From empty prompt to hardware-ready

An empty prompt stalled new engineers. Guided setup lifted project creation from 43.8% to 74%.

From empty prompt to hardware-ready project cover
Client
HydronAI developer platform
Year
2026
Role
Product Designer
Duration
3 days

Impact

  • 43.8% → 74%Project creation rose across two 30-day windows

At hackathons, people never realised they could add context. They grew frustrated and left for Gemini or ChatGPT. I replaced Hydron’s empty prompt with a guided setup that ends at the first question.

Hydron is an AI assistant for engineers who build hardware and the software inside chips. Before it answers well it must read their sources, such as chip parts, files or code. That reading step is called indexing.

The empty prompt offered nothing to connect and nothing to ask, and analytics showed fewer than half of sign-ups creating a project.

The first run moves from sign-in through hardware and indexing to the first question.

Keeping every route in one list

One multi-select list holds pre-indexed hardware, files and a GitHub repository (an online code folder) or a zipped codebase. Hardware comes pre-checked because its 844 parts are ready in seconds with no indexing wait.

Chose

One guided setup that ends at the first question

Rejected

The empty prompt as the only start

Setup proves Hydron isn't just another LLM wrapper before anyone judges its answers.
Chose

Hardware promoted to the top

Rejected

A default that hides repositories

The fastest route should never obscure the other ones.
  1. 01

    Log in

  2. 02

    Attach hardware

    Space adds a source and Enter commits. One registry part or file is the minimum.

  3. 03

    Name the project

    The name is inferred from the part and Enter accepts it.

  4. 04

    Index

  5. 05

    Ask the first question

Project name field prefilled with stm32h743-bringup, inferred from the selected part

I designed the flow with one engineer in three days and wrote the measurement requirements so the result could be read.

Leaving setup without losing it

Skip sits on its own s key outside the checklist, and an amber note explains what skipping costs. The hydron onboard command later resumes the setup instead of restarting it.

Amber skip confirmation warning that answers will come from general knowledge until a source is indexed

Making the change of screen the reward

Onboarding fills the screen with a centred layout and a large dot-matrix wordmark. The first prompt drops back to normal scrolling so the change itself tells people Hydron is ready.

Chose

Normal scrolling at the first prompt

Rejected

A prompt that stays on the onboarding screen

The switch back to normal scrolling marks the moment Hydron becomes ready.
Chose

One indexing line and a table that collapses

Rejected

Apologies for the wait

Slow sources get a single promise of an email when indexing finishes.
Bare Hydron prompt after onboarding, ready for the first question

Computing answers against your own hardware

Hydron computes each answer from the parts you attached instead of approximating from a similar chip. The AI is only as precise as the parts it has read, so setup leads with hardware.

Turning more sign-ups into projects

The flow shipped on 27 August 2026. Across two 30-day windows, project creation rose from 43.8% to 74%.

Grounding every answer from the first minute

An assistant that starts from an engineer’s own chips answers like a colleague who has read the manual. Every later Hydron feature can assume that grounding is already in place.