GAIA R&D Workbench · Hericium Texture Programme

§ 01 · ORIENTATIONWhat you are testing

100 chemistry-scored variants converged to a master formula. From that, three top variants emerged — all use dried Hericium erinaceus as substrate, all are 100% plant-based, all halal-compatible (apple cider vinegar substitution). Your job: execute one variant per trial, score sensory output against the predicted chemistry score, log feedback via @Gaiachef_bot.

The three variants — at a glance

V1 BALANCED · master formula

predicted 8.94/10 · COGS RM 14-18/portion · total 3h 30min (active 45min)

V5 KOJI-MAX · max ferment depth

predicted 8.72/10 · COGS RM 19-24/portion · total 15h 45min (12h overnight preferment)

V2 TEMPEH-MAX · steam-only profile

predicted 8.85/10 · COGS RM 16-21/portion · total 4h 00min · aquafaba surface wash · NO sear

Suggested trial order Execute V1 first (the master). If V1 passes sensory at ≥8/10 across all 9 nodes (see scoring card per variant), proceed to V5 + V2 in parallel. If V1 fails on a specific node, fix that node BEFORE attempting V5/V2 — same node will likely fail there too.

Chef-flex zones Marked with ± below — these are deliberate tolerance windows where your kitchen judgement applies. Hard-locked numbers (no ± symbol) must be hit exactly; deviation invalidates the trial.

🌱 100% plant-based Zero animal-derived. Verify every ingredient before opening package. No eggs, dairy, honey, fish-sauce.

🕌 Halal-compatible Apple cider vinegar ONLY (not wine vinegar). No mirin, sake, alcohol-trace. TGase microbial cert required.

⚠️ Kill-zones Wine vinegar · sake mirin · non-cert TGase · undeclared egg-albumin. If in doubt, /ask the bot before opening.

§ 02 · VARIANT V1V1 BALANCED — the master converged formula

The chemistry-rules engine ran 100 variants over 4 mutation loops. V1 emerged as the converged winner — every node ≥ 8.0/10, no extremes. Use this as the baseline for all sensory comparison.

L2-004 · BALANCED

V1 — the master

No extremes, all 9 nodes ≥ 8/10. The reference profile for everything that comes next.

8.94/10

predicted composite

COGS RM 14-18 · 40-46% margin

Total 3h 30min · Active 45min

§ 1 · Pre-treatment (the chitin softening sequence)

Alkali soak: 90min @ 22°C · pH 8.5 ± 0.2 · NaHCO₃ 8g + sea salt 12g + water 1500ml
Vinegar bath: 60min @ 22°C · apple cider vinegar 5-8% acetic · 150ml
Salt cure: 15min · sea salt 3g rubbed on surface · 1.5% by weight

§ 2 · Marinade (per 200g rehydrated substrate)

Shio-koji paste: 40g · F4 ferment · drives N1 free-Glu liberation (PMR7)
Tempeh powder: 40g · F18 disulfide · protein bind + Glu density
Miso: 0g (omitted in V1)
Amazake: 32g · F2 sugar-binder → Maillard precursor for sear
Soy protein isolate (90%): 40g · F3/F19 pI gel · PMR1 disulfide pool
Vital wheat gluten: 40g · F18 — the egg-cure chemistry replacement
Pea isolate: 0g (omitted in V1)
TGase (Activa RM): 0.25% of total protein mass ≈ 0.5-1g · F11 enzymatic cross-link
Aquafaba wash: 0ml (V1 sears bare)

§ 3 · Hydrocolloids + umami stack

Psyllium husk: 8g · F2 fiber-gel
Konjac glucomannan: 12g · F17 alkali-deacetylated · X5 network density
Methylcellulose (E461 A4M): 6g · F12 hot-set — DO NOT omit in V1 · X4 water lock
κ-Carrageenan: 4g · F2/F10 ion-bridge · X2 cross-link
Kombu (dried): 8g · N1 free-Glu primary (PMR4)
Shiitake powder: 12g · N2 guanylate · drives M1 ×30 synergy
Nutritional yeast: 32g · N1 + B-complex

§ 4 · Cook

Steam: 85°C × 10min low-pressure · sets F12 methylcellulose + F18 disulfide
Sear: 180°C × 60sec · Maillard ONLY (N4 volatiles) · do not cook through
Cut direction: WITH-SPINE — lobster-tear anisotropy (PMR6)

§ 5 · 13-step process spine

1

Rehydrate substrate30min @ 80°C

60g dried Hericium + 600ml hot water (80°C), submerge 30min until pliable. Reserve soak water — it becomes the kombu broth booster in step 10.

