KyrosWorks
AI-native software for the kind of problems people reliably forget to solve, fail to sustain, or can't get right by hand.
Products
Software that does the thing, end to end
Each product solves a problem its users had been hand-piecing together for years. The math has been figured out. The interfaces are calm. The work that should have been infrastructure is now infrastructure.
Live · iOS · App Store
Open Road Ace
Pacing app for open-road time-trial racing. Replaces a human navigator with a virtual rabbit and millisecond-accurate finish-gate detection.
openroadace.app →Live · iOS · App Store
ChoreKey
Parental-control app for the kid's phone. Chores unlock the apps — only at home. Leave the home geofence, the lock lifts automatically. No reading messages, no scraping social, no subscription.
chorekey.app →TestFlight · iOS
Climb Test Prep
Adaptive iOS app to prep for the SAT® exam, built around mastery, not engagement. AI tutor (Wren) with persistent memory, sherpa-guided climb to your target score. No XP, no streaks, no leaderboards.
climbtestprep.app →Coming soon · Android · Play Store
Kyros BT Connect
Free check: does your Android aftermarket head unit do Bluetooth A2DP audio? Finds out whether your exact unit can stream real Bluetooth audio to your car's built-in Bluetooth radio, no aux cable, no hardware adapter, no wiring.
kyros.works/bt-connect →Private beta · Web
PROP
An AI-native quoting engine that matches free-text customer requests to the right products in a supplier's catalog. A deterministic matching core resolves most lines instantly with no model call; a confidence-gated LLM handles only the ambiguous tail, and anything below the confidence bar routes to a person. Accuracy where the machine is sure, a human where it isn't.
Internal · Infrastructure
Wren Engine
Coherent AI. Customizable. Self-improving. Time-aware. Durable memory. A custom AI harness — the runtime under every KyrosWorks product — that solves what most LLM stacks still can't: amnesia between sessions, drift across time, single-machine isolation, and single-agent ceilings.
People
Built for the people who already have the problem
Each product is shaped by a specific group of users — people who have been working around the gap for years, stitching the workflow together by hand, waiting for the right tool. KyrosWorks software is what they would have built for themselves.
Open Road Ace
Open-road racers
TSD competitors at sanctioned events like the Big Bend Open Road Race, the Silver State Classic Challenge, and the Sandhills Open Road Challenge. For forty years they've paced 100-mile courses with stopwatches, paper charts, and a navigator calling splits. Open Road Ace does the timing math at GPS precision so the driver can focus on the road.
PROP
Quoting teams
Suppliers and distributors whose customers order in free text: emailed part lists, spreadsheets, RFQs in a dozen formats. Someone on the team translates every line into a catalog product by hand, and it eats the day. PROP does the translation and surfaces only the lines it isn't sure about.
Climb Test Prep
High school students prepping for the SAT
Students with a target score, a test date, and the wrong tools — generic curricula that ignore where they actually need work, gamified apps that optimize for streaks over mastery. Climb adapts to the student, teaches what's missing, and remembers what's already been worked through.
AI
AI-native development, end to end
KyrosWorks builds with AI as a primary collaborator — from architecture decisions and code review through testing, marketing, and customer support. The AI is the second engineer, not a feature you turn on.
The pattern
AI in our workflow does specific work: writing code from a clear specification, parsing data we'd otherwise process by hand, drafting marketing copy from verified facts. It reviews pull requests against project conventions, maintains a memory layer that survives between sessions, and orchestrates multi-agent workflows when problems span more capability than one model can fit. None of this is a chatbot.
Inside the products
Where AI shows up in shipped products, it shows up quietly and where it changes the outcome — not as a marketing feature. Open Road Ace doesn't have an LLM in it; it has a deterministic timing engine that AI helped build, validate, and calibrate against ten years of historical race data. PROP uses LLMs where the workflow is unstructured text and rule-based code where it's deterministic. AI shows up where it actually outperforms the alternative. Nowhere else.
What this means in practice
Smaller team, faster cycles, deeper testing than a single human could produce alone. The products ship more thoroughly tested than most companies ten times the size. The marketing copy is verified against primary data. The code review is two-pass and adversarial — a second model tries to break the first's work before it ships, and safety-critical math is checked against known fixtures, never on a single pass. The memory persists across sessions. None of this is achievable without AI in the loop — and none of it is sold as "AI-powered," either.
Talk to KyrosWorks
For partnerships, product questions, or interesting problems: [email protected]