Anthony Minessale II and the trouble with a perfect demo

Anthony Minessale II’s keyboard froze during his ClueCon keynote in August 2026. He was demonstrating technology to a room full of people who build communications systems. The equipment had picked an excellent audience and an inconvenient moment.

He recovered the demo. Later, he included the mishap when sharing the talk. A founder could have posted the polished argument and left the keyboard out. Minessale kept the troublesome detail. It belonged in a presentation about the difference between a system that performs beautifully under friendly conditions and one that has to work when circumstances become rude.

That distinction runs through his career. The founder and CEO of SignalWire, based in Brookfield, Wisconsin, created FreeSWITCH and founded the ClueCon developers’ conference. His work concerns the machinery underneath a conversation: how calls connect, how audio moves, how software decides what happens next. His current subject is voice AI. The old telephone problems have come along for the ride.

A browser full of possibilities

In the early 1990s, his computing world included a university shell account, an IBM terminal, file transfers and multiplayer text games. The web browser expanded the view. By 1996 he was working in web hosting, looking at a web server, a database and the possibility of running a program whenever somebody clicked a link.

He tried building browser email, tools for making websites and software for managing customer relationships. Other people’s versions of such ideas became familiar businesses. His account of those years leaves room for the experiments that did not become household names. The attraction was the ability to put the pieces together and see what they would do.

Eventually the hosting business needed telephone calls to join its web tools. Staff should be able to work from different places, answer calls and see information about the caller. Asterisk, an open-source telephone system, offered a way in. Getting all the parts to cooperate supplied his next education.

The answering machine grew ambitions

Minessale’s early telephone experiments included making a desktop PC serve as an answering machine. There is a pleasing extravagance to using an entire computer for a job already handled by a small plastic box. The point was that software could change what the box did.

He became a contributor to Asterisk. As his ambitions expanded, so did the difficulty of fitting them into its design. He and other developers spent months discussing a different approach, twice talking themselves out of proceeding. The new project changed names three times before becoming FreeSWITCH.

Those hesitations matter. An origin story can look inevitable after the software has shipped. At the time, there were existing tools, limited resources and the prospect of considerable work. Starting another project meant accepting responsibility for an additional set of problems. FreeSWITCH began with people who had already spent enough time inside the old problems to describe the new ones rather precisely.

Writing it was only the beginning

He began coding FreeSWITCH in 2005; the project opened to the public on January 1, 2006. Its subsequent life involved maintaining the software, leading development and explaining how other people could use it. A released program creates a new constituency: everyone who now depends on it.

Minessale also became a technical author. His books include collaborations on FreeSWITCH 1.0.6, FreeSWITCH Cookbook and Mastering FreeSWITCH. The last, coauthored with Giovanni Maruzzelli, appeared in 2016. Documentation and books belong in this story because the work repeatedly returns to the distance between a tool existing and another person being able to do something useful with it.

His engineering career included Barracuda Networks, where he oversaw development of the CudaTEL telephone appliance using FreeSWITCH. That placed the open-source engine inside a commercial product, with the obligations that come when somebody expects to plug in a device and get on with their day.

The room around the code

ClueCon started in 2005, giving developers a place to meet around the communications technology they were making. Brian West and Michael Jerris became fellow creators of FreeSWITCH. A community supplied something a download could not: people who could challenge a design, bring another problem and carry the work into a different setting.

The conference and the software developed together. A project that can support many applications needs conversations about what those applications actually require. A meeting of developers also makes it harder to confuse an elegant design with a useful one. Someone in the room has usually tried the awkward case.

In ClueCon’s account of the project’s evolution, SignalWire follows the same effort to make telecom accessible. The audience grew beyond people willing to become specialists in the underlying engine. That creates an interesting obligation for a creator: success brings users whose first request is that they should need to understand less of your invention.

A useful interface is an act of hospitality

Long before SignalWire, Minessale wrote an Apache module that could make a new hosted website work when its directory appeared in the right place. A repetitive configuration task became a convention. The operator did less because the software understood more about what the operator intended.

His fascination with APIs follows that logic. An application programming interface lets one piece of software ask another to do something. The quality of the interface determines how much of the machinery the next programmer must understand. An interface can offer an invitation or a small entrance examination.

