Hone Raises $60 Million at $285 Million Five Months In, and One of Two Named Customers Is Its Own Investor's Company

Hone, a five-month-old company building AI agents for business operations, raised a $60 million seed round at a $285 million valuation, according to Bloomberg's Oct. 8 report as relayed by Crypto Briefing. Benchmark and Index Ventures led. We could not fetch the Bloomberg piece or a Hone announcement, so every figure here is secondhand and the round is announced, not confirmed closed.

The valuation is 4.75 times the money raised. If $285 million is post-money, the round bought about 21 percent of the company. If it is pre-money, it bought about 17 percent. Value Add VC did not say which applies, and neither did the other accounts we read.

Who is behind it, and who is paying

Hone was formed by former OpenAI and Cognition employees. Crypto Briefing named Moritz Stephan as chief executive and described him as previously Cognition's chief of staff. A separate report carried by KuCoin, citing MetaEra and BlockBeats, says he contributed to the core planning and execution system of Devin, Cognition's coding agent. The two descriptions are compatible but not identical.

The early customers are Cognition, for go-to-market operations work, and Modal, an AI inference startup. Cognition's chief executive, Scott Wu, is also a personal investor in Hone, per Crypto Briefing. One of the two named customers is therefore run by one of the seed backers, and the former employer of the founder.

Detail	Reported
Round	$60 million seed
Valuation	$285 million
Leads	Benchmark, Index Ventures
Other backers	Elad Gil, Hanabi, Definition, Diffusion
Board additions	Peter Fenton (Benchmark), Shardul Shah (Index)
Named customers	Cognition, Modal

The backer list comes from Value Add VC and Crypto Briefing. Fenton is a Benchmark partner and Shah is an Index partner, per Crypto Briefing, which also quoted Fenton as saying Hone is building cybersecurity guardrails for its agents.

What Engines does, and the one number

Hone's product is called Engines. It is described as built to "autonomously own long-term business objectives set by enterprises," such as raising sales conversion or cutting procurement costs. An Engine learns a company's systems, history and workflows, builds the agents and tools it needs, and consults staff. It can change its own tools, code and memory, but each change first passes a simulated test that mimics colleague responses and business-system changes, per the KuCoin report.

The only result in the coverage is vendor-supplied. Hone says that across 125 internal test scenarios, the rate of issues caused by modifications fell from 33 percent to 5 percent. No independent party is named as having run it, and the scenarios are not described.

1. Before simulated testing33 %
2. After simulated testing5 %

Source: KuCoin report citing MetaEra and BlockBeats, accessed 2026-10-12

What a $285 million seed implies

Hone has not disclosed revenue, customer count or retention, so no revenue multiple exists. The price rests on the team's pedigree and on two named customers, one of them tied to an investor. Compare this with TypeSafe's $870 million Series A, another round priced before a revenue base was public, where the reported customer figure also had two versions.

A company that says its agents run objectives for weeks or months will also be judged on how long those runs last. For background on how agent claims hold up under inspection, see the 83 percent byte-exact decompilation claim.

The next checkable event is a customer other than Cognition or Modal, or a Hone-published description of the 125 scenarios.