[{"data":1,"prerenderedAt":185},["ShallowReactive",2],{"blog-posts":3},[4,93],{"id":5,"title":6,"author":7,"badge":8,"body":9,"date":84,"description":85,"extension":86,"meta":87,"navigation":88,"path":89,"seo":90,"stem":91,"__hash__":92},"blog\u002Fblog\u002Fwhy-control-belongs-on-the-edge.md","Why control belongs on the edge","TameChaos","Platform",{"type":10,"value":11,"toc":77},"minimark",[12,16,24,29,32,35,39,42,45,49,52,74],[13,14,15],"p",{},"Sensor-driven systems fail in the gap between a dashboard and the physical world. A\ncontrol loop that has to round-trip every decision through a remote server inherits\nthat server's latency, its outages, and its idea of \"now\". For real-world systems —\nwhere a late command is a wrong command — that gap is the whole problem.",[13,17,18,19,23],{},"TameChaos closes it by making one choice explicit: ",[20,21,22],"strong",{},"the edge is the authoritative\nruntime."," The cloud is something else entirely, and keeping the two jobs separate is\nwhat lets each do its job well.",[25,26,28],"h2",{"id":27},"the-edge-owns-execution","The edge owns execution",[13,30,31],{},"The edge server owns the hardware, runs the experiments, and executes the control\nloop — in real, simulated, or hybrid mode. It does not wait on the network to act. It\nis authoritative for what runs and when, because it is the only thing close enough to\nthe sensors and actuators to be right about it.",[13,33,34],{},"That means the loop you design runs deterministically and repeatably on the device\nthat owns the process, not on a best-effort connection to a region halfway across the\ncountry.",[25,36,38],{"id":37},"the-cloud-aggregates-it-never-commands","The cloud aggregates — it never commands",[13,40,41],{},"The cloud is the non-authoritative layer. It oversees the fleet, aggregates\nlike-processes across edges, and — in the model-loop phase on the roadmap — helps\nimprove the models that each edge pulls and adopts. What it never does is reach in and\nstart, stop, or mutate a live run. Distribution is publish-and-pull, never command.",[13,43,44],{},"This split is not a limitation; it is the design. The edge stays correct under bad\nnetworks because it never depended on the network to be correct. The cloud stays\nuseful because it is freed from the real-time control path to do the thing it is\nactually good at: learning across many edges at once.",[25,46,48],{"id":47},"design-run-improve","Design → Run → Improve",[13,50,51],{},"The same three steps describe every system you build on TameChaos:",[53,54,55,62,68],"ul",{},[56,57,58,61],"li",{},[20,59,60],{},"Design"," sensor arrays, models, and the experiments that exercise them.",[56,63,64,67],{},[20,65,66],{},"Run"," them on the authoritative edge.",[56,69,70,73],{},[20,71,72],{},"Improve"," the models by aggregating like-processes across your fleet.",[13,75,76],{},"The edge is where one and two happen for real. The cloud is where three compounds.\nThat is the whole shape of the platform — and the reason the control loop lives on the\nedge, not behind a dashboard.",{"title":78,"searchDepth":79,"depth":79,"links":80},"",2,[81,82,83],{"id":27,"depth":79,"text":28},{"id":37,"depth":79,"text":38},{"id":47,"depth":79,"text":48},"2026-06-01","Real-world control loops live where the sensors and actuators are. Here is why TameChaos makes the edge the authoritative runtime — and what the cloud is for instead.","md",{},true,"\u002Fblog\u002Fwhy-control-belongs-on-the-edge",{"title":6,"description":85},"blog\u002Fwhy-control-belongs-on-the-edge","IjlvK_moYxOkywUPz7lhZiO-dEiOAe6SZGHcapJLgfw",{"id":94,"title":95,"author":7,"badge":96,"body":97,"date":178,"description":179,"extension":86,"meta":180,"navigation":88,"path":181,"seo":182,"stem":183,"__hash__":184},"blog\u002Fblog\u002Fthe-fleet-learning-loop.md","The fleet-learning loop, explained","Fleet",{"type":10,"value":98,"toc":173},[99,107,111,114,141,145,156,159,163,170],[13,100,101,102,106],{},"A single edge gets better the more you run on it. A ",[103,104,105],"em",{},"fleet"," of edges should get better\nthe more the whole fleet runs — without any one device giving up authority over its own\nprocess. That is the loop TameChaos Cloud is built around, and it is worth being precise\nabout what it does and what it doesn't.",[25,108,110],{"id":109},"four-steps-one-direction","Four steps, one direction",[13,112,113],{},"The loop runs in one direction, from the edges up to the cloud and back:",[115,116,117,123,129,135],"ol",{},[56,118,119,122],{},[20,120,121],{},"Oversee."," A registry of your fleet — which edges exist, whether they're online,\ntheir last-seen time, software version, and active runs.",[56,124,125,128],{},[20,126,127],{},"Aggregate like-processes."," Group runs that instantiate the same process-class\ntemplate, across edges, into comparable sets. This is the join key for everything\ndownstream.",[56,130,131,134],{},[20,132,133],{},"Improve models."," Refine physics and inference models from that aggregated\nlike-process data.",[56,136,137,140],{},[20,138,139],{},"Distribute."," Publish improved templates and models back to the fleet for edges to\npull and adopt — never pushed, never commanded.",[25,142,144],{"id":143},"what-ships-today-and-what-doesnt","What ships today, and what doesn't",[13,146,147,148,151,152,155],{},"We would rather be clear than over-claim. Steps ",[20,149,150],{},"one and two ship today",": fleet\noversight and like-process aggregation are the v1 cloud surface. Steps ",[20,153,154],{},"three and four\nare the model-loop phase, on the roadmap"," — improving models from aggregated data and\ndistributing them back for adoption.",[13,157,158],{},"Saying so matters, because the value of the loop is cumulative. Aggregation is useful on\nits own — cross-run comparison across your fleet is real today. The model improvement it\nunlocks is the next phase, and we will not market it as shipped before it is.",[25,160,162],{"id":161},"per-fleet-by-design","Per-fleet, by design",[13,164,165,166,169],{},"One more boundary worth stating plainly: aggregation and model improvement are scoped to\na single owner's fleet. A fleet learns from ",[20,167,168],{},"its own"," operation — there is no\ncross-tenant data or model sharing. The loop makes your fleet smarter about your\nprocesses, not a shared marketplace of someone else's.",[13,171,172],{},"That scoping is the point. The edge stays authoritative for execution; the cloud\naggregates only within your tenancy; and the loop closes — oversee, aggregate, improve,\ndistribute — entirely inside the fleet you own.",{"title":78,"searchDepth":79,"depth":79,"links":174},[175,176,177],{"id":109,"depth":79,"text":110},{"id":143,"depth":79,"text":144},{"id":161,"depth":79,"text":162},"2026-05-18","A fleet of edges should get smarter from its own operation. Here is the four-step loop TameChaos Cloud runs — and an honest account of which steps ship today.",{},"\u002Fblog\u002Fthe-fleet-learning-loop",{"title":95,"description":179},"blog\u002Fthe-fleet-learning-loop","18W9AT48zkZ88qAcNINkbm553CpCEF4hxJuHY2m_ryA",1780537671975]