I've been working mostly remotely since 2018: first on a consumer-facing clothing search app, then a CNC machining startup with a physical factory, and now at Pydantic. I don't think remote work suits every person or every team; I think it's important to build a team of people who can work independently and responsibly, and to hire for demonstrated ability to operate in that kind of environment. It has real advantages and real disadvantages.
But I'm not writing this to rehash the remote-vs-office debate of 2020. During the pandemic, most people learned that remote work was at least a passable alternative to being in an office. The more interesting question in 2026, especially as RTO mandates sweep through the industry (sometimes as a thinly veiled way to reduce headcount), is whether the benefits are worth the tradeoffs. My opinion is that remote work isn't just acceptable now, AI has changed the calculus enough that it's actually better than in-person for a lot of teams, and it's getting better every day.
The Old Tradeoff
Historically, the case for remote work was straightforward: you could hire from a global talent pool and skip the overhead of office space. The case against it was equally straightforward: collaboration suffered. Hallway conversations didn't happen. Onboarding was harder. Ideas got lost between Slack threads and time zones.
This was a real tradeoff, and reasonable people landed on different sides of it. The research reflected this ambiguity: Nick Bloom's studies at Stanford found that remote work's productivity impact ranged roughly from -10% to +10% depending on the role, the individual, and how well it was managed.
The key word there is managed. The quality of remote work has always been more about the tools and practices surrounding it than about remote work as a concept, and in 2026 both the tools and practices look very different than they did even a year ago.
What Changed: AI Closed the Collaboration Gap
Everyone is (rightly) caught up in the transformation happening to coding itself. AI is enabling entrepreneurs to build things that couldn't have been dreamed of a few years ago, and enabling people without deep technical backgrounds to build valuable businesses from domain-specific knowledge they uniquely possess. At Pydantic, we've been using AI coding assistants to design and ship features much faster and more ambitiously than we could before (though it still requires quite a bit of oversight and human-in-the-loop to get the level of quality we aspire to in our libraries and products).
But there's a less well-publicized transformation happening in parallel, a transformation in how we communicate and collaborate, and I think it disproportionately benefits distributed teams.
Meetings That Don't Vanish
How many times have you had an energizing conversation in a small group about how you were going to build something, gotten excited about the details, and then by the time you actually sat down to implement it, realized you'd already forgotten half of what you discussed? With AI-powered transcription, this is starting to feel like a thing of the past.
At Pydantic, we've developed a habit of catching ourselves in ad hoc huddles right at the moment we're about to dive into implementation details, and someone will say "let's turn on AI notes." I frequently hear "wait, wait, wait" in a huddle right before the Slack voice comes through: "AI notes and transcription on."
Once you have the transcript, you can feed it to Claude and you're 90% of the way to a fleshed-out product requirements document, ready for finishing touches before implementation. The ideas that used to evaporate into the ether now get captured and turned into artifacts. It's a little painful to think about how many great ideas were lost to the gods of forgetfulness and entropy before we had this at our fingertips. There's just so much less excuse for that now.
And this advantage is native to remote work. When you're on a video call, turning on recording is a single click. At our last in-person offsite, I was struck by how frequently I wanted to reach for the record button but couldn't because we were all in a room together. Two people using their own microphone on a Google Meet call still produces a better transcript than any in-room setup I've tried. (If there are good tools for this that we're just not aware of, please share!)
Everyone Gets an Office
Here's something I don't hear get brought up too frequently: remote work means everyone has a private office.
Beyond the obvious benefit of being able to focus without open-plan distractions, a private space means I can use tools like Wispr Flow for voice-based dictation without disrupting colleagues or feeling self-conscious (in fact, I used it liberally while writing this article). It also means having somewhere to retreat when you're frustrated, or dealing with something personal, or when you just need to think.
In most in-person work environments, private offices are typically a perk reserved for executives. With remote work, everyone just has one. This might seem like a less obviously AI-powered point, but the dictation angle is worth dwelling on: if you've ever struggled with RSI from long coding sessions, being able to use voice-based tools without self-consciousness is a huge quality-of-life improvement. AI-powered dictation is good enough that it's a real alternative to typing for a lot of tasks, especially for stream-of-consciousness dumps of brainstormed product requirements for new features, which can be cleaned up and iterated on by a coding agent with full access to the underlying codebase. And that's something that open floorplan offices make much harder to take advantage of.
Transparent Work Without Surveillance
Working from home is a privilege, not a right. It's something we earn by doing work the world (or at least our teams) can plainly see is effective, and I think the transparency that AI can provide reinforces that in a healthy way.
In the past, it was impractical for your manager (let alone your peers) to have a nuanced picture of everything you shipped over the past quarter. The proxies we used were reductive at best: lines of code, number of pull requests, story points. None of these capture what actually matters.
With AI, you can generate a nuanced summary of someone's contributions that everyone agrees is a fair representation, and then have a real human discussion about performance that's grounded in the substance of the work rather than an automated scoreboard that determines your annual compensation adjustment.
