LATEST ANDO LEAVES STEALTH · SEPTEMBER 24, 2026$20M ACROSS PRE-SEED + SEED · ACCEL / INDEX / EMERGENCE

The builders / Sara Du

Sara Du and the trouble with the room

She left Harvard, cofounded Alloy Automation, and learned that building a company could also reveal the work she wanted to do next. At Ando, Sara Du is redesigning the place where people and AI agents meet.

Sara Du wore ripped jeans to her Harvard admissions interview. Years later, after requesting her admissions file, she discovered that the jeans had attracted rather more attention than she expected. In her telling, the committee’s doubts reached the dean of computer science along with a sample of her code. His response concerned Nobel Prize winners and their clothes. The code, apparently, deserved a hearing of its own.

It is a pleasingly small detail in a life that is often compressed into large labels: Harvard dropout, startup founder, Thiel Fellow. A pair of jeans makes the story easier to recognize. Institutions have their preferences. So do people who build things. Sometimes the disagreement begins before anyone has opened a laptop.

Du would spend a year at Harvard and leave. In February 2022, she announced that Alloy Automation, the company she cofounded with Gregg Mojica, had raised a $20 million Series A. She had expected to graduate that spring. Instead, she was explaining a different kind of progress to her parents, who had immigrated in their early twenties. When she called after signing the term sheet, she heard them say they were proud.

The work arrived before the company

Her route into software began in Georgia. During the first half of high school, her time went into basketball, swimming, and golf. Her younger brother’s computer science classes stirred a competitive impulse. She began teaching herself C++, stopping and starting as the difficulty became apparent. Sibling rivalry has funded worse hobbies.

The internet widened her circle. In 2017, while she was still a high-school indie hacker, the Thiel Foundation invited her to a summit. Meeting people there, including DoNotPay founder Joshua Browder, influenced the direction she took. The invitation preceded her formal fellowship by years. A community could become useful well before it supplied a credential.

She left high school early to work at DoNotPay in San Francisco, then gave university a try. Before Alloy, her experience also included engineering at Snap and design at Wish. Those jobs put her on both sides of a familiar software question: what can be made to work, and what will someone actually want to use?

Alloy began with requests from friends. Du and Mojica had met through the open-source developer community and worked on side projects together. Some friends had become influencers and needed custom code for their stores, including a way to move Instagram photographs onto Shopify. The repeated favors suggested a product: let store owners connect their tools without hiring someone to write every connection.

Sara Du working on a laptop at the Ando office
The room, under construction. Sara at work on Ando. Photograph: JC Photography.

There is something instructive about an origin this ordinary. A founder helps someone with a task. Another person needs a similar task done. Eventually, the favors start resembling a job description. The interesting moment is deciding which parts can become shared software and which parts belong to the particular person asking.

Frozen burgers, connected stores

Du and Mojica founded Alloy in 2019 and joined Y Combinator’s winter 2020 batch. Their first product tackled ecommerce automation. That focus mattered: a store’s order contains line items, tax codes, and other nested information that a general-purpose workflow might handle awkwardly. Knowing the shopkeeper’s problem could be more useful than offering a longer menu of actions.

In an early public recollection, Du described a holiday spent with her cofounder, eating frozen burgers while building the first version. A year later, she was celebrating new employees and dozens of new customers. The menu had improved too. Startup mythology tends to prefer a garage; sometimes the freezer supplies the more honest landmark.

Funding was less straightforward. Alloy struggled to raise its first $1 million as investors withdrew from deals at the beginning of 2020. The founders had imagined stretching that money over five years. The Series A two years later changed what they could attempt. Money buys time, but it also asks a company to decide what to do with a larger future.

2019Alloy beginsCofounded with Gregg Mojica
2020YC W20Ecommerce automation
2022$20M Series AAlloy’s financing
2026Ando launchesA new messaging workspace

By the end of 2022, Alloy’s public year-end update described a team of 30 across six countries, more than 220 integrated applications, and hundreds of thousands of workflows running daily. It had launched Alloy Embedded, extending the work of connecting software into customers’ own products. A tool that began with merchants’ requests was becoming infrastructure.

In 2023, Alloy expanded beyond ecommerce. Du described its customers as product and engineering leaders at mid-market and enterprise companies. The company’s relationship with Amazon brought another scale of integration work. The original question - how to make separate pieces of software cooperate - had survived while the customer and the machinery grew more complicated.

