Happy Easter! Its 2026, and in my mind its the year of the agent. Though you are sure to find shops that have been shipping production voice agents for the past two years or more, 2026 represents a tsunami of teams starting to offer such things, enough to make one think the field is quickly becoming a commodity.

Nothing could be further from the truth.

We are still at the beggining to be sure. If you dig in a little, most of the offerings are based on the same platform bits. I won't name them, but it isn't difficult to stitch together a few public APIs and call it a functioning voice agent. But if all that is happening is passing transcriptions to an LLM and then text to speech, you are betting your business continuity on the LLM and its ability to reason. None of them reach 100% on any metric, so what kind of failures are you willing to endure?

Now we will discover who has wrapped these pieces in the required guardrails. We will know who has added static logic to the skillsets and tools the agents are designed to utilize. We will experience good and bad implementations of memory. We will endure skillfully as well as poorly built knowledge graphs and RAG retrieval. Then there is the latency issue.

Isn't our own society the same? No one considers people to a commodity. There are high performers and lazy slackers, and without any actual stats to hand I would guess mostly adequate for their jobs.

Let's face it - in just a few years there will be many millions of agents. They won't all be the same. They won't all be high performers, and competition will be fierce. I'm excited to be here at the very start, where we are all struggling with the same architectural issues. I would like to think we have some unique ideas that will set VoxLayer's models apart. At the end of the day our resellers will only push our product as long as it is on the forefront, and we intend to stay there ;)