YesPress Profile — Technology Executive
Chief of Staff to the CEO — Reflection AI | San Francisco, CA
She grew up in an outback town smaller than most office buildings. Now she's inside one of the most heavily funded AI labs on Earth, helping build something that doesn't exist yet: open superintelligence. The path in between is the interesting part.
The Story
Winton, Queensland sits somewhere in the deep outback, about eight hours west of Brisbane. Population: roughly 800. Famous for: being the alleged birthplace of Waltzing Matilda, and not much else. It is the kind of place that produces people who leave.
Casey Flint left - not fleeing, exactly, but pulled. Her parents were serial entrepreneurs, which means she grew up watching people build things and bet on themselves. That wires you a certain way. At 19, she was studying biochemistry at the University of Queensland when Uber offered her an internship. She took it. Left university eight months in. No passport. No blueprint.
Five years at Uber followed - Brisbane, Sydney, Amsterdam, Korea, Hong Kong, Japan. She became the first official member of Uber's Sydney competitive strategy team. She worked on a project for the person who would become CEO of Uber Eats. She spent eight months running a Korean joint venture. Each move was less a promotion and more a willingness to go where the work was most interesting.
"I've made this jump because I want to be more directly involved in and follow my passion for what AI is to bring."- Casey Flint, January 2025, on leaving Square Peg for AWS
Square Peg Capital came next. Four years as a Senior Associate, focused squarely on AI. She met over 1,000 engineers, researchers, and business leaders in the AI space - a number that sounds like marketing until you realize it's what happens when you make introductions for a living and genuinely want to understand the technology, not just price it. She wrote about AI constantly. Her Substack, "Artificially Intelligent," became a destination for founders and the AI-curious who wanted to understand what was actually happening at the frontier, not what the press releases said.
In January 2025, she made a move that surprised some people: she left VC for AWS. Not a lateral move - a deliberate step closer to the product, the compute, the actual work of building AI systems. She wanted to see AI "from chips all the way up to applications," she said. The bittersweet part was leaving the founders she'd backed. The exciting part was everything else.
She joined Uber at 19 without a passport. Three months after her first meeting with the company, she was an intern. A year and a half later, she was a full-time employee helping build Uber's competitive strategy in Australia. By the time she left, she had lived and worked across four continents.
The AWS chapter lasted less than a year - which is not a failure, it's a tell. When Reflection AI came calling, Casey moved again. This time into something rare: a company with an explicit, literal mission to build open superintelligence.
Reflection AI was co-founded by Misha Laskin, who led reward modeling for Google DeepMind's Gemini, and Ioannis Antonoglou, who co-created AlphaGo. The team is dense with former OpenAI, Anthropic, and DeepMind researchers. In October 2025, they closed a $2 billion Series B - backed by NVIDIA, Sequoia, Lightspeed, DST Global, and others - at an $8 billion valuation. That's a 1,367% valuation jump in seven months.
Casey joined as Chief of Staff to the CEO. Her first visible contribution: helping launch Asimov, Reflection AI's code research agent for engineering teams - a system that reads entire codebases, architecture docs, GitHub threads, and chat history, then builds persistent memory of your systems. It is, in miniature, a demonstration of the company's larger thesis: AI that understands context, not just syntax.
The Company
The premise at Reflection AI is not modest. "Building frontier open intelligence and making it accessible to all" - that's the stated mission. In a landscape where the biggest labs are closed, proprietary, and expensive, Reflection is positioning itself as both the American answer to DeepSeek and the open alternative to OpenAI and Anthropic.
The company's 120-person team has the credentials to back the ambition. AlphaGo came from this group. Gemini's reward modeling came from this group. When they say they know how to scale reinforcement learning, they have the receipts.
For Casey, the appeal is obvious: she spent four years evaluating AI companies from the outside and a year working with them at AWS. She knows what the research looks like when it's serious. This, apparently, is serious.
Career Arc
What's striking about Casey's career isn't any single jump - it's the pattern. Every move goes closer to the action. From a university campus to Uber's fastest-growing market. From Uber's growth functions to its strategy core. From operations to venture capital. From VC to the companies actually building the technology. From AWS to a lab that's trying to change what AI means entirely.
She is not a careerist rotating through impressive logos. She is someone following a genuine obsession with a very long way of doing it. That tends to produce people who are very good at the work by the time they arrive.
Perspective
Casey Flint's unusual value is that she has seen the AI industry from every seat in the room. As a VC at Square Peg, she met the researchers, evaluated the roadmaps, and watched which teams executed. As an investor, she learned that incumbents are best positioned to improve existing processes - and that startups need to build genuinely new things, not AI-layered versions of old ones.
At AWS, she saw the infrastructure layer - the compute, the cloud, the enterprise sales motion. She saw what it takes to get AI from a research paper to a company's production system. And she did it while running a newsletter that demanded she translate all of it into plain language for founders who didn't have time to read the papers.
That combination - research literacy, operator discipline, investor pattern recognition, and a writer's habit of explaining things clearly - is exactly what a Chief of Staff needs to be useful in an AI lab moving at frontier speed.
"AI (unlike the internet) is going to enable much, much more novel activity."- Casey Flint, writing in her "Artificially Intelligent" newsletter
The Product
The first thing Casey helped ship at Reflection AI was Asimov - and it's a telling choice of first act. Asimov is not a code-completion tool. It's a code-research agent: it reads entire codebases, architecture documents, GitHub threads, and chat history. Then it builds persistent memory of your systems and holds it in context while your engineers ask questions.
The distinction matters. Most AI coding tools help you write the next line. Asimov is trying to understand the whole program - the decisions behind it, the threads that led to the current state, the context that new engineers lose when senior ones leave.
That's a harder problem. It's also more interesting. And it is, in miniature, a demonstration of the company's core thesis about what superintelligence looks like in practice: not faster code generation, but deeper understanding of complex systems over time.
In the Media
"When something goes wrong...I need to know what went wrong to improve. Who wants to make the same mistake over and over again?!"- Casey Flint
Beyond the Resume
Pitching Advice (From the Other Side)
VCs are learning about you. Advisors block the signal. Leave them out of the first meeting.
Early-stage projections are art, not science. Present them like it. Overconfident models signal inexperience.
ARR, GMV, and revenue are different numbers. Calling GMV "revenue" doesn't end well in due diligence.
VCs don't sign NDAs at first meeting. Asking signals you don't know how the process works.
Early-round valuations are "far more art than science." VCs know this. Detailed valuation frameworks don't impress; they distract.
The Vision
Casey has described wanting to see AI "from chips all the way up to applications" - a phrase that reveals the shape of her ambition. She's not interested in one slice of the stack. She wants to understand the whole thing: the infrastructure, the models, the products, the companies, the people, the consequences.
At Reflection AI, she's as close to the whole stack as you can get. The company is training frontier models, building products on top of them, and trying to do it all in the open - meaning the research, the weights, and the safety work are meant to be accessible rather than locked behind a premium API.
Her Substack continues alongside the day job. "Artificially Intelligent" is how she processes what she's learning and how she stays honest about what's actually happening versus what's being marketed. 3,000+ subscribers read it for exactly that reason.
From Winton to the frontier. The outback produces people who know how to cover distance.