Tesla Energy Screen Analysis: Sign Bug, Range Lies, and a Better Design

This is not a vibe post about “the UI feels weird.” It is a concrete analysis of one Energy-tab failure mode, why it matters for trip safety, and what a better Energy screen + range algorithm should do.

1. Evidence

Context A — mountain regen window: crossing Monarch Pass, elevation loss produced strong regeneration. Projected range ballooned to 999+ mi.

Context B — hot-road range trust: Texas driving in temps in excess of 100°F, where cabin cooling and battery thermal management dominate consumption and projected range drifts from actual remaining range enough to break navigation trust.

2. Defect analysis

# What the UI appears to do
rate = mean_wh_per_mi_last_10   # correctly signed, e.g. -261.1
total_kwh = abs(rate * 10) / 1000
label = f"Consumed {total_kwh:.1f} kWh over the last 10 miles"

# What it should do
energy_kwh = (rate * 10) / 1000  # keep the sign
if energy_kwh >= 0:
    label = f"Used {energy_kwh:.1f} kWh over the last 10 miles"
else:
    label = f"Recovered {abs(energy_kwh):.1f} kWh over the last 10 miles"
# always also expose: Net {energy_kwh:+.1f} kWh

3. Why range estimates strand people

The sign bug is the easy defect. The harder one is range as a false fuel gauge.

In Texas heat above 100°F, real consumption is dominated by:

• cabin HVAC load
• battery thermal management
• speed / aero
• grade changes after the estimate was formed

When projected range is optimistic, navigation plans become fiction: the car “thinks” you can skip a charger, then the estimate collapses late. That is how people get stranded — not because physics is unknowable, but because the UI sells a single confident number instead of a condition-aware band.

4. Recommended Energy screen layout

Layout rules:

1. NOW column: net power (kW), signed Wh/mi, signed 1/5/10/50 mi energy totals.
2. Verb grammar: Used / Recovered / Net — never “Consumed” for negative windows.
3. Sparkline through zero: green below zero (regen), amber/red above (draw).
4. Condition chips: elevation, cabin A/C, speed, hot day — always visible when active.
5. Confidence chip: LOW / MED / HIGH. If not HIGH, nav must not use optimistic numbers.
6. RANGE panel: primary number = conservative nav range; secondary = median/optimistic; always show band.
7. Adjustment list: climate, thermal, grade, speed/wind, closed-loop bias — each with signed mile impact.
8. No 999+ as actionable fuel: fantasy projections stay diagnostic, never route inputs.

5. Range algorithm improvements

nav_range_mi = conservative_range(
    route_grade_profile,
    speed_plan,
    cabin_load,
    pack_thermal_state,
    ambient_temp,
    recent_prediction_bias,
)

if ambient_temp_f >= 100 or confidence < HIGH:
    route_using(nav_range_mi)          # P10 / conservative only
    show_arrival_soc_band = True
    allow_skip_charger = False
else:
    route_using(nav_range_mi)
    show_median_optional = True

6. Priority fix list (ship order)

1. P0 — Sign integrity: stop absolute-valuing window energy; fix Used/Recovered labels.
2. P0 — Shared quantity: Energy tab, trip cards, and nav planner must read the same signed energy model.
3. P0 — Nav conservative mode: under heat or low confidence, route on the low bound only.
4. P1 — Range band UI: replace single-point range with conservative/median/optimistic + confidence.
5. P1 — Adjustment transparency: show mile impacts for climate, thermal, grade, speed.
6. P1 — Closed-loop bias: learn predicted vs actual Wh every trip.
7. P2 — Cap fantasy projections: 999+ can exist as a diagnostic artifact, never as a trip plan input.
8. P2 — Arrival SoC band: “12–21% at charger” beats “17%” false precision.

7. Photo provenance

Bottom line

−261.1 Wh/mi × 10 mi is −2.6 kWh, not “Consumed 2.6 kWh.”

Fix the sign. Fix the verb. Then stop letting optimistic range theater drive navigation in heat and mountains. Energy UX is a safety surface — design it like one.