Du and Mojica also appeared on Forbes’ 2023 30 Under 30 list. Recognition supplies a clean stopping point for a biography. Actual work rarely observes the courtesy. A founder can become good at running a business and still discover that the business asks for a kind of effort she would rather direct elsewhere.

A familiar face, an empty response

One of Du’s more revealing pieces of writing began with a social embarrassment. In October 2023, she counted 2,313 unique people she had met on Zoom over the previous year. At an investor’s happy hour after a Europe trip, someone greeted her enthusiastically. She could not place him. Another guest’s introduction supplied his name and still failed to unlock the memory.

She described the failure in a programmer’s language: “Empty response body.” The joke is precise because the experience is familiar. You have received the request. You know a response is expected. The information is somewhere, probably. Meanwhile, the other person is standing there.

A year of introductions2,313

Unique people on Zoom, by Sara’s own count in October 2023.

Her essay turned that awkward encounter into a question about remembering people and designing tools to help. The small failure seems relevant to her later work: information may exist without being available at the moment it matters. In a company, that gap can mean repeating a conversation, asking the same question, or reconstructing a decision from scattered messages.

It also reveals her habit of examining the interface between a person and a system. The story begins with an evening out. It ends with product considerations. A dinner guest might reasonably wish to leave an embarrassing moment behind. Du gave hers a title and published it.

Letting the agents into the conversation

After leaving Alloy amicably, Du spent time reading papers and experimenting. In early 2025, she built agents and Model Context Protocol servers for companies. MCP offered a general way for language models to use tools. Customers wanted those agents inside Slack, where their teams already worked. Making the arrangement useful proved harder than installing another bot.

The friction helped produce Ando. Du became convinced that permissions and inefficient access to context reflected assumptions in the messaging platform itself. An agent doing work needs a place in the conversation, an identity, and a clear understanding of what it may see and do. Those requirements shaped her next company.

Ando emerged publicly on September 24, 2026, with $20 million raised across pre-seed and seed rounds. Accel led the pre-seed; Index Ventures and Emergence led the seed. The financing belongs to Ando, a separate chapter from Alloy’s earlier $20 million Series A. Identical numbers can conceal very different questions.

Ando’s stated product principles include limiting interruptions and letting useful context accumulate. Its agents can join conversations and take action, while private direct messages remain outside their view unless someone forwards the context. This is a practical design issue as much as a technical one. A helpful participant needs boundaries. So does a tolerable one.

For Du, the attraction includes working on a product she uses herself. The feedback is close at hand. Each notification, sentence, and interaction becomes something the builder has to live with. A messaging tool earns its place repeatedly throughout the day, often while its users are trying to think about something else.

What deserves to stay?

Her tastes help explain the attention she pays to the experience. She names Hemingway and Murakami as favorite authors, and the electric blue #002FA7 as her favorite color. Ando takes its name from architect Tadao Ando. Its company philosophy treats software as an environment people inhabit, with room for attention as well as activity.

In January 2026, Du wrote about editing a personal video. Rearranging clips, adding meme effects, and timing music helped her discover the story she wanted to tell. She could not specify the finished result beforehand. The experience sharpened her argument that some interfaces must support exploration alongside commands. Sometimes making the thing is how you learn what the thing should be.

“Software is so cheap to write these days, but patience has become more expensive.”Sara Du, July 2026

Her July essay on lasting software asks what happens when generating code gets easier than understanding the structure it creates. She argues for patient reading and a willingness to remove work that no longer belongs. The position gives her enthusiasm for agents some useful texture. Producing more software introduces responsibilities for the people who will inherit it.

Alongside building, Du invests through a small fund called *Asterisk. Since 2019, she has backed a few dozen companies and offers help with marketing, brand, and design. The name folds together exception, wildcard, multiplication, and risk. It is a compact expression of how she thinks about founders: a person’s particular strengths can change the possibilities of a business.

Her own second company puts that idea to work. The task is demanding and wonderfully unglamorous in its details: make a conversation useful, let a colleague concentrate, give a machine enough information without giving it every room in the house. The founder who once helped friends connect their stores is now deciding how people and software should share the room. The furniture is still being arranged.