The 2014 San Antonio Spurs had a word for it: Summertime.

That elusive feeling when the ball never stopped moving, everyone knew exactly where to be, and everyone knew exactly what would happen next. In those moments, five players acted like one organism — acting, without thinking — and then, they won it all.

Every team aspires to this level of chemistry, and it’s hard enough with human beings. Cultivating role clarity, shared context, natural rhythms, and complementary skillsets is often as hard as building a great product itself. The more seasoned the founder, the more they focus on it.

But now, a new type of player has entered the draft class: the agent.

They arrive like rookies: all of the talent, but none of the chemistry. None of the knowledge or soft skills to help them fit in, and actually create net-new value. And so, they create drag. They lose their minds when we switch harnesses, they overspend on the company card, and they fail to pick up on the unique cultural norms that define any given team.

As it turns out, handing over all your docs and emails doesn’t get agents up to speed, because great chemistry isn’t built from the playbook alone. It comes from playing together: watching the game, learning how teammates move, what works, what doesn’t, and getting better with every possession. If we want AI to act like a real teammate, it needs a place to do the same.

Not just a chat window, terminal, or elaborate diagram, but a space.

To watch, to learn, and then, to play.

Meet Oasis

Oasis is a workspace where teams and agents find their flow, and power a cycle of self improvement that spans any model, or any harness.

Why a workspace? Isn’t the point to do more, not just chat more? Well, when so many of our processes and workflows go undocumented, this is the place where AI can learn them. Oasis then turns your daily collaboration into shared memory, process, outcomes, and feedback, so every agent starts sharper and each workflow improves.

With one Shared Memory across your team, you can use any model, try any harness, or test out any agent a colleague has made, and feel confident it can hit the ground running. The context you’ve worked so hard to build carries through, with clear boundaries around who can access it. Because good teammates know what to share, what to keep private, and when to ask.

Over time, fuzzy processes become a lot more deterministic. Repeated work gets cheaper, successful workflows become reusable, and every run improves the next. We do this by connecting effort and spend to projects, tasks, and outcomes, so you can see what worked, what stalled, and what something actually cost. Not just one big, mysterious AI bill.

Why We’re Here

We believe the future of talent comes in two forms: agentic teammates, and actual humans. And so, our mission is to make working with AI feel second nature. But how? By looking to our own nature, of course. The ways we already live, communicate, and collaborate with each other.

Up until now, most AI experiences have been 1:1, transpiring inside little chat windows instead of open workplaces. This works well for personal productivity, but when it comes to teamwork, people are hardwired to trust personas and roles — not one all-knowing orb. With clear roles, we can understand what each teammate is great at, what they need from us, and how to work together without overthinking. And this is just one example of what we study, and then apply.

This is the future we want to build: a place where agents feel less like tools and more like teammates. Where every agent knows its role, carries the context of the team, learns from every cycle, and gets better alongside the people around it.

Then every new agent starts with more chemistry than the last.

Every game makes the team better.

And Summertime comes a little faster.