Skip to content

Day-to-Day Use

Once Hungry Machines is installed and your appliances are registered, it mostly runs itself. This page covers what happens automatically, what you can adjust, and what your first week looks like.

Every day, without you doing anything:

  1. Throughout the day, the app collects readings from your appliances so the optimizer can learn how your home behaves.
  2. Once a day, it sends your local weather forecast.
  3. Overnight, Hungry Machines builds a fresh 24-hour plan for each appliance — shifting energy use toward the cheapest hours while staying inside the comfort and charge limits you set.
  4. Every 30 minutes, your Home Assistant applies the current step of that plan: setpoint, mode, and fan for HVAC; on or off for EV chargers, batteries, and water heaters.

You can watch all of this on the dashboard — what’s running now, what’s coming next, and how much you’re saving.

Hungry Machines learns your home, not a generic one, so it needs a few days of running before it can optimize confidently.

WhenWhat’s happening
The first few daysThe app collects readings and keeps you comfortable using safe default settings. Your schedule chart shows a flat comfort band and no savings yet — this is normal.
Once it’s learned enoughHungry Machines has a model of how your home heats and cools. That night’s plan becomes a real optimized schedule, and your savings estimate turns positive.
From then onIt keeps refining the model as it sees more weather and more of how your home responds. Savings tend to grow as the picture fills in.

On the first warm day after you add an HVAC, Hungry Machines runs a short, one-time warm-up for that unit: a controlled morning cooling pattern that lets it directly measure how your specific system responds. While it’s running, the panel shows a banner — and a Skip button if you’d rather it didn’t run that day.

The warm-up only happens once per unit (unless you re-run it later, for example after servicing your HVAC). It’s worth letting it finish: the measurements it gathers make every schedule afterward more accurate.

Open an HVAC’s Edit constraints (from its dashboard card) to tune how it behaves:

  • Base temperature — the temperature you’d like to sit at when you’re home and comfortable.
  • Savings level — how much room you give the optimizer to drift from your base temperature to save money:
    • 1 — Tight (about ±2°F): maximum comfort, minimum flexibility.
    • 2 — Moderate (about ±6°F): a balanced default.
    • 3 — Aggressive (about ±12°F): maximum savings, wider swings.
  • Mode — cool, heat, auto, or off.
  • Away / home times — when you typically leave and return, so Hungry Machines can relax comfort while you’re out.
  • Hourly comfort bands (advanced) — instead of one base temperature plus a band, set your own high and low limit for each hour of the day. Useful if your comfort needs change through the day in ways the simple settings can’t capture.

Wider savings levels give the optimizer more room to work, so they generally save more — the trade-off is larger temperature swings. If your schedule chart looks nearly flat, a tight savings level (or a flat-rate electricity plan) is usually why; see Troubleshooting.

For EV chargers, batteries, and water heaters, Edit constraints sets charge targets, minimums, deadlines, or tank temperature limits instead.

Two ways to step in:

  • Pause one appliance — its per-card toggle takes just that device out of optimization and hands control back to you, without losing its settings.
  • Pause everything — the master optimization switch stops Hungry Machines from applying any schedule at all.

While paused, schedules are still calculated in the background, so turning optimization back on takes effect at the next 30-minute boundary — no waiting for the next night.

If your indoor temperature ever drifts outside your comfort band during a scheduled off period — an unusually hot afternoon, say — the app overrides the plan and runs your HVAC to bring you back into range. Savings never come at the cost of leaving you uncomfortable.