Digital Junk Yards for the Win
Next: is dataset size itself a technology?
In design work it’s a regular phenomena that different approaches, technologies and tools are consider but for a variety of reasons they are not chosen. This might be that initial small calculations show that it won’t work fast or affordable enough, the technology perhaps is still immature and the vendor or open source project has overpromised, its too complex, the existing skills do not favour using it, the team has bad past experiences in some respect or just key decision maker prefers something else etc.
In todays fast paced world where agile methods dominate, these key juncture points are never documented: why team took a particular path and not another one. Or even that a particular decision was made with several viable alternatives on the table.
The reasons remain as silent, tacit know-how, regularly be lost when companies shed workforce.
As technology goes on, old bottlenecks become irrelevant. But this information never reaches mature organisations, They stay firm in their belief that their current approach is the only viable one.
There is a way around this problem that has been known forever in software community: architectural logs. It’s a concept of writing down - say in small blog posts - why a particular approach was taken and what other options were considered and reasons for abandonment.
Very rarely if ever implemented as most organisations have their knowledge-management strategy on powerpoint only and no one has time to read old (aka ancient) blog posts anyhow.
But even this fact is bound in time and no longer relevant.
Today if any material is saved in textual, audio or video format, it can be a source material for a chatbot. For example, it’s today possible to have a dialogue of past online meetings and ask machine learning models to make a summary of a past meeting that I did not attend and if I ever get interested, to ask deeper questions. ML models can turn all spoken material to text and use it as source for discovery.
This means that for example all past portfolio and key architectural decision meetings could have their spoken material turned to text automatically and interested parties could ask ML models for summaries. And if a person has any improvement idea, they can ask to a ML model to surface any discussions where this idea has been on the table and ask reasons for rejection. Or event ask how many times the same idea has taken the spin on management’s table. The idea holder can then check if rejection reasons still hold and go for another spin.
Even better if the material is saved in some more concise format for example in a short blog posts or even as separate audio recording just decisions and reasons. Like a private audio diary of the head architect.
It’s like a private digital bin or digital junk yard of all past bad ideas and why they were rejected. Rejection to be challenged with today’s tools.
Anyone with somewhat longer history in any industry knows that same ideas rise again and again at regular pace and that every now and then concepts that are known to be losers start winning.
There’s gold in them junkyard hills.