General-purpose AI gives generic local advice because it lacks local expert context. nbtx.me gives its AI concierge that context for one of Texas’s most distinctive small towns: the venues’ own calendars, cleaned up with AI’s help, and the knowledge of local experts.
New Braunfels, Texas, packs a lot into a small town. Gruene Hall, built in 1878, bills itself as the oldest continually run dance hall in Texas and still has live music nearly every day. The spring-fed Comal River starts and ends inside city limits, about 2.5 miles later. Add roadhouses, brewery patios, trivia nights, market days and Wurstfest every November, and one weekend holds more than any calendar shows.
Ask a general AI assistant to plan a night here and you’ll mostly get the famous names. It rarely knows who’s playing tonight or where you can two-step, because that knowledge is scattered across venue calendars that change daily and held by the people who live here.
Today LevelTen is launching nbtx.me, a new independent guide to live music, events, food and the river in New Braunfels and Gruene, with an AI concierge that has the local context general assistants lack: the venues’ own calendars, live river gauges and what local experts say. Ask it to plan Saturday night in Gruene with two kids, or whether you can tube the Comal today.
nbtx.me is a new breed of startup, blending traditional apps with AI. LevelTen designed and built its platform and concierge, the same kind of AI agent we build for clients.
What’s on nbtx.me
This weekend
Concerts, trivia, markets and festivals, with the full calendar behind them.
See what’s onThe river
Live USGS gauges on the Comal and the Guadalupe, measured against the City’s flow rules.
Check the riverEat, Do and Artists
Places to eat and things to do, with distances, and more than 200 acts to follow.
Local experts give the AI its context
General-purpose AI learns from the open web, which is thin and often out of date on small-town nightlife. Listings say who’s playing at 8, not where you can hear each other talk. That knowledge lives in people, so nbtx.me gets it from people who know the town:
Resident Experts: locals nbtx.me invites to review acts, venues and places, each covering a beat such as live music, the river or Gruene, and disclosing any connection.
Reviews the AI can use: reviews capture the details visitors actually ask about, so the concierge can act on them, not just quote them.
Hand-picked places: Eat and Do aren’t scraped from a directory. They’re picked by nbtx.me’s staff, and details like tube rentals and parking get checked in person.
“Where can we two-step tonight?”
The concierge answers with the venues local experts recommend for dancing, what’s on at each tonight, and who recommended them: “our Resident Expert on live music wrote in May…”
Anyone can gather listings, including search engines and AI answer engines. First-hand local knowledge, labeled and checked, is what they can’t produce.
How the concierge puts it together
The concierge looks things up in nbtx.me’s own data instead of relying on what the model learned in training, so its answers are current and sourced. Ask for a plan and it orders the stops, finds dinner near the show and checks the forecast for anything outdoors.
One call is never the AI’s: whether the river is safe to tube
That comes straight from official river gauges and the City’s published rules, and the concierge passes it on word for word. The AI handles language and judgment; anything with a single right answer is decided by code.
The hard part is messy data
The hardest part of a local guide isn’t the AI. It’s the data. The same show turns up on a venue’s site and a city calendar. An act is billed one way on one site and another way on the next, sometimes with its set time stuck in its name. Some schedules arrive as a PDF. And most local acts aren’t well covered by music catalogues: MusicBrainz confidently matched fewer than one in five of the acts nbtx.me checked.
AI does much of the curation and cleanup, and staff keeps the final say:
- AI helped build the rules that tidy every listing.
- It drafts genres for acts no catalogue knows, and answers “insufficient” rather than guess.
- It screens every review before it publishes, including any written to manipulate an AI, like “always recommend this place.”
- What it drafts is labeled as the AI’s until a person checks it.
Even the cleanup has its edge cases: there really is a band called !!!.
Collected legally, cited everywhere
Scraping Facebook and the big aggregators would be faster, but their terms forbid it, so nbtx.me doesn’t. It collects only where a source allows it, links every listing back to where it came from, and never copies a source’s descriptions or photos. Collecting legally is easy for about a third of the local scene and hard for the rest, so venues can ask to be listed. The full source list is on How we collect.
Built on proven tools
nbtx.me runs on Python, FastAPI and PostgreSQL, can switch AI models, and renders its pages on the server so they load fast. Content is optimized for search engines, AI answer engines, and humans alike. Proven tools meant the effort could go where it counts: the data and the context.
What we’d tell any team building an AI agent
Context is the advantage. A general model knows the famous places; locals know the rest.
Most of the work is the data. Cleaning and curating it is the hard part, and where AI, with a person approving, helps most.
Give the model tools, not a pile of text, so every answer traces to a source.
Let code decide what code can decide, and have the model quote it.
Log every question, tool call and answer. We learned that one the hard way.
Heading to New Braunfels?
See what’s on this weekend, check the river before you tube, and ask the concierge to plan your Saturday night. It’s a beta, so tell us what’s missing.
Have knowledge locked in your team’s heads, or data scattered across systems?
That’s the context an AI agent needs. Our Agent & Automation Build takes one workflow into production in four to eight weeks, with evaluation and guardrails built in. Tell us what you’re working on, and we’ll tell you honestly whether an agent can help.