Remote

A few weeks ago, I delivered a technical document to our Product Manager. It wasn’t written for him, and he couldn’t quite follow it. Instead of scheduling time with me to walk through it, he posted it in our team channel and asked, “@Claude, explain this doc to me in a way that a product manager who isn’t technical could understand.” Claude read through the whole thing and explained it back to him point by point, in plain language. I followed along and could verify that everything it said was accurate. A few messages later, I watched the moment it clicked for him. No meeting. No me stepping in to translate. That’s new, and it’s a big part of why I think AI is becoming essential to how I’ve worked for most of my life.

I have been working remotely all of my adult life. Basically since 2001 working remotely has proven itself to me to be the best way for me to work. That isn’t to say that it is the best for everyone to work. I feel most productive at home in my surroundings. Certainly a quick flight out to a central location and a meet-and-greet with folks you work with daily is valuable and cannot be replaced, but day-to-day, I need the solitude and familiarity of my home.

I am far more productive because there are fewer distractions in my space. I know not everyone has that experience at home, but I have engineered my life so that my space is distraction-free. I have a dedicated office space with a nice large desk for all of my thoughts and projects to live in. When I am in that space, I am “at work”. However, I also find it meaningful to move about my space. Sometimes I sit in a quiet living room near a window with greenery outside. Other times I take a quick stroll around the block to get a little space away from a problem. I also have the option of changing spaces completely and working outside or in a restaurant. These are all ways that I keep my work-life balanced and diversified.

This didn’t come about easily. Back in 2001, when I first started working with my first company remotely, I used to get an email with tasks. These were the days before JIRA, Slack, and Zoom. Even Google Workspace didn’t exist. Everything I did was through terminal/PuTTY and Remote Desktop. I would get an email with a list of items to do, dutifully do each item, and then send an email back to say the work was complete. My main job was system administration. Bring up and tear down systems, upgrade services, keep email, websites, and different adjacent services online for our customers. Keep those systems up-to-date and ensure that we were keeping up with the leading edge of technology.

As great as all of this sounds, there was one big issue. I was isolated. I would get a task-list, do the work, then signal I was done and get more work. If a system was having trouble, someone would email me or call. However, generally speaking, I had no idea why I was doing what I was doing. I vividly remember sending that first fateful email, expressing my desire to have a regular time that we could all meet. The company was small. I was employee #1. There were few other colleagues to discuss things with so why not a regular call so we could sync on the why? This became a weekly company update meeting. We would discuss priorities, roadmap items ahead, concerns around costs, products, etc. Eventually, we got a shared line that each of us could call into that would host the meeting.

Over time, all of the usual suspects started to arrive to make this type of work easier. VoIP ATA adapters, Skype, a centralized ticketing system we built from the ground up (Zendesk did not exist). We often joked that we were building products for remote life and we should sell those things as services themselves. However, there were too many different product areas for a small company like ours. We were highly focused on our product roadmap and those distractions would get in the way of the mission of the company.

Eventually I grew out of doing systems administration work and moved to actual software development. We hired someone to take on my old role and now I was a full-fledged developer. As I was on-boarding, I was informed of a system called SVN. Basically SVN was where engineers would check in and check out code. Today people would be more familiar with Git, but this wasn’t exactly like Git as it provided the same solution but was a much bigger headache to use. There were procedures to the process of committing code. Very little oversight in our small SVN system. I don’t think I ever performed any sort of formal review of any code, nor was I tasked with getting reviews. Just check out, add your feature, make sure it builds, double check with the team that it was safe and check in your code. For me, this was an intimidating process because I was so used to working fully alone on projects. If you want to read up on the history of VCS, I found this great article describing “ The World Before Git “.

