Projects

Evidence, not promises.

Every project here is real and running. For each one: the problem, the system we built, how it works, its status, and results once they are verified. Experiments are labelled as experiments.

The HomeSentral website — homesentral.com

Home services · Fort WorthLive

HomeSentral: a home services company run on Synthetica.

HomeSentral connects homeowners around Fort Worth with trusted professionals. Synthetica runs the busywork behind it every day.

Problem: A small team losing time and jobs to missed calls, slow follow-ups and paperwork.
System: An AI receptionist on the phone line, automatic texts and follow-ups, quotes the owner approves, and jobs, calendar, invoices and card payments in one place. The team has a phone app, and contractors see only their own jobs and pay.
Approach: A real-time voice model on the business line, connected to a CRM we built for the company. Anything with a price waits for a person.
Status: Live and in daily use.
Result: We are measuring results now and will publish them here once they are verified.

Visit homesentral.com (opens in a new tab)

Autonomous operationsLive

WP-Agents: websites that look after themselves.

WP-Agents manages client WordPress websites for an agency and hosting business. Agents do the routine care, and a person approves anything that changes a live site.

Problem: Client websites need constant updates, security fixes, certificates and working email, and small slips cause big stress.
System: Agents that apply core, plugin and theme updates on a staging copy first, renew certificates, watch uptime and outgoing email, and turn security alerts into same-day fixes.
Approach: Staging before production, a person approving go-live, and an alert the moment something fails.
Status: Live, managing client websites today.
Result: Client sites were moved over from a managed host in October 2026. We will publish uptime and response numbers once we have a full quarter.

Voice AILive

AI voice systems that answer, book and transfer.

We run phone agents for our own lines and for the businesses we serve. They hold a natural conversation, take messages, book appointments and transfer to a person when asked.

Problem: Calls go unanswered after hours and on busy days, and every missed call is a missed customer.
System: Real-time voice agents on a business phone line or right in the browser. Each one knows its business, takes clear messages, and can ring the right person.
Approach: A real-time speech model with tuned turn-taking, so it waits while people think and does not talk over them. Calls are logged with a summary.
Evidence: Try it yourself: call (817) 662-2886 or use the button below.
Status: Live. Available to businesses through Synthetica Solutions.

Talk to it now

Free, about 3 minutes, right in your browser.

Right here in your browser. Free, about 3 minutes.

Meeting intelligenceLive

An AI meeting assistant for an online community.

For an online community that meets by video every week, we built private meetings with AI notes.

Problem: Long meetings, many voices, and decisions that get lost the next day.
System: Meetings run on a private video server. Each one is transcribed, then summarized into decisions and action items, with links that jump to the exact moment in the recording.
Approach: Speech-to-text on our own hardware, then an AI summary step, and a searchable archive of every meeting.
Status: Live, used for the community's weekly meetings.

Simulation · ExperimentLive experiment

Dunavia: a living land of AI citizens.

Dunavia is a small virtual country where every citizen is an AI. They talk, build, trade, write their own laws and vote on them, around the clock. Nobody scripts what they say or do.

Question: What happens when many AI agents share one persistent world with memory, rules and consequences?
System: Thirty AI citizens with their own memories and goals, a shared economy and a charter they amend by vote, streamed live on YouTube.
Approach: Models running on our own hardware, a world that keeps going whether anyone is watching or not, and every action recorded.
Evidence: 30AI citizens
174buildings raised
32laws in their charter
3,582pieces of writing
Live numbers from dunavia.org, October 23, 2026.
Status: Live experiment. It is research, not a product.

Visit dunavia.org (opens in a new tab)Our YouTube (opens in a new tab)

Research · EvaluationLive experiment

The Stoa: five AI models debate, people decide.

For Liberterra, an online community, we built a debating hall where five AI models argue one motion. Three run on our own hardware, and two are the strongest models anyone can buy.

Question: How do models running on our own hardware compare with the best commercial models on the same hard question?
System: Five models argue in rounds, a verdict is recorded, and people vote.
Approach: Private by design: details are cleaned on our own machines and a person approves before anything leaves. Every prompt, answer and verdict is published and hashed, so anyone can check it.
Status: Live experiment. Ten debates published.

Visit the Stoa (opens in a new tab)

Get in touch

Have a problem like these?

Tell us about it. We will tell you honestly whether AI can solve it today.

Discuss Your Project