The useful clue was hiding inside somebody else's dashboard. At Datadog, Brad Menezes was responsible for Application Performance Monitoring, a product that helped engineers see what their software was doing. Customers kept showing him the things being monitored: onboarding systems, finance dashboards, support consoles, operational tools. They were unglamorous, custom, and indispensable. They were also backed up in queues because engineering teams had more requests than hours.
A less curious product leader might have celebrated the monitoring sale and moved to the next call. Menezes noticed the backlog behind it. Datadog could reveal whether an internal app was healthy, but it could not make the missing apps appear. Somewhere between the dashboard and the queue sat a company idea.
That idea became Superblocks, the New York enterprise software company Menezes co-founded with Ran Ma in 2021. Its current proposition sounds perfectly timed for the prompt era: tell an AI agent what internal application you need, refine the result visually or in code, and let IT control permissions, data access, audits and deployment. The sentence is brisk. The career that produced it was anything but.
One reply can redraw a career
Menezes studied mechanical engineering at the University of Toronto from 2007 to 2012. The degree supplied a way of breaking complicated systems into parts. It did not supply an obvious invitation into Silicon Valley product management. So he manufactured one. He has said he cold-emailed roughly 100 companies after university. Yelp replied and hired him onto its product team shortly after the company's public listing.
At Yelp, he worked on reservations and delivery and got an unusually close view of founder-led product making. The experience offered more than a recognizable line on a résumé. It connected an engineering education to the daily work of choosing markets, talking to customers and turning imperfect information into product decisions.
Then he tried to become a founder. He teamed up with Ma, a friend he had met in high-school physics class. They had begun as academic rivals, which is a charmingly efficient way to test both ego and stamina. Their first company entered Y Combinator's Summer 2015 batch and had early backing. It still failed.
“The market you enter is the highest leverage decision in your career.”Brad Menezes, career advice published in 2020
Menezes does not varnish the postmortem. The founders lacked the deep industry advantage required to win. Failure gave them a conclusion more useful than generic resilience: enthusiasm cannot substitute for a market edge. Rather than immediately spinning up another company, they went looking for one. Menezes joined Datadog. Ma joined Confluent. Each entered an enterprise software business where the hard parts lived in production, not in a pitch deck.
A startup inside the startup
Datadog hired Menezes in 2017 to work on APM, its first significant expansion beyond cloud infrastructure monitoring. He describes the job as building a startup inside a startup. The product team had to establish a market, recruit colleagues, earn customer trust and prove that Datadog could expand from watching machines to tracing what applications did across complex systems.
By Menezes' account, the business grew from no revenue to $100 million in annual recurring revenue during his four years. Okta placed him fifth on its 2020 Up-and-Comers list, crediting his leadership of the product vision, customer relationships and product team. The recognition matters less as a trophy than as evidence of the missing ingredient from his first company: an earned view into a market with expensive, persistent problems.
The Superblocks thesis did not arrive as an ecstatic napkin sketch. Menezes and Ma tried to disprove it. They spoke with potential customers and examined the reasons a platform for internal software could fail. Scar tissue had become a research method. Only after those conversations held up did they commit.
Their relationship carried a second kind of infrastructure. Menezes leads product, sales and go-to-market. Ma leads engineering. Product is the large patch of shared ground. Menezes describes trust as the first requirement of a co-founder relationship, and their trust has been tested by more than a decade of rivalry, friendship, one failed venture, separate apprenticeships and another company.
When the wave changes the boat
Superblocks began as a low-code internal app builder. Its customers could connect databases and APIs, assemble interfaces and add custom code. Then generative AI changed what users expected from software creation. A prompt could produce an application-shaped object in minutes. The old distinction between developer and user began to look negotiable.
Menezes and Ma chose an awkward response: rebuild while their existing product was working. They shifted the underlying representation of an app toward code and redirected much of the engineering organization around AI generation. Clark, their AI agent, became the public face of that bet in 2025. It could generate React applications from natural-language instructions, after which users could edit visually or work in an IDE.
Menezes' most revealing observation about AI products is that the system prompt may account for only 20 percent of the result. The rest is prompt enrichment: the context attached to a request, the tools a model can use, and the checks applied before and after it acts. In his formulation, the glamorous incantation is a minority shareholder. Infrastructure still owns the company.
“You basically have to speak as if you would to a human co-worker.”Brad Menezes on instructing AI systems
That helps explain why Superblocks concentrates on enterprise internals. A consumer app builder can optimize for the pleasing astonishment of first creation. An enterprise builder has to survive the second meeting. Which database can this touch? Does the user have permission? Can security inspect it? Is there an audit trail? Can the app run inside the company's cloud environment? Production has a talent for replacing magic with paperwork.
Superblocks announced a $23 million Series A extension alongside Clark in May 2025, bringing the total Series A financing to $60 million. The launch campaign included a collection of system prompts from prominent AI tools and spread widely on social platforms. Customers publicly named around the period included Instacart and Papaya Global. Attention arrived, but the more interesting demonstration happened inside the company.
Menezes told TechCrunch that Superblocks' software engineers are not supposed to write its internal tools. Business colleagues build agents for lead identification, support metrics and sales-engineer assignments. It is a neatly severe form of dogfooding: if the platform cannot serve the company selling it, the engineers do not get to rescue the premise with bespoke code.
The grown-up is a control plane
By 2026, Superblocks was pushing further into governance. Version 2.0 centered on AI-generated enterprise apps under IT control. Subsequent releases added ways to import apps from consumer builders, monitor the software estate through MCP, manage AI spending, and put app databases inside AWS or Snowflake environments. In August, the company announced Superblocks 3.0 with AWS, emphasizing private VPC deployment, model routing and security agents.
The sequence reveals Menezes' larger ambition. He is not merely trying to make developers faster. He wants domain experts in finance, operations and other teams to become builders, while central IT retains a map of what exists and a hand on the permissions. The potential prize is a class of software that was previously too custom to buy and too low on engineering's list to build.
There is a paradox here worth keeping. Superblocks argues for democratized creation and centralized governance in the same breath. The cheerful version of AI software says everybody can build. The enterprise version adds a footnote in permanent ink: everybody operates inside a system. Menezes is wagering that the footnote will be where durable value accumulates.
His path makes the wager feel less like trend-chasing. Mechanical engineering taught decomposition. Yelp taught product. The failed startup taught domain humility. Datadog supplied scale and the original customer clue. Ma supplied a partnership capable of surviving a rewrite. Even the cold emails fit the pattern: when the official route does not exist, assemble one from available parts.
AI may make the first draft of software nearly free. The next questions remain stubbornly expensive: whether it is safe, useful, connected and allowed. Brad Menezes has built his company around those questions. In a market intoxicated by instant creation, he is selling the morning after, complete with access controls, audit logs and a surprisingly tidy kitchen.