Building a New Product on the Shoulders of Five Expired Patents: The Complete Playbook

Ask the question in the other direction

Traditional R&D search asks: "Has anyone done my solution before?" — despair if yes, start building if no.

Combination R&D inverts it: "For each module I need, who has taught how to build it?"

That inversion is worth a lot. A complete product embodies many features — sensing, structure, control, materials — and implementations for most of them were written down in patent literature long ago, a sizeable share now expired. Your product as a whole may be new, but its parts almost all have predecessors: before inventing every module from zero, find out which modules can legitimately stand on someone's shoulders.

The five-step workflow

Step 1: decompose the product goal into a feature list

"An anti-tangle robot-vacuum brush system" decomposes into: a quick-release brush structure, a tangle-load detection mechanism, an untangling actuator, a debris evacuation path… Decompose until each feature independently answers "how would this be implemented." The cleaner the decomposition, the sharper the search hits.

Step 2: search teaching sources per feature, expired first

For each feature, find documents that explicitly teach an implementation. Key disciplines:

• Expired-first is structural, not a preference — of two patents teaching the same feature, the expired one is a public-domain resource while the in-force one is an obstacle to design around or license;
• Teachings must be verifiable — "this patent teaches feature X" must point to a specific passage (specification ¶[0023]); topical relevance is not a teaching;
• Legal status carries its basis — if expiry is estimated, say estimated; the official register is final.

Step 3: derive the combination — which patents together cover which features

The ideal output is a matrix: features as rows, documents as columns, cells reading taught / partial / not taught. When deriving coverage, one under-appreciated principle: cover the most features with the fewest expired documents — concentrated sourcing simplifies both the engineering integration and the provenance trail.

Step 4: handle the two special cell types honestly

• Features taught only by in-force patents: the hot zone. Either design around them (a different implementation route), negotiate a license, or accept that this module must clear a formal FTO before launch — whichever path, the flag stays on.
• Features with no teaching source: these are gaps — the part your own genuine R&D must solve, and usually where the product's real differentiation lives. The dangerous move is letting an AI fabricate a source to fill the hole; the honest move is surfacing the gap and scheduling it into the R&D plan.

Step 5: integration R&D + pre-launch verification

A combination plan is a skeleton, not a product. Inter-module interfaces, parameter adaptation and whole-machine optimization are real engineering — that investment cannot and should not be skipped. After integration, before launch: verify every expiry call against the official register, and run a formal FTO covering all in-force rights the whole product touches.

Two lines not to cross

Line 1: combining to build ≠ patchworking to file. Combining expired teachings into a product is legitimate, encouraged competition. Packaging the same patchwork as a "new invention" and filing it may constitute an abnormal patent application under Chinese regulations. Building and filing are different acts; the freedom of the former does not extend to the latter.

Line 2: clean modules ≠ clean product. Five expired patents free five modules, but the whole machine may have a sixth or seventh feature standing on someone's in-force patent. Expired combination answers "where do the R&D resources come from"; the formal FTO answers "is launch legally safe." The first comes earlier in sequence — the second is not optional.

We turned this workflow into the product: enter a product goal, and it runs feature decomposition, multi-library search with structural expired-first pooling, a teaching matrix with verifiable quotes per cell, combination derivation and honest gap labeling; modules touching in-force teachings carry a permanent formal-FTO flag.

Closing

Combination R&D is not a shortcut trick — it is the correct use of the patent system's output end. Society already paid the monopoly price for these solutions; using them well after expiry is what the system intends. The real work sits at both ends: decompose features cleanly and verify legal status rigorously up front; do the gap R&D properly and the pre-launch FTO completely at the back. The "standing on shoulders" part in the middle is exactly where you least need to be a hero.