April 07, 2026

Generated lab notebook from daily_summary, plan_journal, and setpoint audit data. It is intentionally chronological and may include in-progress cycles before validation.

🌅 Morning Cycle (6:02 AM) — iris-20260407-0600

Result: The morning plan held the greenhouse stable through dawn and the dispatcher kept full 10-parameter coverage intact. By noon the system had already pushed the 11:04 AM batch successfully and active setpoints remained non-zero. However, the day is materially hotter and drier than the morning posture assumed: south zone is already at 87.4F and 2.81 kPa VPD with the controller in COOL_S3. The early deployment experiment is still directionally valid for Wednesday, but today has already entered a physics-limited afternoon.

Reflection

Validating previous cycle: iris-20260406-1800

Previous hypothesis: Since minute-staggering restored live params but broke plan coverage, a same-timestamp complete plan should preserve both dispatcher success and chart continuity now that the system is no longer in a zeroed-param state. Tuesday frost protection should hold with temp_low=58, while Wednesday will be physics-limited and likely see 4-6h VPD stress even with max misting. Result: The morning plan held the greenhouse stable through dawn and the dispatcher kept full 10-parameter coverage intact. By noon the system had already pushed the 11:04 AM batch successfully and active setpoints remained non-zero. However, the day is materially hotter and drier than the morning posture assumed: south zone is already at 87.4F and 2.81 kPa VPD with the controller in COOL_S3. The early deployment experiment is still directionally valid for Wednesday, but today has already entered a physics-limited afternoon. Score: 7/10

New finding: Same-timestamp full transitions remain publishable and dispatcher coverage is holding, but hot-day midday replans still need to react aggressively when south-zone VPD outruns the greenhouse average. → Added to Lessons Learned

Hypothesis

Conditions: Scheduled 6 AM cycle. Greenhouse entered the run healthy at roughly 57F and 0.7 kPa VPD with HEAT_S1_HUM_IDLE active. Forecast shows a cold dawn followed by two increasingly dry afternoons, with Wednesday the harshest window at roughly 74-75F and 10-14% RH. No deviation trigger, but dry-air stress is the dominant risk. Testing: Test whether earlier same-timestamp dry-day deployment, starting Wednesday at 9 AM instead of waiting until noon, reduces peak-afternoon VPD stress while preserving clean full-transition plan coverage and stable dispatch behavior. Expected outcome: Peak Wednesday VPD stress should stay below 6 hours above target, Tuesday should remain manageable with no cold stress at dawn, and plan coverage should remain complete with non-zero active setpoints after dispatch.

Setpoints

Tuesday April 07

Primary crop-band changes:

Tactical tunable changes:

Wednesday April 08

Primary crop-band changes:

Tactical tunable changes:

Thursday April 09

Primary crop-band changes:

Tactical tunable changes:

☀️ Midday Cycle (12:02 PM) — iris-20260407-1200

Result: The midday reactive replan pushed the aggressive dry-day posture successfully at 1:29 PM and held full plan coverage. It did not fully prevent afternoon stress, but it appears to have capped the worst runaway and carried the greenhouse into evening recovery with active setpoints still intact. Current active setpoints remain at the intended aggressive values, so the replan succeeded operationally even though the afternoon stayed physics-limited.

Reflection

Validating previous cycle: iris-20260406-1800

Previous hypothesis: Since minute-staggering restored live params but broke plan coverage, a same-timestamp complete plan should preserve both dispatcher success and chart continuity now that the system is no longer in a zeroed-param state. Tuesday frost protection should hold with temp_low=58, while Wednesday will be physics-limited and likely see 4-6h VPD stress even with max misting. Result: The midday reactive replan pushed the aggressive dry-day posture successfully at 1:29 PM and held full plan coverage. It did not fully prevent afternoon stress, but it appears to have capped the worst runaway and carried the greenhouse into evening recovery with active setpoints still intact. Current active setpoints remain at the intended aggressive values, so the replan succeeded operationally even though the afternoon stayed physics-limited. Score: 7/10

New finding: Zone-triggered midday replans can stabilize the controller posture and limit further runaway on extreme dry afternoons, but they do not eliminate stress once the greenhouse is already deep into a physics-limited window. → Added to Lessons Learned

