The Race
I like safe bets. I like that little psychic click when you knock over the first domino and walk away because you don’t have time to sit there and watch them fall, but you’re damn sure the pieces are falling into place. For some unfathomable reason nobody seems to care about my preferences though. They’re always showing up with opinions and agendas and introducing uncertainty and unknowns and generally just making a big mess of things. Sometimes I like to just wait a bit and see how the situation evolves before making any decisions.
It feels easy. It takes very little energy. Sometimes the payoff is huge, when you discover something that changes everything and you entirely save yourself the pain of starting off in the completely wrong direction. I know I’m telegraphing and you saw this coming a mile away, but you don’t get payoffs without decisions. Waiting is a decision; it’s the active bet that your chance of moving in the right direction is so low that it’s better to not move at all. Waiting to build your context lake is the wrong decision.
The environment around this decision isn’t uncertain. It’s noisy, it’s chaotic. There’s so many things quickly happening all at once; it’s easy to miss the guy in the gorilla suit. The noise is specifically because we’re not uncertain about LLMs anymore and they’re permeating everything. Buried in that chaos is where you find the combination of tools to dramatically improve the consistent quality of organizational decision making. The search, the cost, the novelty all mean that for now most folks are clustered together on this path, even if you haven’t started. Only the really lucky few are going to find out they were on the optimal path by accident. Waiting for them to share what they’ve found in a place you can see is the wrong decision. You are not going to step out to the market and grab a context lake off the shelf. Being on the less-right path with a foundation beneath you and tools to iterate on will keep you a hell of a lot closer to the front of the race than the starting line.
Everything is always a gamble. Technically anything is possible, etcetera, etc. Some bets are still bad bets. By all means, sit at that blackjack table and stay on your five. Let everyone know that you also like to live dangerously; making the wrong choice sends a loud signal. Waiting to see how other organizations build their context lake is betting against the combination of Large Language Models and four decades of supporting satellite Computer Science technologies coming together in a way that just makes sense. Building your context lake is about protecting your seat at the table and preserving your ability to play the game.
There’s always an uncertain future if you look far enough ahead. Over the past five years we’ve built abstract cognition out of probabilistic next token prediction in high dimensional latent space. Forget about pointing at Large Language Models and generically waving our hands about scaling and emergent behavior though. Think specifically about the risk of formalized decision making becoming computationally tractable. An artificial version of the process in the Human mind that we take for granted, as uniquely ineffable as the spontaneous production of natural language was just a moment ago. I don’t know the substrate it might show up in, I’m not that smart. I do know that formalized decision making on top of a healthy context lake wouldn’t leave room for more than a footnote about this thrilling era of individual productivity that we’re so enmeshed with today.