Give a large ambition a specific first step
Hard-tech roadmaps can be long, but an early investment discussion still needs to understand the immediate step. A robot completing a motion, a chip passing a test or an algorithm working on a fixed dataset can each be valuable. The interpretation depends on the conditions and the risk the team proposes to investigate next.
Jared Friedman's YC hard-tech essay explains how an ambitious project can begin with a smaller demonstration that builds confidence. Jimhang's pilot framework below is an original regional application, not a universal timetable or financing model.
Agree the pilot brief before arriving on site
Imagine a mowing robot pilot. The brief should identify ground conditions, the permitted area and hours, arrangements for human intervention and how incomplete tasks will be logged. Reporting that a lawn was finished leaves too much unexplained: area, slope, weather and staffing can alter the meaning of that result.
Narrow the task until it can produce an interpretable conclusion. Agree an owner, stopping conditions and what happens after a failure. Reducing a pilot's scope does not remove necessary safety validation. This is a planning framework; the responsible, appropriately capable participants must determine the actual operating boundaries.
Ask four different questions of the evidence
The worksheet applies to chip evaluation, robotics deployment or an industrial AI trial. Define relevant measures before observing the results. Preserve versions, raw records and anomalies rather than selecting the best-looking outcome afterward.
| Evidence | Keep | Review question |
|---|---|---|
| Technical performance | Versions, conditions, valid runs and failures | Under which conditions can it be repeated? |
| Site value | Existing and trial workflow records | Whose task improved, and how? |
| Delivery burden | Setup, training, supplies and intervention | What work remains at a second site? |
| Commercial next step | Acceptance owner and payment arrangements | Which decision has an actual commitment? |
Bridge the laboratory and the customer site
A university commercialization project can maintain three related records. Research testing explains the measurement and technical contribution. Engineering records explain manufacturing, interfaces and repeatability. Customer records explain use, acceptance and maintenance. The same result should not be presented as three independent advances.
For a Hong Kong team working with suppliers in Shenzhen or another Greater Bay Area city, assign responsibility for versions and issues. A component substitution, firmware change or assembly revision may change the test conditions. One accountable owner should maintain the version log so the customer demonstration matches the version described in the materials.
Finish with a decision that another person can follow
Report successes, failures and unresolved questions together. Then decide whether to continue in the setting, change the design, try an adjacent setting or pause. A letter of intent may document interest; it does not substitute for technical acceptance, a purchase contract or money received. Keep those events separately dated.
When discussing a semiconductor, AI or robotics project with Jimhang Capital, a reviewable pilot pack can support a deeper conversation than a polished demonstration video alone. Include the task brief, results, issue list and next-stage budget. This connects the long-term ambition with the engineering and customer decisions immediately ahead.
Leave with a concrete work product
- Define the setting, version, acceptance method and stop conditions.
- Preserve technical, customer and delivery evidence together.
- Assign an owner for cross-team test version records.
- Turn the pilot into a decision to continue, change or pause.