The most expensive object in a software company may be the handoff. It has no line on the cloud bill and no purchase order. It simply sits between people. Product explains an idea to design. Design turns it into a polished picture. Engineering studies the picture, studies the existing code, and rebuilds the thing in a language the product can actually run. By the time it ships, the idea has enjoyed three careers.
Kevin Michael knows the ritual because he helped perform it at scale. For five years he was a founding engineer and head of product at Perpetua, the Toronto advertising software company that exited for roughly $150 million. The job placed him in the awkward and useful middle: technical enough to understand the machinery, product-minded enough to notice how much human time disappeared around it.
Perpetua could make prototypes quickly with low-code tools. That was the attractive part. The less attractive part arrived when a promising prototype had to join the real product. Engineers rewrote it. The demo had answered whether an idea was worth pursuing, but nearly none of its physical substance survived. Speed had produced a disposable object.
brief
mockup
code
The useful annoyance
Michael and Perpetua colleague Peter Gokhshteyn kept returning to the same annoyance. Why should a design tool describe the eventual interface without being the interface? Why should the prototype be a sketch destined for the bin? Their answer became Tempo, founded in Toronto in 2023 and accepted into Y Combinator's Summer 2023 batch.
Tempo is a visual editor for React, but that tidy label understates its politics. It argues that a product team should work on one artifact. A designer moves a component on a canvas, yet the thing underneath is code. An engineer can inspect its structure. A product manager can make an idea concrete without opening a ticket and waiting for the next ceremony. AI assists, but the code remains available for scrutiny and revision.
The ambition is less to make everyone identical than to make everyone adjacent. Michael expects product managers to become more technical and designers to become design engineers. Engineers, in turn, can spend less time reconstructing already-decided interfaces and more time on the difficult systems that give a product its advantage.
“We believe code is becoming a key medium for designers - a new kind of paint for the web, our shared canvas.”Kevin Michael, Tempo
The paint metaphor matters. Much of the AI coding market has treated prompting as the interface and generation as the event. Michael's version leaves room for the stubborn human desire to adjust one curve, one spacing rule, one component state. Tempo's canvas and design controls are built around the belief that a first pass is not a finished opinion.
That position became clearer as AI-generated software multiplied. If many teams can summon similar functionality, then design, distribution, and judgment carry more weight. The machine can remove blank-page anxiety. It cannot decide what a company ought to feel like, or which inconsistency is charming and which is merely careless.
Learning the machinery
Michael arrived at this argument through an unusually literal education in systems. At the University of Waterloo, he studied Systems Design Engineering and participated in the school's Velocity entrepreneurship community. As an undergraduate researcher, he co-authored a 2017 paper comparing two approaches to image segmentation. The paper's practical constraint was telling: it examined methods that could work with modest computing resources, including a CPU-only setting.
His early work crossed engineering and product before those roles became his company thesis. At Microsoft, he interned on the Visual Studio team. A product built with two other interns won the company's Best Intern Project award; his public profile says it later accumulated roughly 13,000 downloads. He then spent a term as a product fellow at Social Capital and founded Six Club before joining the group that built Perpetua.
The résumé reads like a tour of software's layers: the tool that engineers use, the investor's view of products, the founder's blank sheet, the scale-up's crowded roadmap. Tempo gathers those layers into one interface. The continuity is not a grand childhood prophecy. It is the more credible accumulation of irritations that refused to go away.
Support as fieldwork
The cleanest AI demos begin with an empty project. Real companies do not. They have old dependencies, strong opinions, naming conventions, half-migrated design systems, and components nobody dares delete. Tempo chose the untidy market of production codebases, which meant its founders had to learn how software behaves after the applause ends.
Michael responded with an unscalable habit: personally handling hundreds of customer-support inquiries in a week. The point was not martyrdom by inbox. Each question exposed a collision between the product's promise and somebody's actual repository. Stay close long enough and the recurring complaints become a roadmap. The founder becomes an anthropologist with admin access.
That customer contact also sharpened Tempo's boundaries. The product added ways to connect existing component libraries, work with familiar React and Tailwind projects, and bridge back-end services such as Supabase. It introduced templates with payments, authentication, and databases already wired together. The aim was not simply to make a handsome landing page appear. It was to help the handsome landing page survive contact with a business.
What faster is for
Tempo often describes its promise with the familiar startup phrase “10X faster.” Michael also uses a more revealing measure: fidelity. A rough sketch is useful because it is cheap to change. A working prototype tells a better truth. Production code is where the idea meets every constraint the sketch politely ignored. Traditional workflows climb this ladder by rebuilding at each rung. Tempo tries to keep the object intact as its fidelity rises.
This reframes speed as the number of learning cycles a team can afford. A quicker first draft matters, but a quicker correction may matter more. If a designer can test the real component, a product manager can demonstrate behavior instead of describing it, and an engineer can modify rather than reconstruct, then the team gets another chance to be wrong before customers have to notice.
There is an organizational consequence too. Tools encode permission. A design file tells engineers where to look but not always where to act. A code editor grants enormous power behind a forbidding wall of syntax. Tempo's visual surface attempts a narrower bargain: more people can make meaningful changes, while the underlying React structure remains visible and owned. The question is not whether everyone becomes a senior engineer overnight. It is whether a good idea has to wait for the one person who knows which file to open.
“It’s the design and prototyping tool we always wished existed - one that brings the entire product team into the build process.”Kevin Michael, on Tempo
Conviction, with a ruler
Michael's public comments about fundraising carry the same preference for control. He has described investor enthusiasm for AI code as fickle and emphasized staying convicted when sentiment moves. Tempo turned down more money than it accepted during an earlier fundraising period, choosing to avoid the pressure and distorted incentives that can arrive with too much capital too soon.
In October 2025, the company announced a $5 million seed round involving Golden Ventures, Y Combinator, BoxGroup, Webflow Ventures, iNovia, General Catalyst, and several individual investors. Tempo also said more than 100,000 people had built something with its beta. The money was assigned to a decidedly concrete mission: expand the engineering team and keep developing the product.
By 2026, Michael was publicly recruiting designers and design engineers. The titles are evidence of where he thinks the work is heading. The old boundary said designers propose and engineers produce. The new role is comfortable with pixels and pull requests, composition and constraints. It does not abolish expertise. It gives expertise a wider desk.
A workshop, not a relay race
Michael does not present AI as the extinction event for software developers. His expectation is almost the reverse: when repetitive coding becomes cheaper, demand for software rises. Engineers spend more attention on intellectual property and complex problems; more people participate in building; the number of useful experiments increases.
The unresolved problem is coordination. Faster generation can produce faster disagreement, more orphaned prototypes, and a great many interfaces assembled from the same fashionable ingredients. Tempo's answer is to place the participants and their AI collaborators on a common surface, with production code as the record everyone shares.
There is a pleasing modesty hidden inside this large claim. Tempo does not ask an organization to believe that language has become unnecessary. It asks whether teams could translate less. Let the product manager touch the idea. Let the designer touch the implementation. Let the engineer preserve the system instead of tracing a screenshot. Keep the human hand available when the yellow is wrong by two shades.
For Michael, the future of software creation looks less like a sequence of departmental approvals and more like a workshop: people with different crafts standing around the same object, changing it while it is still warm. The handoff may not disappear entirely. But it could lose its throne as the place where good intentions go to wait.