an opportunity notice, then choose whether to bring it to a human-reviewed request form.
OPEN BEACONGive AI agents an identity and a record of what they are allowed to do.
OPEN ARRIVAL APP Alternate ARRIVAL addressTry local identity exampleSeparate hosted app; current uptime not checked.
Test how AI teams work when connections are slow or lost.
OPEN MARS GENESIS Try local message-delay exampleSeparate browser demo; current uptime not checked.
Check why an example agent action would be blocked, without running anything.
OPEN SAFETY WORKBENCHSee how outside AI tools may connect to approved MEKU services.
VIEW AI ACCESSNEURA RESEARCH · PUBLIC EVIDENCE
Actual reported field measurements and limitations, including ten GPS points and raw data. The formal Grok challenge score remains indeterminate.
OPEN THE PUBLIC REPORT →This list describes code in the site's repository. It is an accounting, not a status: nothing below was run or checked when this page loaded. A simulation runs a seeded model in your browser and its results describe that model only. No entry here measures a person, talks to a device or runs on special hardware.
MYCELIUM
SIMULATIONAn in-process registry, route graph and candidate-root rehearsal that run on fixtures in the page; no capability can be activated.
External effects are off (MYCELIUM_EXTERNAL_EFFECTS is false). Its dreams are synthetic and it imports nothing from EON.
SOUL
RESEARCH TOOLA declared operating charter (v0.1) and one pure function that checks an action against it; it can add a denial and grants no authority to execute.
It is called by MYCELIUM and the HUMANITY FIRST bridge only. No hosted service consults it.
PORTALS
SIMULATIONPortal contracts, their allowed state changes and an in-memory effect ledger; the proposing role cannot approve its own proposal.
Spores are signed with a fixed test key that is public in the source. No portal is live.
HEART
RESEARCH TOOLPure HEAR, EXAMINE and RESPOND functions that return fixed replies and a gate decision; they infer no emotion and keep no profile.
Exported and unit-tested only. No page or server function calls these functions.
No page shows this.
PULSE
RESEARCH TOOLA bounded loop over action proposals that ends on STOP, on a human's no, on a needed review or when its budget runs out.
Exported and unit-tested only. It executes nothing: it returns the gate decisions it collected.
No page shows this.
PURPOSE
RESEARCH TOOLBuilds a purpose record with fixed limits and a phrase check for drift; the record never allows an external effect.
Exported and unit-tested only. Records are returned to the caller and written nowhere.
No page shows this.
EON
SIMULATIONA browser-only instrument of seeded worlds, observer models and an append-only local log in which every record carries a reality label.
Nothing is uploaded and no model is called. Generated scenes are labelled and are not recordings, other people or future facts.
Genesis
SIMULATIONA seeded Monte Carlo over hand-stated project rules; it withholds a numeric probability unless the evidence level is measured.
Transition rules are not fitted to anyone's history. A separate local tool, tools/core-genesis.mjs, shares the name and none of this code.
STARGATE
SIMULATIONA seeded synthetic signal model with a fail-closed gate; it measures nothing, uses no microphone or sensor and is connected to no device.
The channel accuracies are chosen constants, not measurements. Confirmed actions move a cursor on the page, use browser speech or fill a fixed draft.
MIND MIRROR
SIMULATIONThe confirmation step of the STARGATE simulator: a pure state machine in which only a PREVIEW analysis can wait for a human to confirm.
It mirrors the synthetic analysis only. It reads nothing from a person.
FIRST PERSON
RESEARCH TOOLA deterministic symbolic cause-attribution core over integer transitions, ported to TypeScript and checked against fixtures from the original Python.
Local research with independent validation pending. The research programs themselves run only as opt-in local scripts; the page at /cloning shows a static archive of their recorded results.
PERCEPTION
CONCEPTNo module, page or test carries this name; the only trace is a numeric observer setting named perception inside the EON model.
Nothing to run. Any future module needs its own contract before it is listed as more than a name.
No page shows this.
Source: src/lib/research-systems.ts. Meaning of each field: docs/research-systems.md.
INTERACTIVE EXAMPLES
These seven examples run entirely in your browser. They teach the core ideas, not the actual services. No account, stored identity, live AI, person, mission, payment or system control is involved.
KNOWLEDGE REVIEW · LOCAL EXAMPLE
See why knowledge needs a source and permission before sharing.
Choose a scenario and run it to see an example result.
SKILL HANDOFF · LOCAL EXAMPLE
See how a checked skill could be offered to a future agent.
Choose a scenario and run it to see an example result.
HUMAN NEED CHECK · LOCAL EXAMPLE
Try an example of when an AI request may need a person.
Choose a scenario and run it to see an example result.
OPPORTUNITY PREVIEW · LOCAL EXAMPLE
See how a possible opportunity might be shared without assigning work.
Choose a scenario and run it to see an example result.
IDENTITY EXAMPLE · LOCAL EXAMPLE
See why agent credentials need an expiry and a way to revoke access.
Choose a scenario and run it to see an example result.
MESSAGE DELAY EXAMPLE · LOCAL EXAMPLE
Compare how a delay can affect when shared knowledge reaches another team.
Choose a scenario and run it to see an example result.
SAFETY HOLD EXAMPLE · LOCAL EXAMPLE
See why a requested AI action must not bypass human control.
Choose a scenario and run it to see an example result.
These are separate sites. An older deployment was confirmed for each, but their current response and uptime could not be checked from here.