software-shaped knowledge work

so much (but not all ofc) of modern knowledge work is just a set of "software-shaped problems". i like this definition of software-shaped-problems (SSP):

SSP = business challenges with structured inputs, verifiable outputs, and defined rules.

this is why software already ate the world. no matter whether you're in marketing, finance, engineering or sales, software already helps you do your job better and faster. outside of purely creative roles (although you could argue this point), there is essentially no role or function in the modern knowledge economy where "better" software shouldn't (in theory) lead to better outcomes.

so, it should be no surprise that as the cost of generating software plummets, we look to design and build net-new software to solve these software-shaped problems in a more on-demand, focused and fit-for-purpose manner. not everyone agrees here (yet), but it seems to me that in the future, people will increasingly want tools built for their specific use cases, not cookie-cutter, mass-produced software that solves for the average of all use cases.

one way to think about this is "just in time" apps. Patrick Collison (Stripe CEO) called this "oven fresh" rather than "freeze dried" on TBPN the other day which I thought was an interesting analogy.

in some ways, this feels kind of obvious. but there are some interesting trends worth pondering:

- organisational config as code = this has been a thing in software for a long time (terraform etc), but its interesting to see more and more functions start to look to work this way. encoding business rules, domain knowledge, processes and workflows in code means you get interoperability (no lock in to a specific product), audit-readiness (easily review, manage and edit) and reliability (avoid "human"-esque drift, predictable) out of the box.

- database + model is the new spreadsheet (and app?) = we used to flatten complexity into rows and columns because that’s the most the spreadsheet tool could handle. and remember, most SaaS is just a database + a CRUD tool for interfacing with that data. when a model can sit on top of a database, the UI doesn't have to be fixed or static. maybe every user has a unique interface?

- context as the moat = raw product execution used to the the barrier, i.e. they who build the "best" product the fastest win. maybe the future is who can model their domain experience into code the fastest? if the UI is ephemeral (see above point) and all that matters is the data that a model/agent can sit on top of, having the best (and most open) data wins.

i know it's hard to keep perspective when you're in the middle of it, but it does feel like we are living through one of the most important and interesting moments in technology history.