Who Is Jay Hack?
Jay Hack has been building AI since high school — from a university AI lab in Michigan to Stanford multimodal deep learning research to founding two AI companies before turning 30. His first company, Mira AI, exited for $28M. His second, Codegen, was one of the first coding agent companies and was acquired by ClickUp in December 2025. Now he's Head of AI at a $4B work management platform, leading the team building Super Agents — AI agents that live inside the work graph rather than alongside it.
What makes Jay worth paying attention to isn't the resume. It's his thesis. While the rest of the industry races to build better models, Jay argues that context — not capability — is the real bottleneck. He believes vertical AI is doomed, that the best model is already the best model, and that the companies that win will be the ones that give agents the richest possible understanding of how work actually flows. He sold his own coding agent company because he saw that coding agents were actually just the first wave of a much larger transformation hitting every knowledge work industry.
He says it with the certainty of someone who stopped debating it months ago and started building on it.
The Archetype: The Magician
The Magician
The Sage
The Return
Jay's core move is transformation through hidden knowledge. He sees connections that are invisible to others — that coding agents are actually general-purpose knowledge work agents, that Merkle trees from cryptography solve AI context management problems, that the progression from Assembly to C to Python to LLMs maps onto how every industry will be transformed by AI. The Magician doesn't just build things; he changes how people see the landscape.
His secondary archetype is The Sage. Jay accumulates knowledge systematically and genuinely loves explaining what he finds. "The reason is..." is his natural lead-in to any answer. Galaxy Brain, his personal open-source eval repo built with no commercial purpose, is pure Sage behavior — building a knowledge tool because understanding is intrinsically satisfying. But in Jay's case, the Sage serves the Magician: he gathers knowledge specifically to see the transformations others miss.
"The progression that the coding industry has gone through will be the progression that many other industries go through."
The Hero Match
Hermes
In Greek mythology, Hermes is the god who moves between worlds — Olympus and the underworld, gods and mortals, the visible and the hidden. He invented language and commerce. He's not the ruler of any single domain; he's the one who connects them all, carrying insights from one world that turn out to be answers in another.
Jay's career follows the same pattern. From Mira (computer vision) to Codegen (coding agents) to ClickUp (horizontal work management), each move carried knowledge from the previous world into a new one. His insight that coding agents are actually knowledge work agents is a boundary-crossing move — he saw that the message from the developer world was the answer for marketing firms, construction companies, and legal services. The Merkle tree memory system is another crossing: importing a cryptographic concept into AI context management because the underlying problem was structurally the same.
What makes the Hermes parallel specific to Jay: he doesn't stay in one domain long enough to become its undisputed master. He stays long enough to understand its deepest pattern, then carries that pattern somewhere it's never been.
Tony Stark — MCU, post-Avengers arc (Iron Man 3 through Endgame)
Not the brash inventor of the first film. The later Stark who has already proven he can build anything and now spends his energy on the systemic threat that nobody else sees coming. Jay has already had his exits. He's not trying to prove he can build. He's trying to prove that the entire industry is misreading the map — that context, not model capability, is the real bottleneck, and that agents need to live inside the work graph.
"Don't pitch your business. Let the person arrive at the conclusion."
That's Stark energy. Build the argument so well that the conclusion feels inevitable. And like the later Stark, Jay channels his energy into building the infrastructure that will matter when the wave hits — not into winning today's battle.
The Story Behind ClickUp
Jay didn't sell Codegen because the company was failing. He sold it because he saw something nobody else in the coding agent market could see yet.
The coding agent space was getting crowded. Devin dropped pricing from $500 to $20 a month. Foundation model labs started shipping their own coding tools. Every week brought a new entrant, each one building essentially the same thing — an AI that writes code in a sandbox. Jay watched this unfold and saw a pattern he'd seen before: commoditization was coming, and the companies still raising Series A rounds for domain-specific models were about to get wiped out. "Vertical AI is doomed," he said — not as a prediction, but as a settled fact.
What Jay realized was that coding agents were actually just the first implementation of something much bigger. The same multi-agent orchestration, the same sandbox execution, the same context engineering — all of it applied equally to marketing teams, accounting firms, construction companies, legal services. The progression that the coding industry had gone through would be the progression that every knowledge work industry would go through. And the company best positioned to serve all of those industries was the one that already had the work graph — the one that already understood how tasks, projects, and workflows connected across an entire organization.
So Jay brought Codegen's technology to ClickUp, where it became the foundation of Super Agents — AI agents that don't just generate output but actually operate inside the flow of work, with a proprietary memory system inspired by Merkle trees that gives them persistent, verifiable context across tasks.
The Founder's Journey and The Company's Journey
AI lab researcher at Michigan, Stanford deep learning, founded Mira AI ($28M exit), founded Codegen (coding agents), saw the commoditization wave coming, sold to ClickUp, now building the infrastructure for AI-native work management as Head of AI.
ClickUp grew from a project management tool to a $4B horizontal work platform. The Codegen acquisition brought multi-agent orchestration and sandbox execution technology in-house. Super Agents launched as AI that lives inside the work graph. The bet: context, not model capability, is the moat.
