Notes From Building a Calling Product Solo

Why a solo developer takes on telecom

Phone infrastructure has a reputation, and it earns it. Carrier rules, number reputation, call attestation, the difference between a call that connects and one that silently dies on the network. None of it is the kind of thing you pick up over a weekend. So the reasonable question is why one person would wade into it instead of building something safer.

The honest answer is that I had built consumer software before, with CanScan AI, and I came away convinced that the products worth making are the ones where the hard part is the point. If placing a real, trusted business call were easy, it would already be a feature inside ten other apps. It is not, because the call itself is where everything gets difficult, and that difficulty is exactly the moat.

Being solo changes how you approach that. I cannot brute-force a hard problem with headcount, so I have to actually understand it. Every layer I lean on, from the carrier routing to the voice model, I have to be able to reason about end to end, because when something breaks on a live call there is no one else to escalate to. That constraint is uncomfortable, and it has also made the product better than it would have been with a team to hide behind.

What is genuinely hard about phone calls

The thing nobody tells you is that a phone call is a real-time conversation with no second chances. On the web you can show a spinner. On a call, a pause of even a second reads as the line going dead, and the person on the other end starts saying hello, hello. So the whole system is a fight against latency: hear the words, understand them, decide what to say, and start speaking, all before a human notices the gap.

Then there is the matter of knowing when to stop. People interrupt. They trail off. They say one more thing right as you start to respond. An AI that talks over a receptionist, or keeps going after the answer has already been given, feels broken in a way that is hard to forgive. Getting the turn-taking right, the sense of when it is your turn and when it is not, has been some of the most finicky work in the product.

And underneath the conversation sits the network. A call has to come from a real number with real carrier attestation, or it gets flagged, screened, or dropped before anyone hears it. VeraDial routes through Twilio's carrier-grade infrastructure with STIR/SHAKEN A-level attestation precisely because that trust layer is not optional. A perfect conversation that never connects is worth nothing.

• Latency: the round trip from hearing to speaking has to stay short enough that the line never feels dead.
• Interruptions and turn-taking: knowing when to keep quiet is as important as knowing what to say.
• Carrier trust: the call has to look legitimate to the network before it ever reaches a human.

The things I obsess over

Call quality comes first, and not in a slogan way. If the voice stutters, or the AI mishears a confirmation number, or it sounds like a robot reading a script, the whole premise falls apart. The AI always identifies itself as an AI at the start of every call, and then it has to be good enough that the person stays on the line anyway. That bar is high and I do not let it slip.

The second thing is the transcript. After the call, you get a full transcript and a summary, and that record has to be true. If it says the appointment is Thursday at two, it had better be Thursday at two, because you are going to act on it without having heard the call yourself. An inaccurate summary is worse than no summary, so accuracy there is not a nice-to-have, it is the product keeping its word.

The third is your number and your reputation, and this one I treat as close to sacred. VeraDial is verified business calling. That means I will never let the product behave in a way that gets your number flagged, screened, or burned. No spammy patterns, no tricks, nothing that trades your long-term standing for a short-term result. The whole point is that when you call, people can trust it is really you.

Why I build products I have personally needed

VeraDial exists because I kept hitting the same wall. There were calls I needed made, the kind every small operator knows, checking a detail, confirming a booking, chasing down an answer, and they always landed at the worst time or fell off the list entirely. I wanted something that could take a goal and a caller profile, place the call, and hand me back exactly what was said.

Building for yourself is a useful forcing function. I am the first person who has to live with every rough edge, so the feedback loop is immediate and unsentimental. If a flow annoys me on a Tuesday, it gets fixed, because I am the one who has to use it again on Wednesday. There is no committee deciding whether the friction is worth addressing.

It also keeps the scope honest. The live product today is outbound calling plus the verified-identity foundation around it: real US and Canada numbers, business SMS, call recording, voicemail transcription. I would rather ship the parts I actually use and trust than gesture at a roadmap I cannot stand behind. The trust here is built by the work, not by the pitch.

What being small actually buys you

There is a version of this story that is all hardship, and that would be dishonest too. Being one person has real advantages. I can hold the entire system in my head, which means a bug report and a fix are rarely more than a short hop apart. When something on a call feels off, I am not filing a ticket into a queue, I am opening the code.

It also means I can be candid with you, which is most of why this post exists. I am not going to overstate what VeraDial does or invent numbers to look bigger than I am. The inbound AI receptionist people ask about is coming, but it is not live yet, and I would rather tell you that than let you assume otherwise. The product is supported by ElevenLabs Grants and Cartesia, runs on SOC 2 infrastructure, and costs $9.99 a month with a hundred credits included. Those are the real facts, and they are enough.

The trade is that I move at the pace one person can move at, and some things take longer than I would like. I have made my peace with that, because the alternative, shipping things I do not understand or cannot stand behind, is the one outcome I am not willing to accept on a product whose entire promise is that it can be trusted.