FreeSWITCH exposed controls through modules, network connections and language integrations. Minessale’s design interests kept moving outward, toward programmers working somewhere else, in another language, with another application. The telephone engine had to be useful beyond the desk where it was built. That requirement eventually became a business proposition.

The mechanic’s dilemma

Before SignalWire, the team earned money helping enterprises use the technology. In a 2020 interview, Minessale put the earlier consulting business’s first revenue milestone at $2 million, from roughly a hundred enterprise customers. Those are his historical figures for the services business, rather than a statement about SignalWire’s current revenue.

The work demonstrated demand. It also revealed a constraint: each customer could need knowledgeable people to help assemble and maintain a system. Growing the business meant supplying more expertise. The team wanted to put more of that expertise into the product.

Imagine buying a telephone service and receiving a perfectly good engine plus a requirement to recruit its mechanic. For some customers, that arrangement is sensible. For a developer trying to add calling to a website, it creates an entirely new occupation. Minessale’s commercial problem was how to make the specialist work reusable without making every customer a specialist.

“The solution got bigger than the problem.”Anthony Minessale, describing FreeSWITCH’s growth

SignalWire, founded in 2017, offered communications infrastructure that developers could control through software. Instead of assembling their own network of telephone servers, customers could concentrate on application behavior: what happens when somebody calls, where the call goes, what information an application needs.

The founding rationale was consistent with his earlier work. More people wanted communications capabilities than wanted a career operating communications infrastructure. Cloud delivery could move the maintenance burden into a shared platform. The customer still wrote the application; the team running the platform took responsibility for more of what made it possible.

That is a demanding promise. Every simple interface places complexity somewhere. The work behind SignalWire is deciding how much of it the platform can absorb, and how much control a developer still needs to retain. Minessale had already spent years discovering how often those two requirements collide.

Remote before it was a pitch

The organization behind the code also reflected its purpose. Minessale has described working remotely from 1995, years before starting FreeSWITCH. The project’s distributed team used its own communications software as part of working together. SignalWire was founded with a distributed team too.

He recalled investors who were more accustomed to companies gathered near their offices. Geography, for those investors, supplied a familiar picture of what a startup looked like. For a group already building communications across distance, a shared address was a less persuasive organizing principle.

There is a practical test inside that arrangement. Software meant to let people work together from different places has colleagues available to try it. Their own meetings and collaboration put everyday demands on the idea. A grand claim about connecting the world eventually encounters the ordinary requirement that somebody else on the team must be able to hear you.

The pizza that wasn’t sold out

The arrival of conversational AI gave Minessale a fresh version of his recurring problem. Software could speak with callers naturally. The question was how to let it do useful work while keeping the application’s decisions under control.

A SignalWire pizza-ordering bot furnished an example. It began telling callers that menu items were sold out, although nothing was. A fluent sentence had become an invented business fact. To a caller trying to order dinner, the distinction between a model’s improvisation and the restaurant’s actual inventory is rather consequential.

The SignalWire Agents SDK embodies his proposed response. Developers define workflow steps, decide which tools are available and check requested actions in their own code. The model handles language within that structure. A refund, a reservation or an order should require the application’s real rules to be satisfied. Eloquence alone does not confer purchasing authority.

The call still has to connect

In April 2026, Minessale described voice AI failures through three familiar engineering concerns: who owns the conversation’s state, how delays accumulate between systems, and how rules are enforced. Those concerns survive every improvement in a model’s vocabulary.

A caller interrupts. A call transfers. Two events arrive almost together. An application must keep track of what has happened and what should happen next. Minessale argues for handling AI within the infrastructure that already manages the audio and call state. That is SignalWire’s architectural position, shaped by his experience operating voice systems.

It also explains his impatience with a beautiful demonstration that leaves the underlying machinery unresolved. People on a call encounter a single experience. They do not divide an awkward silence into the separate vendors who contributed to it. Somebody has to take responsibility for how the whole conversation behaves.

His August essay supplied the career’s current refrain: “Nobody should have to solve an escape room to ship a product.” The FreeSWITCH creator still enjoys elaborate machinery. He would prefer the next developer to spend less time trapped inside it.