Closing time, Heroku
Every new beginning comes from some other beginning's end
While the SaaSacre has dominated headlines this month, another major development has been rippling through the software world, one that hits especially hard if you’ve spent time in the developer ecosystem. Late on the Friday afternoon of “Claude Crash” week, Salesforce quietly announced it would end new enterprise contracts for Heroku and move the platform to a “sustaining engineering model”. While the announcement post was inconspicuously titled as a product update, for millennial app builders, the news felt more like the end of an era, particularly as Heroku had sun-setted its popular free tier just three years earlier.
As I’ve written previously, innovation in the developer economy has always been restless with platforms rising and falling constantly. This is part of what makes the dev realm such a dynamic area to invest in as a venture investor. But Heroku’s fade carries more weight than most. At its peak, Heroku wasn’t just one of the most beloved developer platforms of all time; it inspired a new standard for how app building should feel: inclusive, intuitive, and magical. As Meagan Gamache of Render writes in her post “Heroku changed everything”:
For nearly two decades, Heroku was where you shipped your first app, launched your startup, and fell in love with building for the web. It earned a place in the hearts of an entire generation of developers. Heroku didn’t just build a platform—it proved that powerful developer tools could be simple. Before Heroku, deploying an app meant wrestling with servers and manually configuring infrastructure. Days of work that had nothing to do with your actual product. The founders believed deploying should be as simple as git push. That idea changed the industry.
For me, the nostalgia is deeply personal. I remember the days of crashing at 650 7th St with my SalesforceIQ team so we could sit alongside and absorb the energy of the Heroku OGs. In the 2010s, everyone at Salesforce knew Heroku’s office had the best snacks, but more than that, it had a kind of creative gravity. You could feel that something important about the future of software was being figured out there by developers who loved developers.
While Heroku has ceded developer mindshare over the past two decades, the standards it pioneered will endure. In fact, AI is accelerating the trajectory of developer principles that Heroku set in motion. Next-gen dev platforms that don’t take heed and evolve will face the same fate as this once dominant PaaS platform.
For instance, consider how AI is democratizing software development to levels unimaginable in Heroku’s heyday. About eighteen months ago, I wrote about how more stakeholders, especially non-technical ones, are now empowered to participate in product development since natural language can be used to develop code. Then nine months ago, I highlighted how the definition of a developer continues to be expanded with AI coding agents entering the arena.
This rapid transformation has profound implications for what success looks like for developer platforms in the AI age:
DevEx: As the definition of developer widens further and AI agents become “first-class” citizens of developer tools, platforms will need to curate an outstanding agent experience alongside human experience. For instance, documentation needs to be easily machine-readable.
UI and onboarding practices: The best dev platforms have always felt as frictionless as consumer products, and now that standard must extend to agentic onboarding, implementation, and UIs. For instance, how should IDEs evolve for human-agent collaboration, or for when agents swarm or supervise other agents?
Virality and word-of-mouth loops: Developer mavens have always been the best marketing channel since authentic evangelism drives adoption. Elite developers also often serve as high-signal influencers. In the AI age, new questions are arising about how the next generation of network effects will look like: how do agents discover tools? How do they endorse tools to other agents? How do agents collaborate around shared toolchains?
Dev Rel: Will traditional methods such as events, content, and community channels still drive network effects in a world dominated by AI coding agents?
Trust and reliability: This is a sacred covenant with developers that will always be table-stakes. What will be the new standards here around AI provisioning, governance, and entitlement?
Pricing models: Many best-in-class dev platforms have historically relied on granular, metered usage-based pricing. Will the new levers of pricing power shift toward outcomes, agent productivity multipliers, or orchestration-layer control in the AI age? How do builders design natural upsell triggers for AI agents?
Budgets: AI tooling is already escaping traditional “developer budget” silos and pulling from broader operational and innovation budgets. Should every developer be augmented with an “AI coding” tool budget? How will spending evolve as AI agents proliferate?
New productivity curves: As I’ve previously written, AI coding tools have taken the tenet of abstraction to new heights. As agents and AI coding tools continue to remove the “mundane” layers of dev work (e.g. implementation, bug fixes, etc.), what happens to the gap between a 10x engineer and a regular one? Should developers specialize or stay generalist?
We’re not in an incremental tooling upgrade cycle with AI; we’re experiencing a wholesale rewrite of how software creation works:
As the developer economy accelerates into a phase of unprecedented reinvention, perhaps the most enduring lesson from Heroku isn’t merely about platform design — it’s about anticipating where the center of gravity of software creation is heading, expressing that opinion, and having the conviction to build for that future before it is obvious. RIP Heroku, may your spirit live on.