I know this can sound dystopian, and I think it's important to acknowledge that. AI-generated summaries can hallucinate contributions that didn't happen, fail to adequately represent the kind of work that doesn't leave a clear paper trail, or just feel inhuman as a way to evaluate people's livelihoods. Even at Pydantic, out of an obvious abundance of caution, we haven't moved to heavily automated performance review workflows. But I think there's real promise here if we can figure out how to do it well. If I could just remember all the work my colleagues had done over the past quarter, I think I'd do a better job of appreciating it; I believe AI can help with that, and I'd love to see more experimentation in this space. But any proposal along these lines needs to be met with a skeptical, critical eye first, not just accepted at face value because it sounds efficient.
The same applies to meetings. Imagine being able to ask of your team standups: Who is being spoken over? Who is dominating the conversation? How can we restructure this meeting to actually serve our needs? If your meetings are transcribed, this is within reach with AI assistance. Formal minutes have been standard practice in parliamentary proceedings and boardrooms for centuries, and given how much more accessible this has become, I think there's really no excuse not to be doing something similar for a large fraction of the meetings we have at work. The benefits are significant.
To be clear, this isn't about recording everything, all the time. I've actually heard stories of offices that have recording devices to capture all in-person interactions, and that sounds a lot more Orwellian than my day-to-day. I like being able to speak off the record, and I do so more often than not; it's important that people have space for half-formed ideas and private gut-checks with trusted collaborators. The point is having the option to capture what matters, when it matters, at the click of a button, and remote work makes that option frictionless in a way that in-person work simply doesn't.
What the Research Says
A lot of the points I've made above, especially around AI-powered workflows, aren't backed by formal research yet (if you know of any, especially research that contradicts what I've been saying, I'd love to hear about it). But there is useful research on remote work more broadly that's worth calling out for people who might not be aware of it.
The strongest evidence on remote work itself comes from Bloom's randomized controlled trial at Trip.com, published in Nature in 2024. The study followed 1,612 employees over two years with a proper control group, and found that hybrid/remote arrangements reduced quit rates by 33% with no measurable impact on performance or promotions.
On the other side, companies like Amazon, JPMorgan, and others have pushed aggressive return-to-office mandates. But a University of Pittsburgh study examining S&P 500 firms that issued RTO mandates found no significant improvement in financial performance, while employee satisfaction declined measurably.
Gallup's data tells a similar story: hybrid workers report the highest engagement levels, and fully in-office workers who could be remote report the lowest.
None of this means remote work is automatically better. The research is clear that it requires intentional management. But the old assumption that you're trading productivity for flexibility? The data just doesn't seem to support it.
What Remote Work Still Requires
AI can't compensate for bad practices. I want to spend some time on the requirements of remote work because they're often glossed over in pro-remote discussions online, and I think painting a realistic picture is more useful than an idealized one.
Push Communication
Remote work demands that you proactively share context rather than waiting for someone to pull it from you. This is especially critical for junior developers, and in 2026, that includes many of the growing number of people building software primarily through AI-assisted prompting.
If you're new to a codebase, I think one of the most important things you can do is explicitly ask your coding agent about pre-existing patterns and infrastructure relevant to your task before you start vibing from scratch. Another is asking a senior teammate for a pair session to review what you've built. If they can't find time for that, it might be a sign you're working on something that isn't high enough priority to warrant senior attention, and it's worth asking whether your time could be spent elsewhere. If you are working on high-priority tasks and still can't get review time, that's a management failure. It's admittedly easier for this kind of thing to slip through the cracks in a remote environment, but it's also something that can be readily addressed when the team has a culture of pushing information proactively and flagging blockers early.
Synchronous Time Still Matters
I've heard of work environments that optimize for making absolutely everything asynchronous. I think that's an interesting exercise, but it's not what I aspire to. I enjoy face time with my colleagues, and I want to work with a team where that feeling is mutual.
At Pydantic, despite having a significant split between Eastern and Western hemispheres, we fit plenty of synchronous interaction into the timezone overlaps. We do have a couple of larger all-hands meetings each week, but the bulk of our meetings are small, targeted discussions between the people actively collaborating on something. I think that balance matters a lot, because large meetings where most people are sitting quietly are mentally draining, while small focused conversations are where the real work gets done.
In-Person Gatherings
We get together a few times a year for company-wide offsites and occasionally at conferences. The team-building that happens at these events is significant, and I haven't seen it fully replicated through remote work alone.
Despite how rarely we're together in person, one of the strongest impressions I have from these gatherings is how frequently the right thing to do turns out to be for everyone to take out their laptops anyway. Want to show a demo to ten people? Crowding around a laptop isn't practical, but watching a screenshare from your own machine is. Want to record a Q&A? Tossing around a microphone doesn't work well, but everyone using their own audio channel on a recording service does. Have a colleague who couldn't make the trip? If people aren't dialed in, they just can't participate, and trying to include one remote person in a room full of people is always awkward.
