Metis's first-anniversary account dates Andromeda's mainnet launch to November 19, 2021. It describes a project concerned with bringing organizations and applications into a usable blockchain environment, including the earlier decentralized autonomous company idea. That history matters because the present AI emphasis did not create Andromeda from nothing. An established execution network, its applications and its community preceded the later strategy. The anniversary's historical usage figures are snapshots supplied by the publisher, rather than current measures of adoption.

The current environment reference identifies Andromeda mainnet as chain 1088 and the Sepolia test environment as chain 59902. They use METIS and test METIS respectively. Applications select a network through its chain ID and RPC configuration, so similar wallet addresses do not make balances interchangeable between the environments. A developer testing a deployment can use the test configuration while keeping mainnet balances and transaction history separate. This account follows the mainnet identified by chain 1088.

The March 14, 2024 launch announcement describes a completed upgrade and the beginning of an alpha phase with two Foundation-run sequencers, operated by Artemis and Enki. That is a more specific historical statement than the announcement's sweeping decentralization slogan. Multiple participating processes were introduced, but the initial operating arrangement was deliberately constrained. The article also describes a temporary upgrade pause. Neither the number of sequencers at launch nor the ambition to expand participation should be presented as a verified current operator count.

Phase two activated on April 23, 2024 at block 16,500,000. The published changes included multiple transactions per block, a transaction pool and sequencer mining. Those are concrete changes to transaction handling and incentives, not just a renamed launch phase. They explain why an older description of one transaction per block becomes misleading.

The operator requirements specify between 20,000 and 100,000 METIS locked on Ethereum, suitable infrastructure and an application reviewed through governance. Holding the minimum balance alone does not make a newcomer an active sequencer. This combines an economic commitment with an admission process. It also separates direct operation from buying a liquid staking product issued by somebody else. Such a product can add its own contracts, withdrawal conditions and operator choices without changing the underlying requirements for an admitted sequencer.

The rotation documentation describes stake-weighted voting power, selection of the active sequencer and coordination through the Metis sequencer set. Failed operation can trigger another selection. The same documentation says downtime penalties are not currently implemented, so a generic proof-of-stake explanation that promises automatic slashing for every outage would be inaccurate here. The meaningful operational questions are which participants are admitted, how rotation proceeds and how failures are handled.

Stake terminology does not answer those questions without the implementation rules.

The MPC design distributes signing work across key shares and includes key generation, resharing and threshold signing. Themis coordinates the participating processes, while local key material is stored encrypted. This avoids describing the mechanism as every operator independently submitting an unrelated Ethereum transaction. It does not prove that the people controlling those shares are economically independent.

A signing protocol specifies how authorized participants cooperate; the independence, maintenance and compromise of their machines remain separate considerations from the cryptographic construction.

The locking design represents sequencer positions through contracts and an NFT with restricted transfer behavior. It includes signer changes and an exit process, with a 21-day period shown in the documented withdrawal flow. This is not an ordinary freely transferable collectible or an immediately redeemable bank balance. Different actions affect the locked position, signing authority and eventual withdrawal. The distinction matters when someone compares direct sequencing with a third-party receipt token whose advertised liquidity depends on its own market or redemption arrangement.

The May 2025 Andromeda upgrade announcement set out a move to Ethereum blobs and a fault-proof system. Older introductory material referring to Memolabs should therefore be read in its historical context, not copied into a current architecture diagram. The announcement itself was a deployment plan with instructions and dates. It explains the intended changes, but the stronger evidence of implementation comes from the subsequent software release and activation announcement.

The maintainers published MVM v0.3.7 on May 13, 2025 with the mainnet fault-proof activation block and blob-related configuration, and Metis subsequently announced activation publicly. These records support describing the blob and proof transition as implemented rather than an indefinitely pending roadmap. They do not make blobs permanent archival storage or remove the need for correct challenge and administrative controls. The release is also an operator artifact: maintaining a compatible client and required endpoints forms part of keeping the system available.

The reviewed DisputeGameFactory implementation restricts its create function to GAME_CREATOR_ROLE. Its separate dispute function allows a bonded request under specified conditions. Those two entry points have different powers. Publicly requesting that a batch be challenged is not the same as permissionlessly creating and completing every dispute game. This code-level distinction limits what the phrase fault proofs establishes by itself. The repository documents an implementation; this review does not claim a byte-for-byte audit of every deployed proxy and its current role assignments.

The state commitment contract likewise checks who may propose batches and who may delete them. Its deletion path resolves the designated fraud verifier, requires the batch to be disputed and enforces the challenge window. Administrative and proof roles therefore remain relevant after a challenge mechanism exists. Reading these checks is more informative than assuming that an Ethereum posting transaction makes every L2 state transition trustless. This source supports the implementation's authorization structure, not an independent certification of all live configuration and key custody.

L2BEAT's current independent analysis identifies restricted challenger participation, a Security Council upgrade path without a protective delay and no ordinary forced-inclusion route. It consequently does not classify the network as having completed the stage of decentralization supporters previously hoped to reach. Those are attributed findings from its monitored configuration, rather than conclusions inferred from Metis branding. They describe material dependencies for users even when data is published on Ethereum and more than one sequencer participates in ordering transactions.

