Serge Garden has a professional interest in the moment an impressive piece of technology meets an unimpressed buyer. It is an awkward introduction. Engineers have done their work. The demonstration behaves. The future seems close enough to invoice. Then someone with a budget declines to participate. Garden, the New York founder and CEO of AVB Invest, keeps returning to this encounter in his writing about building companies. A machine can perform beautifully while the business around it remains an unfinished sentence.
His career connects systems architecture, software services, startup operations and investment. That combination gives him several ways to approach the same question: what must happen between making something work and persuading somebody to pay for it? The question travels well. It follows him through accounts of synthetic avatars, a proposed repair-service pivot, and his more recent arguments about the people founders hire when sales refuse to arrive.
The architect acquires a sales problem
Garden’s public career history begins with systems architecture. He lists roles at LANIT-PARTNER in 2004-2005, ISG in 2006-2008 and Jet Infosystems in 2008-2010. He also lists Bauman Moscow State Technical University in his education. These are the technical foundations beneath a working life that would move steadily closer to customers, teams and capital.
In March 2011, his career record turns to QCoders, where he lists himself as founder and CEO through July 2016. The business offered outsourced, on-demand software development. Next came reOrder Technologies, from July 2016 to December 2017, with a retail infrastructure proposition involving interactions between customers and restaurant staff. He then lists a founder and board role at Spetz beginning in December 2017.
By January 2020, AVB Invest appears in that record, alongside WeInvest. Later entries include Starta Institute US and NeuraLoom. Read together, the sequence moves from the architecture of computer systems toward the architecture of businesses. There are more people in the latter diagram, and they are considerably less obliging about following specifications.
A beautiful demo, an absent customer
One of Garden’s published business stories concerns a synthetic-media team led by Anton Nikonorov. Looking back at its 2021 thinking, Garden describes digital avatars and speaking faces, with possible uses in brand communication and education. The technology had possibilities. The purchasing decision was less cooperative.
Garden reports a $900,000 prospective pipeline without completed purchases. His proposed change was to sell advertising integrated into livestreams, bringing the technology closer to an established spending category. He says the team never completed that direction. The striking part of the account is its ending: an attractive proposal remains a proposal. There is no ceremonial acquisition waiting in the final paragraph.
That willingness to leave an unsuccessful decision on the page gives the story its value. A founder’s retrospective can become a rather expensive costume party, with every accident dressed as foresight. Here the interruption survives. The distance between an interesting application and a finished business is visible.
“If the horse is dead, dismount.”
Serge Garden, on the unfinished synthetic-media pivot
Changing the customer before changing the code
Another account starts with a failed startup and an unfashionable ambition: try again using what already exists. Garden describes months spent reconsidering a repair-service business. The change he recounts was toward commercial trucks and a different model for supplying labor. His team called repair businesses, drivers, handymen and fleet operators to test the idea.
In Garden’s telling, funding followed, including a later $1 million check in California. He also recounts pitching Tim Draper. Those are his reported experiences, rather than a reason to turn an unnamed company into a new case study with an invented identity. The detail that matters to his argument is that the software had not changed during the commercial reframing.
The two stories make an interesting pair. In one, a technically appealing product could not make its proposed turn. In the other, Garden says a changed customer and business proposition altered the reception of existing technology. Neither story promises an easy formula. Both place the difficult decision somewhere outside the codebase, where developers cannot simply close the ticket.
The wrong hire can produce excellent work
In an August 2026 post, Garden brings the problem inside the founder’s hiring decision. A company needs a compelling offer. The founder hires a product manager. Everyone gets busy. Customer conversations lead to more development, while the original commercial difficulty persists.
Garden distinguishes the product manager’s attention to users and features from the product marketer’s attention to buyers and positioning. His warning is about assigning the problem accurately. A person can be capable, industrious and working on the wrong assignment. Adding another feature may improve the product without answering why a customer should choose it.
There is a particular sting in this argument coming from someone with an architecture background. The technical work retains its dignity; it simply cannot inherit every responsibility in the company. A founder who asks one role to supply several kinds of judgment may receive an impressive quantity of work and very little relief. The calendar fills up. The commercial question waits patiently at the door.
Building the machinery around the founder
AVB Invest puts Garden’s interest in company building into an organizational form. The business describes a combination of investment, shared operating resources, talent development and AI-assisted management. Its AutoVentureBuilder proposition centers on following project performance and reallocating resources. The ambition is to organize work across ventures rather than treat every new company as an entirely separate construction site.
The human component is visible in AVB’s training offer: remote participation, project work and mentorship. Its JobStarter partner app provides a way for organizations and mentors to follow internship activity and feedback. These are practical pieces of the larger proposition. People need useful assignments, someone to review their work and a way to understand whether they are progressing.
For Garden, that brings the career’s technical and commercial strands into the same frame. Software organizes activity; people supply judgment; a customer determines whether the result has value. Keeping those relationships working is a daily operating task. No diagram is likely to spare anyone the next conversation about an overdue deliverable.
AI still needs somebody to frame the problem
Garden’s interest in AI is hands-on. On Intch, he describes building initial prototypes with Python, SQL, language models and agent orchestration. The stated purpose is to establish a concept before committing expensive engineering effort. It is a recognizable continuation of his earlier work: make enough of the system tangible to discover what the next decision should be.
His ambition extends to AI agents that collaborate, build and transact within larger systems. In an October 2025 post, he said prototyping was already under way. The aspiration deserves its future tense. What makes it relevant to his story is the continuity: the systems architect is again asking how separate parts might cooperate, this time with software performing more of the work.
His writing also assigns considerable importance to the person asking the questions. He argues that an AI-capable professional must identify causes, structure a problem, supply context and evaluate the answer. Experience matters because someone must recognize a plausible response that has quietly missed the point. Speed creates more opportunities to decide; it does not supply the judgment automatically.
There is a revealing small admission in his account of using AI. He once saved its answers carefully into Markdown files, expecting a personal knowledge system to emerge. Later, he writes, he largely stopped collecting the outputs and kept asking questions. Even his own workflow becomes eligible for revision. For someone who spends so much time discussing pivots, it is a pleasantly domestic example.
That revision also gives his AI enthusiasm a useful scale. A founder need not begin with an autonomous company to test a different way of working. Garden’s own description begins with questions, prototypes and specifications. Those are small enough to inspect. They put a boundary around the experiment and give the person paying for it something concrete to judge. The same preference runs through his commercial stories: identify the person involved, understand the decision in front of them and find out whether the proposed change helps. A grand vision still has to survive an ordinary working afternoon.
The decision is the story
In May 2026, Garden began a numbered series of business decisions, introducing it with his account of more than a decade in venture building. The choice of subject is revealing. A company history can be assembled from job titles and launch dates. A decision brings the reader closer to the working day: the uncertain customer, the disputed direction, the idea somebody must finally abandon.
Garden’s public writing is especially useful when it stays with those moments. The promising technology, the changed buyer, the misplaced hire and the unfinished pivot belong to the same working life. Each asks whether the next effort will address the actual problem. He has spent years moving between the machinery of software and the negotiations of business. At their meeting point, there is usually another decision waiting.