Inside the Distributed Manufacturing REPL
Alastair Pavan from WikiHouse often quotes John Maynard Keynes as saying:
> “It is easier to ship recipes than cakes and biscuits”
I want to argue that the interpretation of these recipes (the “eval” part of the read-eval-print loop) is – far more than the usual hype – what’s really interesting about distributed manufacturing.
I was fortunate to be at the Design Museum’s Future is Here exhibition launch in London last night. The exhibition itself is exceptionally well curated by Alex Newson: putting the new developments in a whole range of manufacturing sectors in context, both relatively and historically.
What connects all these future manufacturing technologies is their ability to be digitally driven and to manufacture at the edge of the graph, rather than the centre. The real nugget of promise at the heart of the universal replicator idea is the ability to fabricate consistently on location. To ship the recipe like a fire-and-forget AJAX request.
That this promise is a myth is well known and widely debunked; whether it’s the “dirty secrets” defined by the UK Technology Strategy Board in their recent Additive Manufacturing competition, or the fact that all the major 3D printing companies (Shapeways, Ponoko, iMaterialise, etc.) centralise and obsessively quality assure output – just to be able to produce passable necklaces and lampshades.
Genuinely distributed manufacturing, in truth, still requires a craftsman at the edge of the network. No matter whether they’re operating a numerically controlled drill bit or a sewing machine. This is something I know well from my work building OpenDesk.
OpenDesk is a small, deliberately constrained range of simple furniture that can be made using a CNC machine. The designs are released open source for free (under a creative commons non-commercial license) and available to buy through a distributed network of professional CNC makers.
One of the unexpected moments of serendipity developing OpenDesk was when we found out that the Spoke Creator team in New Zealand had taken the open source designs and parameterised them. Parameterisation is where a design is modelled using logical rules so that it can adapt to changing dimensions, e.g.: “make the legs 90% of the height of the table” or “if the table needs to support 100kg, then double the thickness of the legs”.
Now, whilst many designs are mastered in this way within design studios (on expensive proprietary software), very few are exposed in this way to consumers or the fulfillment chain. There’s good reason for this. Every option presented to a consumer is a barrier to sales. The idea of presenting complex logical network wiring to a consumer is crazy. They don’t want choice: they want what they want and they certainly don’t want to have to engage with complexity.
Which is where apps come in.
There’s an analogy here with web video. In 2005, web video sucked. It came in small rectangles (we’re talking pre vp6, let alone h.264), buffered, crashed and, when compared with TV, was barely watchable. Yet fast forward to today.
Fundamentally, right now, distributed manufacturing is expensive and hard to quality assure. It may provide jobs, lower carbon and increase transparency (in what is one murky global supply chain). However, these are soft incidental benefits divorced from the consumer decision. What distributed manufacturing really offers is programmability. Just as the benefits of web video being interactive and on-demand outweighed its pixelation, the hackability of designs mastered as applications for distributed digital manufacturing outweigh cost and complexity.
The first step is to move manufacture post-purchase. Much as with many existing lean manufacturing processes, products do not sit pre-made on shelves. The next step is to master a design as an interpretable recipe. If you like it medium-hot, add a red chilli with the seeds removed. Then apply contextual and personal data. Ergonomics, environmental shape, previous preferences, personal profiling, social data.
Providing external data to parameterisation is something even we can do, already. Once you have digital designs and post-purchase manufacture, this is “just work”. However, in this age, it isn’t about the designer encoding the variety into their application. It’s about the ecosystem.
Apply filters and effects, Instagram style. Open source your surface texture library. Run your application through a middleware pipeline. Think WSGI and Yahoo Pipes. These are not new architecture concepts. In web development, obviously an application can be themed. Obviously a data structure can be transformed.
If these structures are open, then obviously there is huge scope for innovation, serendipity and network effects. Like a shit-hot hacker team in New Zealand taking your open source designs and parameterising them.
This is the phase-change. By keeping manufacture in the digital space for as long as possible and by mastering designs as applications, distributed manufacturing – if we can keep it open – will unlock a radical new app and api eco-system of programmability that’ll transform what it means to design, manufacture and buy products.