The same archetype drives both: the Magician who sees that the transformation in one domain IS the transformation in every domain. Jay didn't join ClickUp to run an AI team — he joined because ClickUp's work graph is the bridge he's been building toward his entire career.
How Jay Leads
Jay shares the work and owns the direction. When he narrates company decisions, they belong to the team — "we were thinking about the right set of problems," "we built a proprietary memory system." But his directional calls are first-person and unambiguous: "Vertical AI is doomed." "I do not partake in AI semantics debates." He builds consensus for the how, and he decides the where.
There's a specific kind of confidence in how he holds controversial positions — he states them as facts, not opinions, and he backs them with structural arguments rather than appeals to authority. At the same time, he updates his methods when the evidence demands it: testing candidates with AI instead of LeetCode, being willing to "burn down your entire tech stack every four months." He doesn't perform humility broadly, but he changes his mind about execution faster than most people change their socks.
Founder Superpowers
Seeing the bridge before the river
Jay recognized that coding agents were actually general-purpose knowledge work agents before the market did. While competitors were optimizing for developer workflows, he saw the structural parallel — "the progression that the coding industry has gone through will be the progression that many other industries go through." His decision to sell Codegen to a horizontal work management platform was the bet itself. The ability to see that the transformation in one domain IS the transformation in another — and to act on it before the second domain realizes it's happening — is genuinely rare.
Importing concepts across domain walls
The proprietary memory system "inspired by Merkle trees" is the signature move. Merkle trees are a data structure from cryptography. Nobody in the AI agent world was thinking about them for context management. Jay imported the concept because he recognized that the underlying problem — verifying and syncing distributed state — was structurally the same across domains. He does this repeatedly: compiler history informs his view of LLM abstraction levels, coding agent architecture informs his approach to marketing firm automation. Most founders stay inside their domain's vocabulary. Jay treats domain boundaries as opportunities.
Making the other person arrive at the conclusion
"Don't pitch your business. Let the person arrive at the conclusion." This is both his advice and his observable behavior. Throughout the conversation, Jay wove ClickUp's value proposition into nearly every answer — but each mention arrives as a logical consequence of the argument he's building, never as a pitch. He constructs the premises so carefully that the conclusion feels like the listener's own idea. Persuasion through architecture rather than assertion.
What It's Like to Work with Jay
Jay is deliberate. He lets questions land before answering, and he chooses his words with the care of someone who treats spoken language as permanent. In a 79-minute unscripted conversation, he asked for retakes twice — not because he said something wrong, but because he wanted a more precise framing. That precision isn't perfectionism; it's a signal that he treats his public statements as commitments.
He's most alive when exploring ideas at the edge of what's known. The last twenty minutes of the conversation — unstructured brainstorming about 100M-token context windows and the future of UIs — brought out a different energy: looser, faster, more personal. He daily-drives 6+ AI clients not because his job requires it but because he genuinely wants to know what's possible. Galaxy Brain, his open-source eval repo, exists because he finds understanding intrinsically satisfying. If you work with Jay, expect someone who cares about being right about the pattern more than he cares about being comfortable — and who will change his methods the moment the evidence demands it.
"You have to be ready to burn down your entire tech stack every four months."
Why This Matters (For You)
If You're a Team Using AI Agents for Knowledge Work
Jay's core insight is that agents need context, not just capability. If you're deploying AI into marketing, legal, accounting, or project management workflows and finding that the output is generic or misses crucial context, you're experiencing the problem Jay built ClickUp's Super Agents to solve. His argument: the best model is already the best model. What matters is whether the agent understands your work graph — your tasks, your team structure, your project history. Before evaluating another AI tool, ask yourself: does this agent know what my team is actually working on, or is it just generating output in a vacuum?
If You're an Engineer Building AI Agent Infrastructure
Jay's approach to context engineering offers a specific technical lesson: look outside your domain for solutions. His proprietary memory system was "inspired by Merkle trees" — a data structure from cryptography that nobody in the AI agent world was thinking about for context management. He imported it because the underlying problem (verifying and syncing distributed state) was structurally the same. His career trajectory from computer vision to coding agents to horizontal work management is a case study in recognizing when patterns repeat across domains. The next time you hit an infrastructure problem, consider whether someone in an adjacent field already solved it.
If You're Early in Your Career
Jay's career arc is a masterclass in knowing when to leave. He didn't stay at Mira after the $28M exit to optimize what he'd built. He didn't stay at Codegen to fight the commoditization wave. Each time, he recognized that the deepest pattern in his current domain was ready to be carried somewhere new — and he moved. The lesson isn't "jump jobs frequently." It's this: stay long enough to understand the deepest pattern, then carry it somewhere it's never been. That's how you compound insight instead of just experience.
If You're Considering Joining ClickUp
Jay's leadership style is directional clarity with team ownership. He decides the where; the team decides the how. He hires for "AI maximalism" — people who are already deep in the tools, already experimenting, already building. He tests candidates with AI rather than LeetCode because he believes the old hiring playbook is obsolete. If you thrive in environments where the strategic direction is clear and the execution is yours to own, and you're genuinely excited about the intersection of AI agents and real-world work, this is someone who will share credit for the wins and take personal ownership of the bets.
Go Deeper
The full conversation with Jay Hack is on its way. Check out other episodes in the meantime.