Gongkai
Gongkai (公开) is the informal IP-sharing ecosystem of the Shenzhen shanzhai hardware scene, named and described by hardware hacker Andrew “bunnie” Huang in a series of posts culminating in “From Gongkai to Open Source” (Dec 2014). The term is deliberately not the Chinese word for open source (kaiyuan, 开源): kaiyuan refers to openness inside a Western-style IP framework, which gongkai explicitly is not. Gongkai means copyrighted documents — sometimes stamped “confidential” and “proprietary” — being made public and shared overtly, just not according to the letter of the law.1
Broadcast vs. network IP
Huang frames gongkai as one manifestation of a deeper cultural gap in how IP and ownership are conceived:
- The West has a “broadcast” view: good ideas are credited to a clearly specified set of authors/inventors, and society pays them royalties. Rule of law lets IP be amassed over long periods into “impenetrable monopoly positions” — good for incumbents, brutal for upstarts (Western hardware startups “spend more money on lawyers than on tooling”).
- China has a “network” view: the far-sight needed to innovate is attained by standing on the shoulders of others, so ideas circulate as favors within a network. In a system with loose IP attitudes, sharing with the network is individually rational — tomorrow you may need to stand on your friend’s shoulders.
Crucially, gongkai copying is not a one-way flow of value like pirated movies or music. The shared documents are the knowledge base needed to build products around the chip vendor’s silicon, so the sharing promotes chip sales — “a quid-pro-quo between the copyright holders and the copiers.” The vendor’s secrecy is nominal and its tolerance deliberate.2
What it enables
Huang’s canonical datapoint is the **3 in single quantities (at a time when a Western hobbyist paid $6 for a 16 MHz 8-bit AVR with kilobytes of RAM). Because small teams can obtain complete design packages for working phones — case, board, firmware — they can fork the design and innovate only on the parts they care about. Phone models churn seasonally, with constant low-stakes experimentation: cigarette-lighter phones, huge-button phones for the visually impaired, watch form factors.3
The Western contrast: Huang reviewed 100+ hardware-startup business plans and found most building on “overpriced chipsets built using antiquated process technologies” — not because the founders lacked ambition, but because the parallel constraint of scarce documentation and scarce supply for cutting-edge parts funnels everyone onto Arduino, Beaglebone, and Raspberry Pi. His own Novena laptop used a Freescale i.MX6 for the same reason: not the best chip, but the one with downloadable documentation and Digi-Key availability.4
The Fernvale experiment: repatriating gongkai into open source
Huang and collaborator xobs tested whether gongkai IP could be legally absorbed into the Western ecosystem. The result was Fernvale (~2014): a clean-room-ish reverse engineering of the MT6260, releasing CC BY-SA-3.0 hardware (with an Apache 2.0 patent rider) and a BSD-licensed NuttX RTOS port. Their legal theory, worked out with reference to the EFF Coders’ Rights Project:
- Copyright: reverse engineering to understand ideas and achieve interoperability is fair use (Sega v. Accolade, Sony v. Connectix). For pure facts — register addresses, bit definitions — they relied on Feist v. Rural: facts aren’t copyrightable, and a compiler may use another’s facts given original selection and arrangement. They even invented a custom assembler-macro language (“scriptic”) to transform vendor init code into a new expression, guarding against subconscious plagiarism.
- Hygiene: files came from public servers (no CFAA issue), no encryption or access controls were bypassed (no DMCA anticircumvention issue), and no shrink-wrap/click-through EULAs were agreed to (no contract waiver of fair-use rights).
- Patents: essentially punted — the standard open-source practice, since clearance searches are ruinously expensive and familiarity with patents risks willful infringement and triple damages. The license’s patent “poison pill” was the backstop.
The engineering was equally notable: X-raying the $3 chip revealed not one die but a multi-chip module of 4–5 dies — a striking validation of Krste Asanović’s prediction that the future is “system mostly on a chip,” since merging DRAM/flash/analog/RF/digital into one process is uneconomic. The team built an FPGA SPI “ROMulator,” reconstructed the boot chain (SHA-1-checked first-stage bootloader), and attached a debugger by emulating the ARM core in QEMU and reflecting unknown loads/stores to the live chip over a serial shell — bypassing multiplexed JTAG entirely.5
Huang’s normative claim
Huang released Fernvale not primarily as a product but as a rights exercise: “rights tend to atrophy and get squeezed out by competing interests if they are not vigorously exercised.” His formative childhood tinkering, he notes, is “no longer a legal option for the next generation of engineers,” and the shanzhai are the counter-existence-proof: a permissive IP environment spurs grass-roots innovation. The post ends by asking the community whether blob-free firmware is worth continuing for — MediaTek’s own LinkIt ONE (with Seeed Studio) had just offered an officially sanctioned Arduino-shim path to similar silicon, raising the live question of whether API-walled openness is enough.6
Relationship to the patent-econometrics view
Gongkai is the street-level, ethnographic complement to the economist-grade picture in chinese-innovation-patent-ecosystem — and an interesting tension with it. The patent literature (Lerner et al.) finds China’s critical-tech surge is homegrown and university-driven with minimal reliance on imported US know-how after ~2013. Gongkai documents the earlier, grey-market phase of capability accumulation: knowledge diffusing through Shenzhen’s informal networks precisely outside formal IP channels, with the state’s non-enforcement functioning as an implicit industrial policy. Where Ang et al. measure quantity-over-quality gaming of formal targets, gongkai shows the unmeasured substrate: capability built by actors who never filed a patent. The 2014 Huang piece and the post-2015 “internally generated innovation” finding bookend the same arc — the network IP regime as the cultural precondition for the formal surge.
Cross-domain connections
- chinese-innovation-patent-ecosystem — econometric complement/tension (above)
- deepseek — DeepSeek’s open-weights strategy and efficiency-first training is gongkai’s quid-pro-quo logic in AI form: give away the artifacts, sell the ecosystem position. Liang Wenfeng’s stated philosophy (“open source is a cultural behavior, not a commercial one”) echoes the network-IP view almost verbatim.
- institutionally-constrained-technology-adoption — gongkai is the inverse case: institutional non-enforcement as the adoption accelerant rather than the constraint
- norms — the network-IP regime runs on reciprocity norms (favors, standing-on-shoulders) rather than formal law; a working large-scale norm-governed knowledge commons
- mutual-aid — the guild pattern at digital speed: shared designs, mutual training, fierce competition in execution — Kropotkin’s federative principle with supply chains
- division-of-labor — the shanzhai inversion: small teams holding whole designs and specializing by choice on one piece — Smith’s extent-of-the-market explains why Shenzhen’s dense supply web enables it