The first public greeting Austen Collins remembers receiving for his new framework was a Hacker News comment calling it a horrible idea. He tells the story with the dry generosity of someone who understands the venue: perhaps that is simply how Hacker News says hello. The project had a shark for a mascot, a name borrowed from a blockbuster, and a promise that sounded suspiciously neat. It proposed that developers could build whole applications on AWS Lambda while spending far less time managing the machines underneath.
This was 2015. Lambda was new, API Gateway was new, and the vocabulary around event-driven cloud software had not settled. Collins was not arriving from the usual infrastructure lineage. His background included design and motion graphics, followed by work as an AWS consultant. He could write the tool, but he also noticed something engineers sometimes treat as secondary: a technical change needs a story before it can become a movement.
He called the project JAWS, originally an acronym for JavaScript AWS Framework. The README promised a “monstrously scalable serverless application framework.” The logo showed a shark breaking the surface. He even edited a trailer from clips of Steven Spielberg’s film, cutting shark-point-of-view footage against titles about a framework coming for your infrastructure. AWS CTO Werner Vogels retweeted it.
The camp was theatrical. The desire beneath it was plain. Collins wanted to make software without devoting so much of his day to the machinery that kept it alive.
“I simply wanted to build more and manage less.”Austen Collins, describing the theme behind Serverless Framework
The word that made people lean forward
When AWS introduced Lambda, it often described the service as a way to run event-driven code or connect other services. Collins wondered whether it could support an entire application. The process was rough. The infrastructure was young and the convenient patterns did not exist yet. Then he found a sentence in an AWS Compute Blog post by Lambda general manager Tim Wagner. It suggested that developers could build “entirely serverless applications.”
Collins liked the word before its meaning was stable. “Serverless” was not a literal description. Servers still did the work. But it expressed the experience he wanted: technology that moved out of the maker’s way, scaled when needed, and did not produce a bill while idle. The term turned an implementation detail into a compact emotional promise.
He began demonstrating JAWS almost weekly at the AWS Loft in San Francisco. Before one presentation by Wagner, Collins approached him and asked if he could show the audience his application framework. Wagner said yes. The enthusiasm was difficult to miss. Later that year, the Lambda team invited Collins to present at AWS re:Invent.
The attention brought a practical problem. JAWS was also the name of an important screen reader used by blind and low-vision computer users. Collins began receiving one-dollar donations accompanied by messages asking him to change the name. He understood the conflict and moved quickly. Of the possible replacements, “serverless” was the one that made developers react. It sounded odd to investors. That was part of its pull.
Getting the matching domain took about two months of cold-calling to locate its owner. By the end of 2015, JAWS had become Serverless Framework and the company became Serverless, Inc. The technically imperfect category word now sat on the product, the company, and the dot-com.
A framework designed to be interrupted
Open source changed the job. The framework went viral, Collins recalled, and then the issues arrived. Startups and enterprises wanted different workflows. Developers held strong preferences about languages, providers, file structures, and deployment habits. A closed product team could either defend one path or build a place where the disagreements could produce working software.
Collins chose a plugin architecture early. Version zero accumulated roughly 70 plugins. Today the project’s repository describes an ecosystem of more than 1,000. Extensibility was not only a feature. It was an admission that no central team could anticipate the field.
Feedback came through GitHub, email, forums, chat, and social media. In 2016, Collins said a comment arrived at least every few minutes. His way of sorting the noise was unusually editorial. He described himself as theme-based: identify the theme, then judge each choice by how well it expresses it. For Serverless Framework, the theme stayed “build more and manage less.”
That principle helped the project move without pretending the underlying cloud had stopped moving. Version 1 leaned on AWS CloudFormation for stability and added support for other providers. The framework grouped related functions and their infrastructure into a “service,” usually described in a readable YAML file. Collins summarized the tension cleanly: a function is powerful as a unit of deployment and painful as a unit of development. The product’s work lived between those two truths.
- Name the burden the user wants to stop carrying.
- Turn that relief into a memorable product theme.
- Let the community extend the workflows you cannot predict.
Collins’s unusual range mattered. At a 2017 RedMonk talk, he argued that packaging is a story and a vehicle for a product’s message. This is not a claim that story can rescue weak software. His own history suggests a tighter relationship. The shark, the word “serverless,” the live demos, and the plugin ecosystem all made the same promise at different levels: difficult infrastructure could be presented as a coherent experience.
There was also some comic instinct in how he sold the future. One early fundraising deck included an image of Docker’s whale upside down in the ocean with a bite missing. Just before an investor meeting, Collins checked the firm’s portfolio, saw its connection to Docker, and removed the slide. Product storytelling requires taste. It occasionally requires a last-minute edit.
Building a company around an open door
In October 2016, Serverless released version 1 and announced a $3 million seed round led by Trinity Ventures. The company had twelve employees at the time. Collins had been an AWS consultant before starting it, so the commercial problem was familiar: businesses wanted to ship more quickly, while cloud services presented a growing set of knobs, permissions, regions, and dependencies.
Two years later, Serverless announced a $10 million Series A led by Lightspeed Venture Partners, with Trinity also participating. The company paired the open-source framework with tools for visibility, delivery, and integration. Collins said the framework and gateway would remain open source. The business would help teams operationalize the approach safely and consistently.
The product expanded beyond deployment. Monitoring, testing, secrets management, security, and observability entered the experience. In 2022, Collins announced that Serverless Cloud would spin out as Ampt while Serverless, Inc. stayed focused on the framework and observability. Version 4 became generally available in 2024 with native TypeScript support and integrations for tools including Terraform and Vault.
The roadmap also widened the meaning of serverless. A container framework introduced in February 2025 let developers move API workloads between AWS Lambda and ECS Fargate without rewriting the application. Two months later, Serverless MCP brought AWS state, logs, metrics, and documentation into AI-assisted coding environments. The abstractions changed. The old theme remained recognizable.
A different kind of creature
In February 2026, Collins showed a new early-beta project called Creature. It is a desktop environment for MCP apps, the visual interfaces that can live inside AI conversations. Users can build tools, share them with a team, open them in tabs, rearrange them, and let an agent work across several at once. Collins described the approach as closer to an operating system than a set of chat widgets.
The technology is different from Lambda. The product instinct is not. Generic software asks a user to adapt to a fixed interface. Collins is betting that personal AI tools will become specific to a person or team, assembled and operated with help from an agent. Creature tries to make that software visible and manipulable, not merely another command typed into a terminal.
“We think personal AI tools are going to replace a lot of generic software.”Austen Collins, introducing Creature in early beta
There is a pleasing loop in the name. The first project borrowed a creature from cinema to make invisible cloud infrastructure feel dramatic. The current project gives AI-native software a creature of its own. Collins has not presented the beta as finished. He warned that it had rough edges. That candor fits the builder who posted an early framework to Hacker News and learned from the sharp replies.
His public biography remains notably focused on products rather than personal mythology. The artifacts do the explaining: the playful launch, the weekly demos, the architecture that welcomed plugins, and the repeated effort to place complex systems behind a clearer surface. Even his LinkedIn summary begins with “making meaning, value and iconic companies.” Meaning comes first.
The useful lesson in Collins’s story is not that every infrastructure tool needs a mascot or a cinematic trailer. It is that abstraction is both engineering and communication. Code removes steps. Language removes hesitation. Community reveals the workflows a founder cannot see alone. The product becomes real when those pieces reinforce one another.
Servers never disappeared. Neither will agents, models, permissions, or the rest of the machinery behind personal AI. Collins’s decade of work suggests a more practical ambition: make the machinery recede far enough that a person can concentrate on the thing they came to create.