Gridbase · ATOC · North Texas
Jake Knight
Founder of Gridbase and creator of ATOC. A systems engineer and published author with a background in aerospace manufacturing, building the full stack of modern tactical technology: edge networking, situational awareness, secure communications, and AI-accelerated software, from field-deployable hardware to production SaaS. Also founder and executive director of the Big Bend Protection Alliance, a non-partisan 501(c)(3) conservation charity working to protect Big Bend National Park and the surrounding borderlands of Far West Texas.
In the News
AFP News Agency · Agence France-Presse · August 17, 2026
Also on the wire: quoted in the AFP story carried by FOX 26, France 24, RFI, La Presse (in French), The Straits Times, and Yahoo News; report syndicated by the National Post.
Approach
My work runs the entire stack: RF and embedded hardware, edge and self-hosted infrastructure, native mobile and web applications, and the cloud platforms that connect them. Aerospace manufacturing shaped how I build, with rigor, repeatability, and systems engineered to hold up in the field. I pair that discipline with advanced AI agent workflows: orchestrated, multi-agent pipelines that multiply throughput roughly tenfold, compressing ambitious scope into shipped, production-grade product at the velocity of a much larger team.
The AI build log ≈1,090 hours · 4,198 commits
Every figure below is counted from the git history of 81 repositories, not estimated from memory. Consider it the receipt for the paragraph above.
Build it, tear it out, build it again
The flattering number is 6.3 million lines. I don't use it. Most of that is machine output, not work: 3.4 million lines of decompiled bytecode, another half million from a third-party project imported whole, and the rest vendored dependencies and build artifacts. The honest count is 1.49 million lines written and 1.10 million still standing.
The ~392,000-line gap is the part worth pointing at. Anything can be built once.
- A quarter of all code written was later deleted — 187,665 lines cut back out of 701,417.
- The heaviest build, an edge field-gateway platform, threw away half of everything it wrote: 76,934 lines deleted out of 151,749.
- Two shipping iOS apps were prototyped in Kotlin Multiplatform, then deleted to a tag and rebuilt native in Swift when the cross-platform layer stopped earning its keep.
- A mesh-networking plugin was built, signed, and proven to load on retail hardware — then paused on purpose, because it could not name a capability the existing stack didn't already cover. Working code and justified code are different things.
That is what the tooling actually buys: enough throughput to build the wrong thing, find out fast, and afford to rebuild it. The speed is only worth something because the deletions are cheap.
Method — counted 5 August 2026 from full clones of all 81 repositories: default branch, merge commits excluded; vendored dependencies, build output, and lockfiles excluded throughout. Hours are inferred from commit timestamps, where a gap over 90 minutes starts a new session and each session is credited 30 minutes of lead-in, so the figure is a floor rather than a stopwatch — thinking that ends without a commit isn't counted. "AI-paired" means the commit records an AI co-author.
Expertise
Software & product engineering
Native Swift and SwiftUI iOS apps shipping on the App Store with StoreKit 2 subscriptions; Next.js, React, and TypeScript on the web; and multi-tenant SaaS on Supabase and Postgres with Stripe billing. A repeatable geospatial "radar-UI" pattern turns one proven formula into a family of products, backed by custom end-to-end encryption and forensic, root-cause debugging.
AI-accelerated engineering & automation
Advanced AI agent workflows are central to how I operate: orchestrated, multi-agent pipelines that multiply output several-fold and turn scope that would occupy a team into shipped product. It extends into production AI, with local and offline LLMs (Ollama, Open WebUI), hybrid local-plus-cloud inference, per-repository AI assistants with guardrails, and an AI voice agent that answers the business phone line and captures leads.
Infrastructure & self-hosting
Network topology and edge architecture built LAN-first and private-by-default, meshed with Tailscale and WireGuard. Dockerized services and systemd automation behind Caddy and Authelia; Jetson and edge-AI provisioning; offline-first, grid-down system design; and resilient storage, backup, and NVR video pipelines.
Radio, comms & off-grid systems
Ham radio and software-defined radio; TAK server engineering and federation (OpenTAKServer) for team situational awareness; DMR and digital voice; and Reticulum, NomadNet, and LoRa mesh field nodes, including RF-over-voice gateways that put an LLM on the other end of a radio call.
Security, OSINT & hardening
Ethical public-records OSINT pipelines across SEC EDGAR, USASpending, SAM.gov, and mapping data; firewall and egress control; IoT air-gapping; container and network hardening; and deception techniques such as honeypot tripwires that turn any intrusion into an instant alarm.
Hardware & fabrication
Field-deployable hardware product lines including the Pocket, the Jetson Field Gateway / RackPi, and base stations, plus 3D-printing and print-farm workflows, parametric CAD (build123d), and functional enclosure and airframe builds.
Brand, media & growth
End-to-end brand systems, marketing, and growth-funnel instrumentation; long-form builds and walkthroughs published on YouTube; and audio and music-production tooling.
Speaking & writing
As an educator, I speak publicly on radio communications and edge networking, translating complex field systems into practical, hands-on knowledge for operators and builders. I am also the author of "The Baofeng Buyers Guide," a practical handbook for choosing and getting the most out of Baofeng handheld radios.