The Sideshow
How many projects are you thinking about right now? I don’t know if I could answer for myself. To be really honest I might need to sit down with a spreadsheet for a bit. Actually, I wonder if Claude could build something to keep track?
Individual productivity is on an absolute tear and it’s marvelous; it feels fantastic. Every improvement, every incremental gain, every dissolved drudgery grows the horizon (and my backlog) and makes everything feel possible all at once. It’s also scary and exhausting. Most importantly it’s tactile. It’s in front of us and it’s malleable and we can experiment and play and reshape our agent and our prompts and our workflows to our heart’s content and that’s all satisfying and useful and genuinely productive.
So why are you so focused on it? These new powers feel good and useful, but how many projects are you engaging with now? Is your plan to finish each and every one? To rev the dynamo so fast that prioritization doesn’t matter anymore? We can finally stop worrying about cross-cutting concerns and rationalizing resource allocation and avoiding inefficiencies because our individual engines are, or will soon be, limitless. Right?
A short lifetime ago the transformer architecture gave rise to a new era of large language models that catapulted us into the unknown. The sheer speed of that transformation left our thinking poisoned with old patterns. You can hear them bleed out of conversations happening across organizations everywhere. Now that we move so fast, it’s cheaper to prototype. Now that we’re so individually productive, things we used to worry about don’t matter anymore. Now that we have these new toys, we’re having so much fun focusing on our individual workspaces that we don’t need to worry about how decisions are being made at scale anymore; we’ll just do everything.
All those poisons are absolutely and utterly true, as long as you’re comparing yourself to other players that have yet to adopt the individual productivity gains. As long as you’re comparing yourself to last year, to what used to be, then you really can sit on a new rocket-powered LLM-piloted development environment and follow your muse. You’ll be so much faster than you used to be that you can do all the projects you would have done without it, probably with capacity to spare. As long as someone exists with a lunch to eat and a staunch refusal to get on board with the new tools, you won’t go hungry.
Everyone is moving at different speeds right now, but we all know what the table stakes are; agents are here to stay. Some folk are looking past the speed boost and laying the foundation to move better. While you’re celebrating how fast you can swim, the really dangerous players are mapping the ocean.
The context lake is the map; or at least my best attempt to give the map a name. This map is built from the mass of your organization’s context, structured and made legible by LLMs combined with state of the art modeling and retrieval. We’re not going to repeat our previous naive mistakes that hobbled data lakes for so long; there’s more structure here than dumping everything into the mere and praying to the gods of compute for insight.
Think about the first time you realized LLMs were the real deal. This interesting and amusing toy had been sitting on your desk until that moment when you came back to it out of curiosity and it genuinely surprised you. How long until you’re surfacing strategic insight and making better decisions instead of having an interesting chat?