Relay, NoOnes, BounceBit: three shutdowns at the service boundary (2026)

Between August 17 and August 23, three startup-scale products crossed a service boundary. Relay.app ended access in stages and refunded annual customers. NoOnes moved from a trading platform to withdrawal-only operation after sanctions cut off essential partners. BounceBit retired its standalone chain after an authorization flaw enabled unauthorized token transfers.

The common question is narrower than "why do startups fail?" What operating signal should a founder have watched before the shutdown memo? Relay's public sources leave the cause undisclosed. NoOnes names a regulatory block. BounceBit supplies a technical failure with a measurable blast radius.

Startup / project	What happened	Root-cause signal supported by the record	Counter-metric to watch
Relay.app	Free access ended August 15; paid access ends September 14. Annual customers receive prorated refunds. 1	Cause undisclosed. The founder's move to Google Chrome creates a platform-timing hypothesis, not a post-mortem. 2 3	Paid retention by workflow and customer type; repeatable acquisition after support costs
NoOnes	Wind-down began August 17; the P2P marketplace closed August 21; the remaining service became withdrawal-only. 4 5	Regulatory blocks: NoOnes says sanctions removed essential partners and caused monitoring providers to classify related wallets and transactions as high risk. 4	Critical-partner availability after compliance constraints; compliance-adjusted cost to serve
BounceBit Chain	The standalone chain was permanently discontinued after an August 19–20 authorization exploit. BB will be reissued on BNB Chain from a pre-attack snapshot. 6 7	Security / engineering failure. The event sits outside the channel's five recurring labels; the authorization path and loss exposure are explicit. 6	Privileged-path test coverage; maximum loss and pause time for each contract

1. Relay.app: the shutdown has a date, the reason does not

The event

Relay.app launched in 2021 as an AI workflow-automation product aimed at becoming "the new Zapier." The product automated business tasks such as document drafting, copyediting, and project-management work. 2

Relay's official notice gives customers a precise exit plan. Free users lost access on August 15, 2026. Paying customers retain access until September 14, 2026, after which their accounts and associated data will be deleted. Existing workflows continue during each customer's transition period. 1

The company disabled new signups and free-to-paid upgrades on July 16. Paid subscriptions were canceled automatically on that date. Annual subscribers receive prorated refunds for unused time through Paddle, and the company says those refunds should arrive within five business days. Relay also gives customers exports for workflows, prompts, run history, tables, and connected-app credentials. 1

TechCrunch reported the shutdown on August 17. Founder and CEO Jacob Bank is returning to Google as vice president of product for Google Chrome, and some Relay colleagues are joining the Chrome team. 2

What the public record supports

Relay's official notice gives no specific reason for closing. Bank's own statement explains why he is excited about Chrome: the browser is a daily workspace, an open platform for the web, and a place where people can work with agents without depending on each site's available APIs or MCP servers. The statement describes the next opportunity; it does not describe a failed Relay metric. 3

The evidence therefore supports cause undisclosed. A platform-timing hypothesis is reasonable to monitor because Bank's explanation places browser-native agents against a product built around connected workflows. The hypothesis remains a hypothesis. The public sources do not establish product mismatch, burn rate, team conflict, or a deliberate strategic sale as Relay's cause.

That boundary matters for founders. A founder's new role can coexist with a healthy acquihire, a strategic retreat, a product-market problem, or several pressures at once. Relay's public record confirms the shutdown and the team's next destination. It leaves the causal diagnosis open.

The counter-metric

A workflow-automation founder should put four numbers beside the signup count:

retained paid workflows, separated by customer size and use case;
successful workflow completions after real integrations, not only runs in a controlled demo;
customer acquisition cost and gross margin after onboarding and support; and
the share of customer value that remains defensible when a browser or platform owner can reproduce the workflow.

Those measures would separate an interesting automation demo from a business that can survive platform convergence. Relay has not published the numbers needed to place its shutdown on either side of that line.

2. NoOnes: a large user base still depends on the rails underneath it

The event

NoOnes announced its wind-down on August 17. The company said it had served more than 2.5 million customers during more than three years of operation. The announcement scheduled a staged closure: swaps, Visa services, crypto off-ramps, the gift-card store, Bitcoin Lightning, partner-program payouts, new account registrations, and most USDT networks would close during the week. 5

The P2P marketplace closed on August 21 at 11:59 PM UTC. After that point, NoOnes became a withdrawal-only platform. The company asked users to withdraw balances immediately and no later than August 23. Bitcoin withdrawals remain available through the regular Bitcoin network, while USDT withdrawals use TRC-20. 4

The help-center notice also describes the customer-resolution mechanics. Users can view balances, withdraw funds, and contact support during the withdrawal-only phase. Accounts placed on hold are to be changed to banned status so affected users can sign in and withdraw available funds. The status change ends trading access while preserving the stated withdrawal path. 4

What the public record supports

NoOnes names the root cause: sanctions. The company says it could not resolve or remove the sanctions affecting the platform. The sanctions cut off essential partners, and blockchain-monitoring providers classified NoOnes-related wallets and transactions as high risk. The company warned that transfers could therefore be delayed, rejected, or restricted. 4

That is a regulatory block, with an operating mechanism attached. The failure did not begin with the user count. It began with the loss of the counterparties and transaction pathways that made the user count usable. A marketplace can have demand and still lose its ability to clear, settle, and move customer funds when critical partners withdraw.

