Introducing Workroom

Today, I would like to tell you about the project that I've been working on at Loam.

It's called Workroom. Workroom is an agentic productivity suite.

Background

We're all familiar with apps for docs, spreadsheets, presentations. Depending on our job, we might also use design tools, code editors, CAD tools, etc. Productivity tools are the tools we use every day to do our jobs.

AI has become incredibly capable. Let's define a few terms so we can all be on the same page. An LLM, or model, is a collection of numbers that are used to define the AI's brain. Data, mostly text, is input into the model and the model computes, or infers a response. Agents are bits of code that facilitate this input and output of text. The model can tell the agent to use a tool to create a file or open a web page. If it can be written as software, you can give an agent almost any kind of tool you can imagine.

An agentic productivity suite is the combination of the productivity tools we all know and love (to hate?) and an AI agent.

Can't I already use AI in most of my tools? Isn't that what those chatboxes are about?

You can and you should. What's important to know is that most of the tools we know were not built with the capabilities of agents in mind. Agents have been added after the fact. In many cases, we're working with concepts developed in the 70s and 80s dressed to match the latest macOS or Windows version.

Past Assumptions

We're dealing with assumptions that were made 40 or 50 years ago. Now, just because they are old doesn't mean they are wrong. What it means is that they were built for a different time, a time when computers were unable to understand as much. Most of the software we use today is a combination of a way to arrange data and tools to transform that data. With AI, we have a new layer. Rather than having to say "draw a circle. Make it red." we can say "draw me a ball" and the computer understands what a ball is and draws the appropriate shapes. That a computer can understand our intent and translate it into operations changes the game completely.

And with the advent of AI and agents, we have the opportunity to reimagine how we interact with our computer tools. Where we used to have to be incredibly specific, computers can now handle our natural ambiguity. With agents, we can even delegate tasks to the computer and it will do the work without us outlining every single step. We can express a goal and it will come back a little while later with the result to show us. All that busywork we used to dread can be offloaded to an agent.

This makes it a perfect time to ask ourselves the most human of questions: Are my tools serving me? How would I change them so I can focus on the tasks I care about most?

First Principles

Exploring answers to these questions is what Workroom is about. We want to get to the root, to the first principles of how people work and how their tools need to be shaped in an era where your tools can actually understand you.

More than reimagine our tools in a world with agents, we can tackle problems that were previously inaccessible for the amount of effort they took to build. Part of the reason your document software hasn't been reinvented is because it's been insanely expensive to build what you have today. Trying new things took too long and cost too much money. With agents who can code, the cost of that exploration has flipped upside down. It has gone from being insanely expensive to nearly free.

This means we can build tools that better match with how the human body and mind work. We can do more to support how we think, how we create. Instead of it taking decades and hundreds of millions of dollars, we can do it for far less.

There is a significant piece of the creative process and the thinking process that has been largely ignored in the current generation of tools. I'm uncertain as to why exactly because people have been aware of these needs since the dawn of computing. I chalk it up to the effort. We've been so busy trying to get the shapes on the screen that it's been difficult to do much else.

Alternatives

The piece that's been missing is support for the exploration of alternatives and the recording of history. You and I, we solve problems by trial and error. Even in writing this essay, I have tried different phrasings, moved paragraphs around, and tried a variety of passages. Think about the last time you wrote a document, or made a design. Did you nail it on the first try? I'm guessing you didn't.

We try alternatives in even the most basic communication and supporting this activity has largely been ignored in our tools. What do you want to do if you want to try a different phrasing of a sentence? Copy and paste. Modify. What if you want to read the document with both sentences to compare? It's awkward. Now imagine, this kind of activity in each of the tasks you do on the computer. You're left to workarounds.

History

Not only do we create alternatives, we don't think in straight lines. We think in loops and often revisit past ideas. And because of the ephemeral nature of computers, if you don't save something it ceases to exist. Our past ideas, sometimes our best ideas are lost. Only the most organized of us manage to create the systems to capture their past work. Even so, it's work that we have to think about. It shouldn't have to be this way. Computers should be able to do this for us.

Next

Therefore, it's time for a new productivity suite, one that reexamines generations of assumptions, deeply considers what works for the human body and mind, and leverages the newfound capacity of computers that can think.

This is why I'm working on Workroom.