AI is the new frontend
By "frontend" I mean the surface where product meets user, not the React stack. Today, product discovery and iteration happens with AI, UX has become User/Agent Experience and the roles of PMs, designers, SMEs, consultants and frontend engineers are converging through use of AI tools. The frontend is AI.
With an explosion of AI-enabled products comes new engineering problems. Backend engineers must rise to the challenge. We need engineers in New York City and elsewhere to work on the new platforms and distribution mechanisms that will safely deliver the AI solutions to customers.
Building on the new frontend
When AI-assisted coding arrived, it brought with it a new category of engineer: sometimes called Product Engineer, Applied AI Engineer or Forward Deployed Engineer (FDE). The role prioritizes the use of AI-assisted coding to rapidly produce customer solutions. Sometimes pitched as "half PM, half engineer" or "a micro-founder within a startup". In fast moving consumer or medium-sized B2B spaces, this approach proved an effective way to hook customers looking for the latest AI.
But custom point solutions, while they might lead to short-term growth, don't achieve long-term stickiness. The point solutions are hard to maintain. Expanding a product into adjacent use-cases is impossible. You've built cool demos, but they're brittle. There's no path to a broader platform play.
And for the individual engineer, you're now playing an entirely different game. Your engineering skills are no longer your core asset. Sales, product and design skills become the differentiator. PMs, consultants and Subject Matter Experts (SME) can all vibe code, and may bring more to the table where it counts. Sure, for the moment it may be easier for an engineer to stitch together output from Claude and Codex, but how long before the use of these tools becomes part of every PM, designer or consultant's toolkit?
At our company, Anterior, building an agentic platform for healthcare, there are doctors and nurses on the team who use the AI tools to build out customer workflows. With their medical training and specialized knowledge of the domain, they're able to speak with customers, understand the problems and use AI to iterate. You can take the smartest engineer in the room and they won't be able to understand the intricacies of a complex medical history, but with the AI tools the clinicians are able to build out agentic workflows and eval datasets without an engineer in the loop.
The new backend
Looking around in mid-2026, there's a crisis in software quality. Platforms that were a byword for reliability in the SaaS-era, are now fracturing under the weight of this front and backend realignment. AWS had outages in March and May, Azure had one in April and Cloudflare in February. GitHub had four incidents in March alone and in the same month McKinsey's internal AI platform was breached.
Requirements of correctness, reliability, performance, scalability and security have not gone away. In this new world of AI coding, agents and new user experiences, what are the protocols and security practices? How can data be protected and served to where it's needed? What happens when you can't trust the code? When an agent acting on a user's behalf hits an API, neither the API nor the auth layer can tell whether the request reflects the user's intent. The protocol to answer that question doesn't exist yet. These are weighty engineering challenges. The SaaS-driven architectural standards of 2020 are not nearly enough to get us where we need to go.
If companies are to take their AI solutions beyond surface-level UX improvements, or have the chance to play with secure data in real-world mission-critical workflows, they need to invest in the new backend platform layer. And yes, at Anterior we are building exactly this kind of foundational system for healthcare.
If some engineers have seen the AI coding tools and used them to move into product and sales, then others want to use the AI tools to become better engineers doing the kind of work where engineering judgment still matters more than throughput.
NYC: built different
New York City's startup ecosystem is at first glance poorly positioned for this shift. New York has always skewed toward consumer products or sales-led SaaS companies. Deep infrastructure is more commonly associated with Silicon Valley. FDEs at New York startups are leading the wave of AI solutions for enterprise. But, if long-term success means converting those AI point solutions into platforms, many New York startups are not well positioned to capitalize.
New York excels at building software for at least one industry: finance. Bloomberg, BlackRock, Jane Street, Hudson River Trading and many more have all built world-leading platforms in these streets. As AI becomes essential to succeed in business, I believe New York has a strong hand to play in leading the design of the software architectures to underpin that success, perhaps taking learnings on security, compliance and handling mission-critical data from those finance platforms.
For engineers entering the job market today, my advice is to stay true to the fundamentals. Your value within an organization has always been to bring the experience and judgment that only an engineer can. There's so much more we need you to build, and so much of it is below the surface. The frontend will keep changing. The platform is where you'll find the lasting value.