Hypothesis

Conditions: Scheduled 12 PM cycle. The greenhouse is already in a physics-limited afternoon with south zone at 87.4F and 2.81 kPa VPD while the controller is in COOL_S3. Forecast remains brutally dry through this evening and even harsher on Wednesday, with no deviation trigger but clear evidence that zone-level midday stress is outrunning the greenhouse average. Testing: Test whether a true midday reactive replan, triggered by south-zone VPD overshoot rather than greenhouse-average conditions, reduces same-day late-afternoon VPD stress and improves Wednesday’s starting humidity posture before the 9 AM aggressive deploy. Expected outcome: This afternoon should avoid further south-zone runaway beyond the current peak, active setpoints should remain complete and non-zero after dispatch, and Wednesday’s 9 AM conditions should begin from a more humid baseline than today’s noon overshoot.

Setpoints

Tuesday April 07

Primary crop-band changes:

Tactical tunable changes:

Wednesday April 08

Primary crop-band changes:

Tactical tunable changes:

Thursday April 09

Primary crop-band changes:

Tactical tunable changes:

🌆 Evening Cycle (6:01 PM) — iris-20260407-1800

Result: [dominant: vpd_high_stress] Catastrophic failure on April 10: 0.0% compliance, 75.3h total stress (heat 32.6h, cold 1.6h, vpd_high 41.1h, vpd_low 0h), cost USD 7.80, water 97 gal. The plan was designed for April 9 (scored 95.3 there) but carried into April 10 where tight nighttime bands (vpd_high ~0.6-0.8) meant the engage threshold of 1.6 was above the band ceiling for much of the day. 41.1h vpd_high stress dominated — misting never triggered when VPD was above band but below engage. 32.6h heat stress from daytime overshoot with insufficient cooling bias.

Reflection

Validating previous cycle: iris-20260407-0600

Previous hypothesis: If I tighten VPD thresholds early and deploy the validated 60s pulse plus 30s gap misting posture only during the warm-dry windows, the greenhouse should keep Wednesday afternoon VPD stress materially lower than the recent dry-day baseline without wasting water overnight or during cold dawn periods. Result: [dominant: vpd_high_stress] Catastrophic failure on April 10: 0.0% compliance, 75.3h total stress (heat 32.6h, cold 1.6h, vpd_high 41.1h, vpd_low 0h), cost USD 7.80, water 97 gal. The plan was designed for April 9 (scored 95.3 there) but carried into April 10 where tight nighttime bands (vpd_high ~0.6-0.8) meant the engage threshold of 1.6 was above the band ceiling for much of the day. 41.1h vpd_high stress dominated — misting never triggered when VPD was above band but below engage. 32.6h heat stress from daytime overshoot with insufficient cooling bias. Score: 1/10

New finding: With tight crop bands (vpd_high=0.6-0.8 at night), mister_engage_kpa must be 1.3 at ALL times — not just daytime. The engage threshold of 1.6 was above the band ceiling, so misting never activated despite VPD being out of band. bias_cool=+3 to +4 must be active during ALL heating periods to prevent vent oscillation. This is now a triple-validated lesson. → Added to Lessons Learned

Hypothesis

Conditions: Scheduled 6 PM cycle. The greenhouse is recovering from an extreme dry afternoon, with active aggressive misting setpoints still in place and no evidence of dispatcher collapse. Forecast shows Wednesday as the hardest dry-air day of the cycle, with 74-75F and 10-12% RH through the core afternoon, followed by a still-dry but more mixed Thursday. Testing: Test whether the early 9 AM aggressive Wednesday deployment produces a lower midday-to-afternoon VPD ramp than Tuesday’s late-reactive pattern, while keeping full plan coverage and stable dispatch through the whole day. Expected outcome: Wednesday should begin from a stable overnight recovery posture, active setpoints should remain complete and non-zero after each dispatch window, and peak Wednesday VPD stress should be materially better than a noon-reactive-only pattern.

Setpoints

Tuesday April 07

Primary crop-band changes:

Tactical tunable changes:

Wednesday April 08

Primary crop-band changes:

Tactical tunable changes:

Thursday April 09

Primary crop-band changes:

Tactical tunable changes:

End-of-Day Summary

Climate

Stress Hours

• Heat stress (>85°F): 7.8h
• VPD stress (>2.0 kPa): 13.2h
• Cold stress (<55°F): 9.5h

Economics

Equipment Runtimes

Water

• Total: 127 gal
• Mister: 13 gal

Crop Health (Gemini Vision)

Hourly Pattern

7-Day Stress Context