Your crop goes through a complete case file, not a single-screen calculator. Every field you fill in exists because some engine will read it later on — there are no decorative questions. This page explains, with the same level of detail our own engines use, exactly what each part of the journey does: from the first form to your formula's recalibration months later.

Truly mandatory: specialty, path, plant, stage, and region. If any of these five is missing, the case isn't even created. The specialty decides which module catalog you can use — the filter compares each module's exclusive species against the five known specialties (rose, tulip, orchid, gourmet, resin & trichomes); any other crop falls under "generic agronomy". For Laboratory, if your specialty is Resin & Trichomes, Roses, Tulips, or Orchids, the plant name is set automatically; only in Gourmet do you type in the crop.

Only in Laboratory: before a single module is shown, a dedicated engine decides the risk level and the phase framework everything else will operate within — it doesn't touch ingredients, it sets the rules of the game.

The diagnosis turns into a persistent, idempotent program — if a program already exists for your case, it is returned as is, never duplicated. A real timeline is seeded: intake (day 0, completed immediately), evidence (day 3), scientific design (day 5, with a dynamic title depending on whether your plan includes formula), first response review (between day 7 and 14 depending on duration), recalibration (if your plan includes it, marked as optional), and case closure. If you signed up for "diagnosis only," the scientific-design milestone is marked complete immediately with the diagnosis's own evidence — because that plan does not include a full formula.

Before showing you anything, the real catalog was already filtered by available time (only modules that fit within your hours) and species (if your specialty is not empty, another species' exclusive modules are excluded) — so a tomato case never sees Resin & Trichomes exclusive modules.

You can choose manual (you check boxes module by module) or automatic with 3 quick combos: Essential (6 modules), Balanced (10, the recommended one), Maximum potential (16) — all three use exactly the same automatic production engine, with the same safety validations; it's not a simplified path. The default boosting mode is "maximum", but it's forced to "balanced" if your program is in conservative mode.

By pressing "generate", the system calculates an idempotency key (so a double click never generates two formulas or charges twice) and triggers the real chain: case context → botany → 25-rule matrix → module combination (beam search, 96 states) → mother architecture → maximum boosting (8 scheduled phases) → plant resource dynamics → compatibility → validation → self-ingredient safeguard → biochemistry → assembly → scaling to your actual batch → preparation planning → chemistry → scientific audit → structural audit (24 checks) → reports → persistence → delivery verification → output approval. Each engine can approve, warn, or completely block — if something does not add up, the flow stops right there and nothing is saved. The full detail, engine by engine with what each one audits, is in Formulas with real science.

After persistence, your formula contains: the modules exactly as they ended up (phase, capabilities, marginal score, why each was chosen), the consolidated ingredients (total amount, unit, which modules it comes from, its NPK composition), a full summary (preparation plan, scaling, structured report, scientific package) and the record of audit that approved it.

If you've already paid, you see: quick stats (structural potential, modules, phases, functional coverage); the quantitative scale panel (plants, ml of concentrate, scaled modules vs. “Withheld,” density per m², with the warning in red "Do not spread this formula across more plants than declared"); the program phase by phase with each expandable module, its exact ingredient table, and the full step-by-step preparation (numbered instruction, duration in hours, temperature range, target pH, how to store); a collapsible technical panel; and the final scientific status, honestly reminding that it is still "a prototype — still on its way to full experimental validation". If you have not paid, you only see the module and phase count, with the option to unlock the rest.

You log notes and photos week by week — up to 5 photos per week on the Pilot plan (unlimited on Premium, with a 5 MB limit per photo in JPG/PNG/WEBP) — and you set up a network of witness plants: a representative subset you observe closely. For Yolia 40, 5 witness plants are recommended for uniform batches and 10 for large batches or ones spread across more than 5 sectors or more than 5,000 m²; for Laboratory, if you have a single plant, that same plant is your tracking unit; with 2 to 4 plants all are observed; with more, a sample of 3, 5, or 10 is recommended depending on size. Each case also has access to a contextualized AI Assistant, with its own message history saved as case events.

In all three cases, the previous formula is kept intact and the new version is linked to the same program, with its own complete pass through the 21 engines, its own audit, and its own integrity verification — it is never a simple patch on the old formula. When the system recalibrates, a visible event remains that you can always check: total transparency, never a silent change.

Upon reaching the final week of Yolia 40 (calculated between 4 and 104 weeks depending on your crop), the cycle's final consolidated formula is generated. In Laboratory, the case closes with a declared result and a complete summary of the accumulated evidence.