The point of being together isn't to simulate an office, it's for the unstructured conversations, the dinners, the shared experiences that build trust.
Reliable Infrastructure
This is table stakes, but it matters: reliable internet and high-quality audio and video equipment. Bad audio on calls is the fastest way to erode the quality of remote collaboration. At Pydantic, we take this seriously and offer our team a dedicated budget specifically for home office setup.
On Hybrid Work
I want to say something specific about hybrid and "in-person days" policies. I think it's worth distinguishing between two things that often get lumped together under "hybrid."
The first is a schedule like three days in the office, two days remote. I understand the appeal of this, and it can work well, but I think it's important to recognize that this is functionally closer to in-person work with some breathing room for focused, heads-down time. You still need office space, your team still needs to live near the office, and the in-person dynamics (politics of face time, hallway conversations that not everyone is part of) are all still very much in play. This is probably more practical for large organizations with a big geographic footprint and the budget to throw at the hiring problem. For startups that don't have enough people for multiple geographic centers, the talent pool cost of tying yourself to one location is much higher, and learning to work effectively as a distributed team can go a long way.
The second is a mixed team where some people are in-person and some are remote. In my experience (at Formlogic, where we had a fairly balanced split between on-site Pittsburgh engineers and remote engineers, largely out of necessity), this is where you can easily end up with the worst of both worlds. You have all the costs of office space, you have the geographic hiring constraints for the in-person roles, and you also have a two-tier system where the remote employees are at a structural disadvantage for visibility, context, and influence. It can work if you're very intentional about it, but it requires constant effort to prevent the remote half of the team from becoming second-class citizens, and we struggled with it despite serious effort.
Fully distributed with intentional in-person gatherings is, in my experience, a cleaner model. You get the benefits of a global talent pool and location flexibility, and your in-person time is concentrated and purposeful.
Why This Matters for Hiring
This is the part that makes remote work a genuine competitive advantage beyond a lifestyle preference.
When we're hiring at Pydantic, we're competing for the same caliber of engineer as companies 100x our size. Being fully remote means a candidate can join our team regardless of where they live without uprooting their life, whether that's Chicago, Berlin, South Africa, or Montana (where I live). In our experience, this amounts to a massive expansion of the talent pool.
I've experienced the inverse of this firsthand. As I mentioned above, at Formlogic we needed engineers on the ground in Pittsburgh, and it was glaringly obvious how much harder those roles were to fill compared to the remote ones. There were only so many software engineers looking for a job in Pittsburgh, and most of them were already at Duolingo. Tying yourself to any single geography, no matter how good, is a major constraint.
The Scoop Technologies Flex Index data backs this up at scale: fully flexible companies have been growing headcount faster than fully in-office ones. Larger teams aren't necessarily better, and in the age of AI I'd expect the trend to move toward smaller teams doing more, but the underlying signal is telling: when you can hire from anywhere, it's just easier to find the right people.
The Virtual Future
On most days during my last few years of high school, I'd come home from school and immediately hop on a shared Ventrilo server to play games with friends, Counter-Strike, Warcraft III, whatever. We'd spend hours in voice chat, strategizing, trash-talking, building friendships. Most of them were friends from school, but some were people I'd never met in person.
I didn't have the vocabulary for it at the time, but I dreamed the future of "remote socialization" was bright. Twenty years later, I'm struck by how little progress we've actually made on that front. Shared virtual spaces are still largely relegated to Discord and gaming, which are great communities, but not really mainstream. I think the corporate world has mostly overlooked the possibility that virtual presence could be genuinely good, not just something you tolerate. I'm sure not everyone would agree with me on this, and I get the instinctive reaction of "I don't want to be on a call with people all day." But being in an office is basically that, and nobody considers it crazy. I think there's something to be found in using technology to get the best of both worlds, and we haven't really tried hard enough yet.
We briefly tried Teamflow at Formlogic, and I loved the concept: move your avatar into proximity with teammates and you're immediately in an ad hoc voice chat. The largest problems were compute overhead (my laptop would get hot and eat battery) and poor noise cancellation relative to Zoom, so I'd always end up switching. But I think there was something special there. If they or anyone else has solved these issues, I'm keen to try again.
Wrapping Up
The old framing of remote work was defensive: it's not as bad as you think. In 2026, I think the framing should be offensive: distributed teams, equipped with modern AI tools, have structural advantages that in-person teams don't. Remote work isn't perfect, and I've tried to be honest in this post about the things it requires and the things it can't do. But the tradeoffs that used to make it a tough sell are getting smaller every year, and for companies like Pydantic that were already committed to distributed work, that's a significant head start.
We're hiring. If the thesis of this post resonates — that distributed teams with good tools and practices have structural advantages — and you'd like to help build the libraries and products powering that future, take a look at our open roles.
