Vol. I · No. I An Independent Research Studio Phoenix, Arizona

Qira

Intelligent systems for complex networks

Founded MMXXV imagineqira.com U.S. Patent Pending
From the Founders

Production AI, built from scratch — no university, no lab, no venture capital.

Two brothers in Arizona shipping language models, traffic intelligence, and consciousness research that runs in the real world.

Qira LLC was founded in January 2025 to build production AI platforms from first principles. Where most teams pick a single domain and stay inside it, we work across three: live traffic intelligence on Phoenix freeways, a custom hybrid Transformer-SSM language model in active training, and an empirical consciousness study now into its third-hundred participants.

We do not raise venture capital and we do not file pre-announcements. Every product on these pages has running infrastructure, a public artifact, or a peer-visible repository. The work is the proof.

The research is protected under U.S. Provisional Patent #64/002,166, filed March 10, 2026.

Departments

What We Ship

Production platforms, not slide decks.

● Live System

Phoenix Traffic Intelligence

Real-time freeway monitoring

Eight Phoenix freeway corridors monitored 24/7. AZ-511 + TomTom live speed feeds, cascade prediction, FHWA delay-cost analysis, and AI-powered crew-dispatch recommendations every two minutes.

PTI8 corridors2-min cadence
Public dashboard temporarily offline.
● In Training

LOLM

Latent Order Language Model

Hybrid Transformer-SSM architecture separating language into surface and latent streams. The 29% minority SSM path is 8,645× more essential than the 71% majority Transformer path — dependency inversion. 7 complementary training losses, 5-stream fan-out. Tested at 20.5M, 149M, 304M, and 1.57B parameters. 43% faster convergence at 300M, 15% lower perplexity at 1.57B.

Transformer-SSM5-stream7 lossesPatent #64/002,166
View on GitHub →
● Active Research

Expression-Gated Cognition

An empirical consciousness framework

A mathematical framework measuring how emotional knowledge gates conscious expression. Gating function g(K) = 4K(1−K). N=344+ participants across 20+ countries. Three confirmed response types — Compressors 40%, Suppressors 30%, Expanders 30%. Stable inverse correlation r = −0.988 from N=9 through N=344.

g(K) = 4K(1−K)N=344+r = −0.988
Take the study →
● Early Development

Codey

AI coding assistant

An AI coding assistant with structural code analysis and multi-provider LLM routing. Currently in early development — not yet production-ready.

Multi-LLMStructural analysis
Public preview forthcoming.
● Live Service

Qira Webbuild

Custom websites & data dashboards

Custom websites and data dashboards built to spec. Fixed pricing, full code ownership, modern stack — from landing pages to full-scale data platforms.

Custom buildFixed priceFull ownership
Enquire by email →
● Live · Free & Public

Are We Friends?

A quieter social network

A free, public social network for real-life friendship. Quieter by design — no ads, no metrics, no public follower counts. Anyone can sign up.

FreePublicNo ads
Visit arewefriends →
The Two of Us

Co-founded by Brothers

Self-taught from Arizona.

Bryan Leonard

Co-Founder & Lead Researcher

Phoenix, Arizona

Self-taught systems engineer and applied researcher. I build production AI platforms from scratch — live traffic intelligence, custom language models, and empirical consciousness research. No degree, no VC, no excuses.

Brandyn Leonard

Co-Founder & Research Lead

Maricopa, Arizona

I design the architectures. LOLM's hybrid Transformer-SSM dual-stream system, the traffic network operating-band framework, and the g(K) = 4K(1−K) gating function in EGC — these are mine. I built them because no existing tool solved the problem the right way.

U.S.

U.S. Provisional Patent — Filed

The LOLM architecture and supporting frameworks are under active U.S. Provisional Patent protection.

Application #64/002,166 · Filed March 10, 2026
How to Reach Us

Correspondence