In the mid-2010s, I found myself going down Arxiv rabbit holes as I explored research advances in natural language understanding. One of the papers that stuck with me was GloVe: Global Vectors for Word Representation (2014) by Jeffrey Pennington, Richard Socher, and Chris Manning. It showed how machines could capture meaning by mapping word co-occurrences into vectors, giving language a shape computers could finally work with. That idea, that computers could comprehend a language built for human communication and understand its nuance, was as elegant as it was powerful.
I was hooked, and chose a career path where I got to team up with AI researchers to help build customizable deep learning models that productized the latest in NLP advances. Those years turned inspiration into cautious optimism. I got to follow along firsthand as sequence-to-sequence approaches unlocked translation services, as new SOTA benchmarks were established for Q&A tasks, and as embedding methodologies would advance the way computers understood the nuance of human language. These breakthroughs weren’t just novel, they actually opened up entirely new use cases. You could watch how, in just months, a computer would advance from grasping the contrast in a word like but, to answering questions, powering bots, and summarizing text.
Then in 2017 came the paper. The paper that changed everything: Attention Is All You Need. Sitting with colleagues as the Transformer was introduced, we knew the trajectory had shifted. From embeddings to attention, the idea that machines could not just parse but generate language opened up a future that’s still only just beginning. Through all of this, I fell in love with the idea that your experience with software could move beyond clicks and code into natural conversation, and that neural nets would continue to reshape how we interact with machines.
But new technology doesn’t solve every problem. Even as hyped “agentic systems” impress us with their capabilities, the same problems from the aughts remain: fragile integrations with slow APIs, enterprise systems weighed down by decades of legacy configurations, solutions that work in silos but never scale across the flow of real work. This is what drew me to DevRev. Because here was a clean slate to deliver on a vision that matched modern technology with the real world realities of how work gets done. DevRev wanted to reinvent how we work, not just asking a model to call an API, but enabling it to read, write, generate, summarize, ask questions, and connect people as naturally as sending a text to a friend.
I wanted AI to feel like that teammate you love to work with. It needed to have style, i.e. be consumer-grade in its design, but also maturity, i.e. be enterprise-grade in trust and security. It took us five years to put those pieces together: standardizing schemas and ontologies, building a knowledge graph that captures relationships across data, and embedding the right context engineering so the system could be precise, reliable, and safe. And, yes, have a vibe.
That knowledge graph gave us confidence that we could build something as powerful as a point solution but general purpose enough to become habitual in how you work. It showed us that the future wasn’t about bolting AI onto old software, but creating a new foundation, one where quality, precision, and craft mattered as much as features.
Now, I could not be prouder: we are ready to put all these years of hard work, risk, excitement, and next-level teamwork into practice: we are ready to introduce Computer.
Computer is our first, and huge, step in making the dream of an AI teammate real. One that doesn’t replace you, but amplifies the work of your entire team. That helps your team’s focus and avoid distraction, blocks interruptions, and contributes ideas because it understands your priorities. That prepares you for meetings, organizes your time, deflects the noise that drags you away from reaching your goals and outcomes. And, most importantly, it collaborates alongside you in the flow of work. Computer is AI built for work, in the very best sense of the word.
When you first meet Computer, it starts like any new colleague, with a new employee “bootcamp”. And, as you are going through your onboarding into Computer, Computer is also onboarding onto your team. It asks about your goals, your team, your tools. It connects to your systems and begins to AirSync data into its Memory, learning how you work. And it learns fast. Soon, it can anticipate your day, pull context from private, organizational, and public data, and proactively support you in ways that fit the way you and your team want to work.
Computer can be prescribed, customized, and shaped to behave not just as an assistant for one person, but as a shared teammate that helps the whole group accomplish more together. And what makes it special isn’t just what it does, but how it does it, with a sense of fit and flow that makes you want to use it, and not just because someone else says you have to.
Computer is the teammate you’d pick first when the team splits up for an exercise. The one that blends right in, but that you know you can always trust (even if it can’t catch you in a trust fall).
I believe Computer will define a new era for knowledge workers. That’s why we think long term, build with craft, and move intentionally in how we release what we’ve already proven works.
Many AI companies are busy hyping another agent for this, another agent for that. But this path will just replicate the UX pitfalls of the past. Agents shouldn’t resemble work today; the hundreds of disconnected apps, thousands of Slack channels, and endless forms and approvals just to move incremental work forward.
Computer takes the opposite approach. It unifies. It works with your tools rather than multiplying them. It brings high intelligence with high EQ. It integrates into your team’s rhythms and makes outcomes collective, not fragmented.
That’s what makes me most excited for what we’ve built. Computer is much more than just another superficial ‘Agent’, its a teammate worth working with. A product designed with technology optimism at its core. An AI built to support, not replace us.
