Era’s $11M bet: the software layer AI gadgets were missing

Era is betting that AI gadgets will not converge on one dominant device the way smartphones did. Instead, it expects a spread of rings, pendants, glasses, speakers, and oddball objects that all need a common intelligence layer to route requests, manage models, and survive real-world connectivity constraints ^[1][2][3]. The company’s New York showcase for artists, where developers built a France trivia souvenir, a stock-checking phone-like device, and an air-quality gadget, was meant to show that the platform can support playful but functional experiments ^[1][3][4].

That framing is more ambitious than a typical wearables launch. CEO Liz Dorman said the point is to replace the app layer with an intelligence layer that lets anyone create intelligent objects and devices, rather than forcing people into a fixed handset-centric model ^[1][3]. In practical terms, Era is arguing that the next interface shift will be defined less by screens and more by who controls the device’s behavior, memory, and voice. That is why the startup’s pitch mixes product philosophy with infrastructure language: it is not selling a gadget, but the logic that lets many gadgets feel coherent.

Technically, Era is building the sort of abstraction layer hardware teams usually avoid until they have to. The platform already exposes more than 130 LLMs from more than 14 providers, giving makers room to tune for latency, cost, task fit, and form factor whether the device is glasses, jewelry, or a home speaker ^[1][3]. Era says the same software layer can also add intelligence to legacy devices such as headphones and support custom voice creation, which broadens the market beyond purpose-built wearables.

Investor Casey Caruso said the platform stands out because it dynamically routes across models while handling real-world constraints like connectivity ^[1][3]. That detail matters because AI gadgets are likely to live in battery-limited, intermittently connected environments where the best model on a benchmark is not always the model that can run now. Era’s claim that the stack can scale to millions of devices and support custom experiments for brands suggests it is thinking like a runtime platform, not a niche SDK. The implication is that reliability, graceful degradation, and choice architecture may become the core product features that determine whether AI hardware feels magical or merely fragile.

Era’s funding round also says something about where smart money sees leverage in AI hardware. The startup has raised $11 million in total, including a $9 million seed led by Abstract Ventures and BoxGroup with participation from Collaborative Fund and Mozilla Ventures, after an earlier $2 million pre-seed from Topology Ventures and Betaworks ^[1][3]. The angel roster, which includes Caterina Fake, Ken Kocienda, Tony Wang, Daniel Kuntz, Mina Fahmi, ShaoBo Z, and Kelin Zhang, looks like a cluster of people who have seen both consumer-device promise and platform pain up close.

The founders’ backgrounds make the software-first strategy easier to believe. Dorman worked on AI orchestration at Humane before moving to HP after its acquisition, CTO Alex Ollman worked on agentic frameworks for enterprises at HP, and CPO Megan Gole came through Sutter Hill Ventures’ work on the Jony Ive and Sam Altman io project before joining Era ^[1][3]. That combination matters because AI hardware has been plagued by glamorous demos and weak developer tooling; Era’s team appears to be trying to solve the less visible problems that sink products after launch. If they succeed, their advantage will come from understanding the operational friction behind the headlines, not from shipping a prettier shell.

The broader market context makes Era’s bet feel timely. Humane’s AI Pin became a cautionary tale after criticism and returns, while Meta’s Ray-Ban glasses have found more traction by leaning on a familiar form factor and narrower use cases, reportedly crossing a million units ^[2][3]. At the same time, Plaud has found a niche in meeting-note capture, and startups are still testing pendants, rings, and other hybrids that sit between jewelry and computers ^[2][3]. The category is clearly active, but it has not yet produced a shared platform standard.

That fragmentation is exactly what Era wants to monetize. The company says it plans to make the platform available to the open-source and maker community, which is a classic ecosystem play: seed experimentation first, then become the default runtime when a real market emerges ^[1][3]. For builders, the lesson is that AI hardware is not one market but many awkwardly different products with different interaction models, privacy expectations, and support burdens. If Era is right, the winning strategy in AI gadgets will look less like designing a single hit device and more like owning the software that lets many devices share intelligence, memory, and trust ^[1][4].