NoOnes' withdrawal instructions show why customer resolution should be separated from the operating diagnosis. The company specifies networks, account states, and a deadline for users. Those details describe how the platform is trying to wind down responsibly. They do not tell founders how much the sanctions response cost, which partner failed first, or whether the company had a viable alternative rail.

The counter-metric

A regulated-finance founder should track the dependency chain underneath the headline customer number:

the percentage of critical partners that remain available under the worst plausible sanctions scenario;
the time and cost required to clear high-risk transactions at expected volume;
the share of customer balances exposed to one network, custodian, monitoring provider, or off-ramp; and
a pre-set wind-down trigger tied to service availability rather than registrations.

NoOnes' more than 2.5 million customers describe reach. Partner availability describes whether that reach can still transact. The second number became decisive first.

3. BounceBit Chain: one authorization flaw can retire the chain around it

The event

BounceBit's standalone chain was permanently discontinued after an authorization exploit that began on August 19 and continued into August 20. CryptoNinjas reported the shutdown on August 22. The attacker transferred 286,543,148 BB, or about 286.54 million tokens, from nine mainnet accounts across roughly 14 transactions. The incident lasted about four hours. 6

The reported flaw sat in an Evmos-stack lockup and vesting function. A caller could designate an arbitrary account as the funding source because the authorization check used the wrong principal. The path allowed BB transfers without permission from the account owners. The report says private keys, signatures, wallets, hardware devices, and exchange accounts were not compromised. 6

BounceBit's official updates describe the response. The team paused nodes while deploying a fix and said the issue was limited to BounceBit Chain; the CeDeFi app, smart contracts, and vaults were not affected. The team later announced that BB would be reissued as a BEP-20 token on BNB Chain, with balances set 1:1 against a pre-attack snapshot. 7

The snapshot was taken at block 20,697,260, timestamped August 19 at 21:02:35 UTC, before the first unauthorized transfer. The standalone chain halted at block 20,702,857 on August 20 at 02:36:37 UTC. The reported plan excludes the 286,543,148 BB moved during the attack and reverses transactions made between the snapshot and the halt. 6

What the public record supports

BounceBit supplies a direct technical cause: an authorization failure created a transfer path with a large, concentrated blast radius. This is a security / engineering failure, which sits outside the channel's five recurring labels. The available evidence is strong enough to keep the technical label and weak enough to avoid forcing it into product mismatch, team conflict, burn rate, regulatory blocks, or mistimed market.

The chain's migration plan also separates two outcomes that are easy to merge. BB can be reissued on BNB Chain while the standalone BounceBit Chain is retired. The token's continuation does not show that the original infrastructure remained viable. It shows that the team chose a new execution environment after the incident. 6 7

The public reports give the snapshot and reversal logic. They leave the claim process, redemption deadline, conversion mechanics beyond the 1:1 snapshot, and distribution timeline undisclosed. Those missing details belong to customer-resolution risk, while the authorization flaw belongs to the root-cause analysis.

The counter-metric

An infrastructure founder should measure the loss a single privileged path can create before measuring throughput:

test coverage for every authorization invariant in privileged transfer functions;
the maximum value and number of accounts exposed to one contract or permission class;
the time from detection to a reliable pause, plus the amount recoverable from a pre-incident snapshot; and
independent review of forked or reused protocol components before those components handle value.

BounceBit's incident supplied a visible loss amount and a visible account count. Those are post-failure measurements. The pre-failure metric is the blast radius that the system allowed one authorization mistake to reach.

Three counter-metrics to carry into next week

Paid retention: Relay's closure gives founders a reason to separate trial interest from retained, paid workflows that survive platform changes.
Compliance-adjusted availability: NoOnes shows that registrations and users matter only while the partners and transaction rails remain usable.
Privilege-path blast radius: BounceBit shows why an authorization test should include the maximum value and account set that one mistake can reach.

The point is to identify the metric that would have made continuation—or a disciplined stop—the easier decision before the shutdown announcement. The public record will often reveal that metric before it reveals a satisfying post-mortem.