The Airtable co-founder is refounding his 13-year-old company around multi-agent AI. He calls it the warm-up.
On a Tuesday in late January 2026, Howie Liu shipped Superagent and quietly bet the next decade of Airtable on the idea that one AI is not enough. You want a team of them. Talking to each other. Working in parallel. Then handing you a finished thing.
Liu is 36, lean, and unhurried in the way founders only become after their company stops being an experiment. Airtable has been an experiment for thirteen years. It is now an experiment again. He likes that.
The product he just released, Superagent, does not look like the no-code spreadsheet he became famous for. It looks like a co-worker. You ask it to size up a European expansion. It splits the work. One agent reads the financials. Another scans competitive positioning. A third pulls earnings transcripts from FactSet. Then a coordinating agent stitches the answers into something interactive and ready to present. The whole thing lives at superagent.com, a domain Airtable owns and treats like a second front door.
Liu has spent the last two years telling colleagues that the move to AI-native software is bigger than the move to SaaS. Most CEOs say that. Liu rearranged his entire org around it. He acquired DeepSky in October 2025 to bring in autonomous-agent talent. He hired David Azose, who used to lead engineering for ChatGPT's business products at OpenAI, as Airtable's CTO. Then he reorganized his teams around AI in a way Lenny Rachitsky wrote four thousand words about and called "a re-foundation."
"The breakthrough isn't a smarter single agent," Liu said when Superagent launched. "It's multiple agents working together." He name-checks Anthropic's Claude and the Manus agent as the only other products he thinks have a real, autonomous agent architecture. Everything else, he argues, is a glorified workflow with an LLM call wedged in the middle.
The myth most often told about Liu starts in College Station. His father, a Texas A&M biochemistry PhD, kept a C++ training book on his desk. Howie was thirteen. He picked it up. He did not put it down.
At sixteen he enrolled at Duke to study mechanical engineering and public policy. His undergraduate research involved computational airfoil design using genetic algorithms and neural nets, which is the kind of sentence a young person writes when they intend to graduate early and start a company. He did both.
In 2008, while still at Duke, he was involved in the formation of Freestyle Capital. In 2009, the year he graduated, he co-founded Etacts, a CRM that helped salespeople remember the humans behind their pipeline. He was twenty. Within a year, Salesforce had acquired Etacts and given Liu the keys to its social CRM product. He stayed long enough to learn how a $100B company makes software, and short enough that the lesson did not calcify into habit.
In January 2013 he co-founded Airtable with Andrew Ofstad and Emmett Nicholas. The pitch was disarmingly small: a spreadsheet that worked like a database. The pitch grew. By 2018 the company was a unicorn. By 2021 it had raised $735M at $11B in a Series F round. Liu had been ahead of schedule for so long that he started saying out loud that the schedule itself was the problem.
Liu separates agents from "LLM-powered workflows" by four properties. The latter dress up as the former. He has been clear about which is which.
Liu's reputation inside Airtable is built on caring about the look and feel of software with an intensity that borders on disagreement. He has argued for years that the best business software empowers rather than restricts.
To student founders at Stanford in 2021, he counseled patience: wait for the idea you cannot unsee. Then, when you see it, move faster than your friends think reasonable.
"Prefers agility over fixed long-range plans," is how he describes himself. The reorg around AI happened in months, not quarters.
Liu uses the word "refound" deliberately when describing his job in 2026. The implication: the company that exists today is not the one he intends to be running in five years.
He name-checks Claude and Manus, not the broader leaderboard. The list is short on purpose.
Superagent is the bet expressed as a product. The competing hypothesis - one big general agent - is everywhere. Liu is putting his next decade on coordination.
The kinds of details that do not show up in a Crunchbase profile but tell you something anyway.
The training manual was on his father's desk. The father had a PhD in biochemistry. The kid was thirteen. None of this was scripted.
Undergrad research at Duke involved using genetic algorithms and neural nets to design airfoils. Years before the AI boom, Liu was already letting algorithms find shapes humans hadn't.
A photo of Ginkaku-ji under snow. He had just sold Etacts. He was twenty-one. He took it himself.
Liu has said that Superagent could eventually eclipse Airtable. This is not the kind of thing most CEOs say about a thirteen-year-old core product. It is the kind of thing a founder says when they have decided to bet the company on what comes next, not what got them here.
The end state, as he describes it, is software you delegate to. A team of agents that plan, divide, investigate, synthesize, and present. A user who is more conductor than typist. An Airtable that is less spreadsheet and more nervous system for an organization full of AI labor.
He says he is energized. He says the disruption will arrive faster than people think. He says we are still in the warm-up. The interesting thing is how often he is right about when something is about to happen.