Blue: an agent that runs tasks on your iPhone

What it does

Blue is a small USB-C dongle and iPhone app that lets a person ask for work to be done inside the apps already on the phone. The YC profile describes tasks such as sending a follow-up, booking an appointment or answering a message, with Blue opening the relevant app and reporting back without requiring the user to touch the phone. The current homepage is sparse and says only that the company is building something that does the next task for you.

The fit is a consumer who has repetitive phone workflows and is comfortable delegating device control to a dedicated hardware/software assistant. The product’s promise is broader than dictation or a voice shortcut: Blue is meant to navigate existing apps and complete actions. That makes the permission, confirmation and failure model the core of the product.

Why I’d look closer

The team combines the right kinds of experience for the hardware/software bet. The YC profile describes Omar Abdelaziz as a former Google Assistant and Scale AI leader with an early LLM computer-use patent, Peter Krogh as a long-time voice-dialog and Gemini Live design lead, and Farhan Hossain as an Apple Vision Pro engineer with robotics, wearables and consumer-electronics experience. Those are public professional backgrounds, not evidence that Blue can safely operate every iPhone app.

The product history also matters. The profile includes an earlier Blue launch for a smart dictation keyboard, then a previous launch for a voice assistant, before the current “run your phone” description. That looks like an evolution toward computer use rather than a reason to merge every past feature into today’s product. No current public pricing, app-store proof or integration list was exposed in the sources checked.

What I’d ask

Which actions can Blue perform without confirmation, and how does a user see or undo a step? What happens with two-factor prompts, private messages, payments, destructive actions or an app redesign? I’d test a low-risk appointment workflow, inspect permissions and logs, and require an explicit approval boundary before delegating anything irreversible.

My editorial take

Blue is interesting because the hardware could give computer-use agents a physical, always-available interface. The buyer question is not whether voice can launch an app. It is whether Blue can complete a multi-step task reliably while keeping the user in control. The current sparse homepage means this should stay a sourced early profile, not a product recommendation.

Quick facts

Field	Sourced detail
Buyer fit	Consumers with repetitive, multi-step iPhone tasks
Product wedge	USB-C dongle plus iPhone app for agentic app actions
Founder context	Google Assistant, Scale AI, Gemini Live and Apple Vision Pro backgrounds
Public pricing	Not exposed in the sources checked

Sources checked

Checked 2026-09-19.

Source	Used for
YC company profile	Product, founders and launch history
Blue homepage	Current public status and sparse product surface

Cohort context

Blue is listed in Summer 2025. In our 2026-09-18 directory snapshot, 12 of 166 listed companies in that cohort have YC’s primary industry label Consumer (7.2%). This is a current-directory comparison, not an original intake count or a performance ranking. Nine-cohort dataset.

Public website snapshot

Observed 2026-09-19T16:17:58.187Z in raw homepage HTML. This records visible metadata and advertised links, not agent execution or product quality.

Signal	Homepage observation
Product description metadata	Observed
Canonical link	Observed
H1 or H2 heading	Observed
Typed structured data	Not observed in this response
Docs/developer link	Not observed in this response
Pricing link	Not observed in this response
llms.txt link	Not observed in this response
Markdown alternate	Not observed in this response

Public observations · Collection method. Missing links here do not establish that a capability or file is absent elsewhere.