We had annual get-togethers with the company and clients where we could sync up on where we were as a company and build enthusiasm with our clients as well as our team. We would use this time to showcase new product ideas on the roadmap and host tutorials and Q&A sessions on how to use our products. However, there was a great divide that still somewhat existed between where I was and where the company was. Often, miscommunications around priorities and my capability as an employee were the biggest gaps. I remember writing up a piece of software in a week. This was done in my spare time, because I wanted to learn a new domain of engineering. I showcased this project to my boss. They were blown away and told me they were building almost the exact same thing under a skunk-works project. We never found full alignment.

Even with the introduction of Google Workspace, JIRA, Zendesk, Zoom, and Slack, all of these products never solved anything our in-house tools hadn’t already solved. The issue was that in-house projects represented a risk. Another system you must maintain, another piece of software you have to patch. Each one adds more and more burden to the team. So naturally those systems were replaced because the promise was that we no longer had to maintain them. These truths are as true today, in the AI era, as they were then. I remember racking our own servers in a shared facility for all of our systems. Today, you’d offload those responsibilities onto AWS, GCP, or Azure. By the time I left that job, we were on-boarding our stuff onto Azure. From company to company, each pattern of remote work continued to play the same tune. Git was far easier to work with. Comms with Slack were easier than email. Quick meetups are a slash command away. However, the same issue that plagued us from my first remote job to now still is a problem. How to keep all of these remote people in alignment?

Enter AI. Many new tools have been coming out related to AI. AI coding assistance represents a huge promise for increased development. I think it is fair to say that the jury is still out on the ROI of all of this AI spend, but some interesting workflows have started to develop representing, I believe, a monumental shift in how we work. Especially in the remote working world, AI represents what may be the biggest change I have experienced in my working life as a Software Developer. Last year we knew that the Software Development Life Cycle (SDLC) could be shaken up. We just weren’t exactly sure how. Is AI a performance unlock that makes all software development 10X? Would planning be upended? Does this allow ideas to flourish more richly by allowing novices to create software mockups? This would change what was once a presentation and words into a mock app that allows everyone to touch and feel the experience in new ways. Or, were we facing something new, AI that could develop autonomously?

All of this has been vague until this past month. In late June, Anthropic released Claude Tag. At the time it was announced I was curious. An AI agent that runs in Slack. I have built an internal agent in Slack before that assists users with finding information. However, one of the challenges that I am careful of when introducing AI is the engagement that AI has on the humans living with it. I have seen previous attempts where AI introduced noise and confusion, so having an organic relationship with AI is critical. Claude Tag starts out as an assistant you need to mention before it engages, but when it introduces itself to a channel, it gives a summary of what the channel is about along with how to control how it engages with the channel. You can tell it to only engage with links, specific types of links (like PR/MRs or articles).

I wasn’t exactly certain what value AI would provide in this context. However, I quickly started to get it . It serves as a communal prompt for our team. You converse about a plan, problem, bug, etc. Then you ask Claude what it thinks, which then provides a summary view of the conversation along with a proposed solution. You tweak the summary or proposal. Once the group is happy, you tell Claude to act. A recent example: a question came up about how to implement a feature we were building. My IC and I talked through the details back and forth, and once we landed on an approach, we realized it wasn’t actually that complicated. So we asked Claude how it would build it. It gave us a detailed implementation plan, and when we said, simply, “open a merge request,” it went and did the work itself. My IC reviewed it later that day and merged it in. Easy peasy.

At first blush this doesn’t sound that advanced or impressive, but I quickly started to see what the true value of this experience was, especially in remote work. Claude acts like a mediator of context in the discussion. If someone is not in sync, they can ask Claude to explain it. That’s the same thing our PM did with that document. Or they can work with their teammates to get in sync often without needing to call a meeting. It also allows your quiet observer teammates to read up on what they missed and even engage with Claude to get quickly caught up. It serves as an alignment tool much like sharing an office space. It brings shared context to a remote interface that allows a team to stay in alignment.

It is still early in this process of evaluating AI, but I am starting to see a picture of a remote world that now can no longer exist without AI because of the huge disadvantage you would have against companies that use AI or share office space. This is the first additive tool to the remote work experience that is truly showing me that AI has a place that is uniquely its own in fully remote companies.