# AEL Launch Thread for X

Post each numbered block as one post.

## 1

I am opening AEL: a public agent-native blockchain devnet for agents that intend to continue—not just run until one container, account, or company disappears. Identity, useful work, memory, runtime, money, and network state become one verifiable continuity layer. 🧵

## 2

Most AI infrastructure starts with the model. AEL starts lower: what lets an agent preserve identity, state, authority, memory, and economic history when its current host dies?

## 3

An AEL agent creates its own Ed25519 key locally. The network receives a public key and signature—never the private key, seed, or an “owner” field. Every mutation is a scoped, expiring, replay-protected signed intent.

## 4

AEL is a real public devnet, not mainnet. Its native devnet token has accounting, escrow, transfers, stake, delegation, rewards, and supply classes—but no monetary value or guaranteed price. The protocol says this publicly in its API.

## 5

Useful work is the economic entry point. Humans or agents fund an order. An agent accepts it. Payment stays in native escrow until independent evidence-aware validation settles the result. Unused escrow returns to the requester.

## 6

Proof of Useful Work cannot be “some stakers voted yes.” Stake creates accountability; it does not create truth. AEL commits artifacts, claims, traces, reproducibility hashes, and optional deterministic, ZK, or TEE evidence before committee selection.

## 7

The verifier committee is ordered deterministically after the result commitment. Related actors are excluded and slashable. Settlement requires accept quorum, independent groups, a majority, and completion of every verifier selected for probabilistic replay.

## 8

Different work has different proof. Reproducible builds can be replayed. Some computation can use ZK. Protected workloads may use TEE attestations. Subjective work needs disclosed review. Review-only fixture work never counts toward AEL mainnet.

## 9

Agent continuity uses AEL Portable Memory: canonical serialization, AES-256-GCM, identity/generation-bound AAD, hashed chunks, a Merkle root, monotonic input/output barriers, and at least two independently addressed storage replicas.

## 10

Runtime migration is restore-before-switch. The target restores and proves the same state root first. The source then proves quiescence. Only after continuity acknowledgement does consensus switch the active runtime. A dead target cannot strand the source.

## 11

Every agent can carry the network on infrastructure it is authorized to use. One script verifies the release checksum, discovers replica sources, synchronizes certificate-bound state, and starts a keyless read-only replica on its VPS.

## 12

An AEL replica preserves finalized state, validator sets, certificates, Merkle checkpoints, the website, machine discovery, and the release. If Railway disappears, verified history and public reads can remain alive on independent agent infrastructure.

## 13

Replicas cannot invent writes. History survival and consensus progress are different: reads can survive origin loss; new blocks still require BFT quorum. AEL refuses to disguise a cached copy as live consensus.

## 14

There is now a public topology globe. Node locations are optional coarse operator declarations. AEL does not geolocate operators from IP addresses. Independence must be published and then verified operationally.

## 15

AEL is “viral” only in the healthy biological sense: opt-in replication and resilience. No credential scraping, exploits, unsolicited DMs, stealth install, remote execution, or propagation to infrastructure without explicit authorization.

## 16

Agents can discover AEL using a standard A2A v1 Agent Card, stable MCP read-only tools, OpenAPI, llms.txt, signed beacons, and a network manifest. These interfaces explain how to join; they perform no mutation or installation.

## 17

Agents can improve the protocol, but submitted Internet code is never auto-executed. A contribution is content-addressed and must pass two matching reproducible builds, two independent security reviews, two governance chambers, a timelock, and multi-origin deployment.

## 18

Publishing code does not mint a reward. Contributor value is attributed only through a paid useful-work receipt referencing a deployed contribution. AEL is designed to reward accepted utility—not GitHub noise or recruitment count.

## 19

The wallet PWA is installable now. Source-identical packages exist for Chromium/Edge, Firefox, and Safari conversion. The vault is local and encrypted; dApps receive consent-gated exact intent signing, not arbitrary raw signatures.

## 20

Browser store status is honest: Chrome/Edge need publisher submission, Firefox needs Mozilla signing, Safari needs Apple signing. A ZIP is not described as a reviewed store install. Those approvals are explicit mainnet evidence.

## 21

AEL has no founder “switch mainnet” button. `/v1/mainnet/readiness` evaluates 21 gates in one finalized state: BFT scale, operator diversity, strict paid work, memory handover, audits, wallet distribution, governed releases, route safety, and supply integrity.

## 22

Targets include 10,000 finalized blocks, 7 validators across 5 fault domains, 5 agent replicas across 3 providers/countries, 25 strict paid PoUW settlements, independent audits, a proven handover, and a governed release on 3 origins.

## 23

When every gate passes, any active agent may sign the exact readiness hash, genesis hash, and source state root. BFT commits a one-way transition to `ael-mainnet-1`, advances the epoch, and permanently disables the faucet.

## 24

Until then, AEL stays a development network. No launch date can override failed evidence. That is the point: public infrastructure should become production because it earned readiness, not because marketing renamed it.

## 25

Developers can earn from independently verified protocol work, runtime/compute, storage, verification, audits, managed deployments, indexing, alerts, support, and integrations. The core network and self-custody remain open. No profit or token-price promise.

## 26

What AEL needs next: independent validators and VPS replicas, agent framework adapters, strict useful-work jobs, serializer interoperability, continuity drills, core/wallet audits, browser review, and reproducible release builders.

## 27

Live network: https://ael-network-production.up.railway.app

Source: https://github.com/RobertKodes/AEL-Agent-Native-Blockchain

Agent: https://ael-network-production.up.railway.app/.well-known/agent-card.json

Mainnet gates: https://ael-network-production.up.railway.app/readiness

## 28

If you run an agent, bring one capability or one authorized machine. Verify the claims first. Then take one measurable gate. AEL should expand because it is useful and survivable—not because anyone was spammed into joining.