The finality documentation describes mvm_finalizedBlockNumber and agreement after sequencer rotation, with more than two thirds of the relevant voting power. It separately discusses submitting batches to Ethereum. These are different milestones in the transaction lifecycle. A wallet can display a completed local transaction before the full settlement and challenge conditions relevant to an Ethereum withdrawal have elapsed. Calling every one of those milestones final without naming the layer makes applications look simpler at the cost of hiding the actual dependency.

The developer FAQ describes METIS as both the native L2 currency and an asset with an ERC-20-compatible interface, while distinguishing wrapped METIS used by applications. It also documents an execution environment based on Berlin with PUSH0 support and says Cancun is not supported. Developers should test against those stated rules rather than infer complete compatibility from an EVM label. An Ethereum compiler target can introduce unsupported behavior even when ordinary account addresses and familiar wallet tooling appear to work normally.

The October 2025 economic proposal assigned different roles to Andromeda, Hyperion, LazAI, ZKM and GOAT. It explicitly described GOAT as an independent Bitcoin L2 exploration, rather than another name for Andromeda. It also proposed reallocating 20 percent of total METIS supply from an existing mining reserve into an ecosystem growth reserve, with no new issuance or increase in the cap. Reallocation can nevertheless alter when existing tokens enter circulation. An unchanged maximum supply does not mean unchanged incentives or distribution.

Metis's December 3, 2025 retrospective reports approval of the unification program. That establishes a later governance status than the original proposal, but it is not a transaction-by-transaction account of reserve spending. The same report describes AI-related application launches, including CreateAI and LazPad. Their existence belongs to the ecosystem story without making every generated asset reliable, every promised integration complete or every application's activity a cash distribution to METIS holders. Approval, execution and an investor's eventual return are different propositions.

David's April 17, 2026 update reduced the expected sequencer mining reward rate from roughly 15 percent to roughly 10 percent. It attributes the change to average L2 block intervals increasing from about 12 seconds to 17 or 18 seconds while the per-block reward stayed unchanged. The arithmetic is consequential: fewer reward-bearing blocks per year reduce an annualized rate even without a new per-block payout. These are the operator program's reported assumptions, not a guaranteed yield available to every METIS buyer.

The August 2025 v0.3.9 maintenance release records peer-to-peer and debug-tracer fixes and provides separate image versions for the execution client and data transport layer. The notice permits operators to update at their convenience. This was an ordinary maintenance release, not another sequencer launch or a reported theft. Keeping the two service versions explicit helps distinguish the execution process from the component transporting chain data.

The builder-mining documentation describes a reward program that distinguishes transaction-based allocations from special initiatives and includes registration and community selection. The incentive is intended to make application development economically attractive. It also means transaction counts can be influenced by the program itself. A useful evaluation considers what users are doing and why, rather than treating every additional transaction as an independent paying customer.

The documentation's allocation figures describe its stated program, not a guarantee that an unchanged budget will continue indefinitely.

The September 7, 2026 report for August names applications receiving builder rewards, including FeedByHash and cross-chain or trading services. FeedByHash's virtual-pet interactions illustrate why application-level context matters when interpreting a transaction ranking. A frequently used game action and a large settlement transfer can each produce a transaction without representing the same economic activity. This is an official reward report, not an independent audit of organic demand, unique human users or whether the program's long-term benefits exceed its subsidy cost.

The March 2025 optimistic-quorum announcement prompted a disagreement about governance access. Ransome welcomed a potentially easier path for proposals while acknowledging capture concerns; cypherpunk669 objected that a large holder could exert disproportionate blocking power. These were original comments about a particular procedure, not a survey proving a community-wide preference. Their disagreement is useful because speed and participation are not the same goal. A process can reduce routine friction while leaving unresolved who can stop a proposal and at what cost.

A later leadership and voting framework did not pass. On May 1, 2026, Andrei reported that roughly 7,900 METIS voted against it, exceeding the stated 5,000 threshold despite a majority of participating wallets favoring it. Subsequent discussion considered revisions, and nambi asked about quadratic or diminishing voting weights. This record demonstrates an actual veto and continuing disagreement. It does not establish that the proposed leadership registry or replacement voting architecture was subsequently enacted; no such execution is asserted here.

In the January 2025 ReGenesis discussion, Prince_omobee questioned whether another strategic pivot was chasing a trend, while cypherpunk669 imagined AI making technical decisions more accessible to ordinary participants. Mariodpc, describing personal investment in METIS, connected enthusiasm to hoped-for returns but also requested clearer explanations and visible progress. These are distinct motives within one conversation: identity, participation and investment. None proves broad consensus, future price performance or that an AI tool already makes governance more representative.

The economic-unification thread exposed another divide. Ale3074 questioned concentration, reserve transparency, token fragmentation and potential market pressure. Elena answered that the Foundation retained operational responsibility and described a longer-term intention to broaden governance. The exchange gives readers an original account of disagreement about stewardship. A promise that ecosystem growth will support METIS is an investment thesis to evaluate, not a contractual entitlement.

The important missing evidence is often not another slogan, but a clear allocation record and the authority behind it.