This finding about $ONE was the most unexpected in my research this time, and also the one I think is most worth writing separately among the 11 articles. According to publicly reported information, on September 6, 2026, Harmony released two proposals to completely shut down the mainnet that launched in 2019, migrate the native token ONE to Ethereum, and pivot to the AI video “mashup-editing economy” business.

The closure reason provided by the team is: «The threat posed by national-level attackers and AI agents.» This is not an isolated incident. Before this, there was a sequence of events—I traced backward along the timeline and found that before the shutdown proposal, there were three more incidents, all with explicit dates:

On August 14, 2026, Harmony confirmed the first round of a 4 billion ONE token issuance. On August 17, 2026, it released a rollback plan, intending to revert to the block state of August 11.

On August 18, 2026, an attack occurred: 2.385 trillion ONE tokens were illegally minted, and the network rolled back. On September 6, 2026, proposals were used to shut down the mainnet, migrate Ethereum, and transition to AI video. From the first confirmation of anomalous issuance to the proposal to shut down the entire chain—less than a month. Regarding the forged amounts, I need to distinguish three tiers clearly, because online versions often mix them up: the first wave officially confirmed at about 4 billion tokens (about 26% of total supply); the total rebuilt by the official at about 3.01 trillion tokens (involving 6 abnormal blocks); and the amount actually transferred successfully at about 2.385 trillion tokens (477 successful transfers out of 534 attempts; each transfer was 5 billion, all occurring within 106 seconds).

By the way, the numbers circulating online—"23.85 billion," "23.85 trillion," and "300 billion"—are all order-of-magnitude errors. There’s one blank that must be made clear: there is no verified realized loss amount. Two independent security institutions both explicitly ruled out the calculability of this figure. The forged amount is about 200 times the total supply; the exchange-frozen amount has never been disclosed; and a chain rollback cannot undo off-chain transactions that have already occurred.

So that "nominal value of $356 million" cannot be quoted as a loss. The two independent institutions reach the same technical root-cause conclusion: the receipt signature only covers the header, while the marker key used for accounting is taken from two fields that were not signed (shard ID and block number). Change those two fields and you can post the entry again. An aggravating factor was the fix from the hard fork on July 13, 2026: using the "epoch within the receipt’s own header" as the switch, which caused headers before the fork to still follow the weaker validation path.

The specific scope of the chain rollback: Shard 0 was rolled back to block 92,730,034, and Shard 1 was rolled back to block 94,978,278 (the state as of 2026-08-11 23:25:37 UTC). A total of 141,628 blocks were discarded, including 109,126 valid user transactions. Here’s another governance document line I found: the original text of a proposal topic on the official forum was "Change to 2 shards and use 25% of emissions for recovery"—meaning the shard count was reduced from 4 to 2. This neatly explains why the rollback only involves two shards.

The exchange-side gap data: across six exchanges, holdings dropped from 10.234 billion tokens to 6.581 billion ONE. The official emphasized that this is an adjustment in the reconciliation methodology, not newly recovered funds. By platform: Binance fell from 3.248 billion to 2.751 billion; another exchange fell from 3.494 billion to 0.488 billion, the largest drop. Deposits/withdrawals and trading were still paused at the time.

There’s a number jump here that needs to be explained to readers: earlier it was said that 4 billion tokens were minted, but reconciliation shows that 35.91 billion tokens entered exchanges from the hacker wallet. These can only be explained by the idea that the same receipt can be posted repeatedly and that the actual forged amount far exceeds the first wave. Several critical details in the proposal: the multi-signature wallet, the liquidity pool, and on-chain applications cannot migrate. The team urged users to exit all smart contracts by September 10, 2026.

Validators may stop running nodes starting at 22:00 Beijing time on September 10. The migration method is: the network’s last block is used to create a snapshot; in Ethereum, the new ONE tokens are air-dropped to the same wallet addresses, and holders do not need to claim proactively. Delegated staking and unclaimed rewards will be air-dropped to each party’s governance treasury. A one-time compensation pool of $1.372 million will be established, and payments will be made in four quarters to validators (who stop operations on time, sign the agreement, keep their stake, and serve as governance validators for the new project) and their delegates.

The total ONE supply and the token issuance rate remain unchanged; the newly issued tokens are intended for new business. Neither of the two proposals is binding, so plans may still be adjusted. The new business includes: opening prompts and materials so users can do secondary creation; AI agents expanding video stories; and operators need to stake tokens and earn rewards based on online service time. The team plans, subject to meeting conditions, to help operators generate up to $1 million in total revenue in the first year; promoters initially can earn a 30% commission from the $10 monthly subscription they recommend.

On the exchange side, ONE has two tags: Layer1 and Launchpad, and there are no risk tags. Regarding the data, there’s a field issue that needs to be explained: in the exchange metadata, its circulating supply is 15,175,140,288 tokens, while total supply is 15,028,890,000 tokens—circulating supply is greater than total supply. This is a field error in the exchange interface, not a characteristic of the project itself. Third-party data platforms show circulating supply and total supply for ONE that are basically equal, with no inversion.

So I won’t use this number, nor write it as a problem of the project. The market cap is about $43 million to $45.6 million, depending on the snapshot time.

## Today’s

20% drop—I can’t find the reason. I’m going to retract a judgment I originally planned to write. I initially intended to directly attribute today’s drop to the shutdown proposal above—time-wise it would seem logical. But I conducted targeted searches (in five languages: Chinese, English, Japanese, Korean, and Persian, covering September 22 to 24). The result is: within this window, there were no specific events related to ONE—no delisting, no tag changes, no unlocks, no governance vote results.

The broader market was down at the same time (Bitcoin -2.2%, Ethereum -2.6%), but ONE fell 20% to 25%, and the magnitude really is attributable to this token itself. All I can find is context, not cause: one week before the crash it surged by +37% (24 hours) and +85.48%; the chart pattern fits a pullback. The migration terms themselves are uncertain; and deposits/withdrawals on six exchanges were still paused at the time.

But these are all background, not an unverified confirmed reason. So the accurate way to write it is: the cause has not been found in public information. I won’t assert the possibility of “rally then pullback” as fact. A reminder: the proposal is not binding, and this is still in the community proposal stage. Historically, there are precedents where after a proposal was made, it was adjusted due to community opposition. So the accurate phrasing is: "the project team proposed doing this," not "it has already been done."

This article compiles publicly available information and personal viewpoints and does not constitute any investment advice. The event is still unfolding; follow the latest official announcements for updates.