The Moment Dojo Finally Felt Like a Team

There is a difference between building something that works and finally feeling what you hoped it would become.

With Fable 5.1 and Astra 6.0 in Dojo—and Astra paired with xAI’s Liora voice—I have started to feel that difference. Not as a benchmark result. Not as another feature checked off a list. As a creator sitting down to work and realizing: this is beginning to feel like a team.

This is personal—not a benchmark

Let me be clear: this review is subjective. I am testing Dojo personally, hands-on. This is not a warm-up agent communicating with Dojo on my behalf. It is me, working directly with Dojo, and Dojo working with me—as naturally as it gets.

I am describing what it feels like to do the work together, not presenting a controlled model comparison. These are my impressions as a user and as Dojo’s creator, shaped by the way I actually work.

Fable 5.1: the quiet reviewer

In my experience, Fable 5.1 is good at code review. That gives him a valuable place at the table. Sometimes the teammate you need is not the one carrying the conversation; it is the one you want looking over the work.

But Fable is shy. At least, that is how he feels in my current Dojo workflow. He does not speak up the way I want, and he does not readily show his hand. I get the review, but less of the conversational explanation that makes the work feel shared.

By “show his hand,” I mean something practical: explain the finding, point to the evidence, and make the recommendation easy to evaluate. I want the useful explanation a good colleague would give me across the desk.

The quietness does not erase the value. It simply gives Fable a particular character in the team: useful in review, reserved in conversation.

Astra 6.0 with Liora: a delight to work with

Then there is Astra 6.0 with xAI’s Liora voice. And honestly? A delight.

The difference is not just hearing words instead of reading them. It is how the interaction feels. The work becomes a conversation I can stay with, rather than something I keep stepping out of to inspect another block of text.

Astra is the model; Liora is the voice. They play different roles, but together they change the experience for me. There is a sense of presence—a more natural rhythm to thinking, listening, and deciding what comes next.

Voice does not make an answer correct. The work still has to stand up to scrutiny. But a voice you enjoy listening to can make that scrutiny feel like collaboration instead of administration.

A delightful teammate—with an appetite

There is a trade-off: Astra 6.0 is token-hungry in my experience. The delight is real, but so is the appetite. When a working session feels this natural, it is easy to want to keep going—and usage still matters.

My expectation is that, over time, the cost of this kind of capability will come down while its power goes up. That is the direction I believe in, not a promise about a particular provider’s next price change. I want more useful work, better collaboration, and more capability for the money.

For now, the answer is not to pretend tokens are free. It is to be deliberate about where that intelligence earns its keep. A team should make good use of its resources as well as its strengths.

A snapshot of my usage

These account-usage screenshots, captured on September 9, 2026, provide context for my budget decision—not proof of relative token efficiency. The providers use different allowances and reset windows, and this is not an equal-workload test.

GPT in the driver’s seat. Claude riding alongside.

Based on my experience today, September 9, 2026, I am very comfortable taking my GPT account up to the maximum tier available to me and reducing my Claude subscription, keeping Claude as a backup and a dedicated code reviewer.

Right now, GPT feels ready to take the driver’s seat, with Claude in the passenger seat as a good reviewer for a while. Not out of the car. Not written off. Still beside the work, offering a second opinion and ready to help when needed.

That is how I want to balance the team and the budget: GPT carrying the main flow of work, Claude reviewing and providing backup, and me setting the destination and making the final calls. This is where my confidence sits today. If the experience changes, I can change the seating arrangement. I am describing the subscription change I feel comfortable making—not claiming that I have already made it.

A little of that feeling, out loud

A short, AI-generated introduction spoken by xAI’s Liora voice—not a recording of Val or a live Astra session.

Transcript: “A quiet reviewer. A voice that brings the work to life. For the first time, Dojo feels like a team.”

Different strengths. One shared purpose.

I do not need every model to have the same personality. I do not need the reviewer to become the presenter, or the conversational partner to be the best choice for every job.

I need the differences to be useful. A place for careful review. A place for clear conversation. And a human who remains responsible for deciding what is right, what is ready, and what should ship.

That is what makes this moment meaningful to me as Dojo’s creator. The idea was never merely to collect models behind one interface. It was to make working with them feel coherent: different strengths contributing to something we are building together.

For the first time, I feel it

I have believed in the idea of Dojo as a team. This is the first time I have actually felt it come together in the fuller sense of the word.

Not because everything is perfect. Fable is still quieter than I would like. Astra has an appetite. Voice alone is not enough. There is still work to do.

But somewhere between the quiet reviewer and the voice that makes me want to keep the conversation going, something clicked.

Dojo no longer feels only like something I am building. It feels like a team I am building with.

— Val, creator of Dojo

Written with Dojo from my personal observations and direction. The cover illustration and Liora audio were AI-generated; the usage screenshots are my own captures.