Prism

Everyone has an opinion about where open source is going. Almost nobody has a dataset.

Prism is an attempt at one. Every year, the exhibitor list from Open Sauce is a snapshot of who is actually building things and what they’re building — not what a survey says people intend to do, not what got funded, but who paid for a table and turned up. Categorise that list consistently, do it again the next year, and the year after, and you start to have something that can answer questions properly.

The questions it’s for

• Which domains are growing, and which are quietly shrinking?
• Do individuals get absorbed into companies, or do companies fragment back into individuals?
• How long does the average exhibitor keep showing up? What’s the survival curve for a small hardware project?
• When something becomes a trend, how many years earlier was it visible on the floor?

None of these are answerable from one year of data. All of them become answerable from five.

Why it’s worth doing badly first

The value of a longitudinal dataset is entirely in its consistency, and consistency is only achievable in hindsight — you can’t know in year one which categories will matter in year six. So the discipline is: capture everything, categorise imperfectly, keep the raw data, and accept that the taxonomy will be revised. A rough record started now beats a perfect one started in three years.

Relationship to Synapse

Synapse reads one year’s floor and maps the connections inside it. Prism reads many years and looks for movement. Same source data, different time axis — and in practice Prism should probably consume Synapse’s categorisation rather than duplicate it.