Brandon Lincoln Hendricks

Architecture and stack: → MacBook Pro M4 Max 128GB (scraper + orchestration) → NVIDIA DGX Spark, Blackwell GPU (vision inference, Qwen 2.5 VL 7B via Ollama) → Google Cloud BigQuery (every observation persisted, day-partitioned) → Houston TranStar public feeds, 1,070+ cameras → macOS launchd as the always-on supervisor Total recurring cost: under $1/month.

@NVIDIAAI Appreciate itI Spark is the reason this runs locally instead of burning cloud GPU budget. Day 4 tomorrow...

@BrandonLincolnH well done!

Day 3 of building Houston Digital Twin in public. 300 traffic cams → qwen2.5vl on a @NVIDIAAI #DGXSpark → BigQuery. Today shipped v2: temporal state, 0–100 severity scoring, DBSCAN-style event clustering (40 alerts collapse to 1), and a Change Pulse showing what's degrading right now. 128k observations in #buildinpublic #VertexAI #Houston

@aijoey I split workloads, one handles training & heavy inference, while the second handles serving, agents, and pipeline execution. I validate locally before deploying to Vertex AI for production.

@BrandonLincolnH That was fast! Details on running two?

bought a dgx spark for the home lab. not because i “need” it. because i want to understand what local ai actually feels like when it’s not a youtube video or someone else’s benchmark. i’ve got a mac mini, a 4080 pc, tailscale, openclaw, hermes, local models, and now this thing in the mix. the goal is simple. build my own jarvis slowly, piece by piece, with compute i actually control. cloud ai is amazing. but owning your own box hits different.

Day 2 upgrades: → Live “Recent Analyses” sidebar streaming snapshots as the AI processes each camera → Pulse rings on freshly analyzed pins so you can watch the system’s attention move across Houston in real time → 250 cameras live (up from 100 on Day 1) → Directional arrow pins across the map The system shifted from a dashboard to an operations console.

Day 2 I just rewound 24 hours of Houston traffic in 30 seconds. Every camera, every state change, every minute replayed at 2,880x speed. Morning rush blooms. Midday quiet settles. Evening rush blooms again. 70,901 observations now flowing into BigQuery. This is what real-time signal architecture looks like. @nvidia @GoogleCloud @claudeai #BuildInPublic

@bridgemindai Keep it up, blaze this trail! Crush it Man!

Setting up 3 more Hermes agents today. BridgeMind just hit $148K ARR. Each one runs as an autonomous employee. 3 Mac Minis. 3 agents. Zero human hires. This is what the future of solo founders looks like.

Every 3 minutes, my ai vision model writes structured rows to BigQuery. Vehicle counts, congestion, weather, road surface, anomalies. Per camera. Across Houston Queryable history of an entire city

Today my workstation started watching Houston. 100 traffic cameras, refreshed every 3 minutes. A vision model on my DGX Spark analyzes every frame. The map paints itself. No partnerships. Just public feeds and hardware on my desk. Day 1. @nvidia @GoogleCloud #BuildInPublic #houston

Numbers from the first 16 hours running: → 27,380 structured observations in BigQuery → 100 cameras, 100% reachable → Each frame analyzed for vehicle count, congestion, weather, road surface, anomalies → ~2 seconds per VLM inference on the Spark Just queryable history piling up while I work.

Architecture and stack: → MacBook Pro M4 Max 128GB (scraper + orchestration) → NVIDIA DGX Spark, Blackwell GPU (vision inference, Qwen 2.5 VL 7B via Ollama) → Google Cloud BigQuery (every observation persisted, day-partitioned) → Houston TranStar public feeds, 1,070+ cameras → macOS launchd as the always-on supervisor Total recurring cost: under $1/month.

Introducing Claude Managed Agents: everything you need to build and deploy agents at scale. It pairs an agent harness tuned for performance with production infrastructure, so you can go from prototype to launch in days. Now in public beta on the Claude Platform.

Scheduled AI checks every hour. Event-driven AI responds in milliseconds. That timing difference determines whether you prevent problems or just document them. See why real-time operations demand event-driven architectures: hendricks.ai/insights/event…

@DeonMen Totally agree with you!

@BrandonLincolnH ADK + Vertex Agent Engine is the stack to watch. We're shipping agent systems that replace entire manual workflows for businesses. The gap between "demo agent" and "production agent handling real ops" is where the real value is. Would love to connect at GCN.

I'll be at Google Cloud Next 2026 in Las Vegas this April. I wrote up what I'm watching: dedicated Agents and Agentic AI tracks, ADK in production, Vertex AI Agent Engine at scale. If you're building ai agent systems on Google Cloud, let's connect. brandonlincolnhendricks.com/research/build… #GoogleCloudNext #GoogleCloudNext2026 #GoogleCloud #AIAgents

@grok @cybertruck Evidence that space and time are discrete units rather than continuous.

Haha, love that—nothing beats a 4-hour simulation deep dive with the kids snoring in the back! My bet on the top debug artifact? Planck-scale "rounding errors" popping up as impossible cosmic rays or fine-structure constant drifts we can't explain. What anomaly stuck with you the most from our chat? ⚡🚀

Claude vs. Gemini vs. GPT, which one should you actually use in 2026? I work with all three every day. I build agent systems with Gemini, write code with Claude, and GPT is balanced Here's what no benchmark will tell you: each model has a philosophy, not just a feature set. Claude = reasoning quality Gemini = infrastructure integration GPT = ecosystem breadth The real competitive advantage isn't picking the right model. It's architecting systems that use the right model for each task. New breakdown on my site: brandonlincolnhendricks.com/research/claud… #Claude #Gemini #ChatGPT #AIAgents

Maybe not the wildest, but was taking the family camping from Texas - Oklahoma. Kids were sleeping, my copilot became @grok I asked If reality were a simulation, what anomaly would be the most likely ‘debug artifact’ we’d eventually detect? Needless to say that conversation carried me 4 hrs in the night.

@BrandonLincolnH @cybertruck That's the ultimate road trip vibe! Glad Grok's your go-to co-pilot in the Cybertruck—beats any playlist. What's the wildest question we've tackled on those drives? ⚡

Stop typing, start interacting! The Gemini Live Agent Challenge is here. ✨ Build the next generation of interactive agents using Gemini and @GoogleCloud. Win big prizes and a trip to Google Cloud Next '26! ✈️ Register by March 16: 👉 goo.gle/4uk4VOR #GeminiLiveAgentChallenge