Mushroom yields to thumb press · no dry core
2

Press cycles ×315min

Pasta roller settings 1 → 3 → 5. Press, fold, press, fold, press. Drives out interstitial water + opens fiber direction.

Visible fiber alignment (X3) · ≥30% mass-water expelled
3

Alkali pre-soak90min @ 22°C

NaHCO₃ 8g + sea salt 12g + water 1500ml. Stir until pH meter reads 8.5 ± 0.2. Fully submerge mushroom.

pH 8.5 ± 0.2 verified with meter · no acrid smell
4

Vinegar bath (pre-tame)60min @ 22°C

Drain alkali, rinse cold water. Submerge in 150ml apple cider vinegar (5-8% acetic). Slow chitin hydrolysis without heat.

Mushroom firms slightly · surface no longer slimy
5

Salt cure15min @ 22°C

Drain vinegar, blot dry. Rub 3g sea salt + 1g sugar over surface. Rest 15min.

Surface beaded with expressed water · firm to touch
6

Marinade build10min

Whisk shio-koji + tempeh pwd + amazake + soy iso + wheat gluten + TGase into a thick slurry.

No lumps · thick like yogurt · TGase fully dispersed
7

Hydrocolloid integration5min

Sift psyllium + konjac + methylcellulose + κ-carrageenan into marinade. Whisk vigorously.

Smooth · no clumps · holds a soft peak
8

Vacuum-tumble (or massage)20min vac · 45min hand

Vacuum bag substrate + marinade. Tumble 20min OR hand-massage 45min until marinade is fully absorbed.

Marinade no longer pools · substrate visibly plumped
9

Preferment (V5 only — SKIP for V1)12h @ 20-22°C

V5 only. Skip for V1.
10

Umami stack soak30min

Combine kombu 8g + shiitake powder 12g + nutritional yeast 32g + 200ml reserved soak water from step 1. Brush over substrate.

Substrate surface mahogany-brown · umami liquid fully absorbed
11

Cut WITH-SPINE5min

Slice substrate parallel to fiber direction (X3 lobster-tear anisotropy). 1.5–2cm thick.

Visible parallel fibers on cut face
12

Steam10min @ 85°C

Steam at 85°C low-pressure. F12 methylcellulose sets, F18 wheat-gluten disulfide locks.

Internal temp 80°C+ · surface firm but not rubbery
13

Sear60sec @ 180°C

Cast iron 180°C, neutral oil thin. Sear 60sec. MAILLARD ONLY — do NOT cook through.

Surface deep amber-brown · no smoke · internal temp unchanged

§ 6 · Sensory scoring card (fill after step 13)

Score each node 1-10. Composite = average of 9 nodes. Pass ≥ 8.0/10.

Node	What you're scoring	Target	Your score
N1 Free Glu	Umami body — savouriness on first chew	≥ 8.5	___ /10
N2 Ribonucleotide	Umami amplification — deep mushroom-stock	≥ 8.0	___ /10
N3 Aroma volatiles	Earthy mushroom + ferment perfume	≥ 7.5	___ /10
N4 Maillard volatiles	Sear crust — toasted bread / browned-butter	≥ 8.5	___ /10
X1 Chitin softness	No cloth/rubbery jaw-fatigue	≥ 8.5	___ /10
X2 β-glucan + mannoprotein	Cohesive bite — no falling-apart	≥ 8.0	___ /10
X3 Fiber anisotropy	Lobster-tear shred · visible directionality	≥ 9.0	___ /10
X4 Water-content gradient	Juicy interior · dry-crust exterior	≥ 8.0	___ /10
X5 Network density	Springback after compression · steak-like	≥ 8.5	___ /10
COMPOSITE (avg of 9)			___ /10

Photo capture (every trial) 1) overhead whole-portion · 2) cross-section showing fiber anisotropy · 3) sear-crust close-up · 4) interior tear-test (pull apart, photo the strands) · 5) bite-test plate (one bite missing)

Failure-mode quick-fix

Texture rubbery → reduce konjac 12g → 8g; verify TGase ≤ 0.25%
Falls apart on bite → verify methylcellulose was 6g (NOT omitted), increase TGase to 0.30%
Cloth-mouthfeel → extend alkali soak to 120min, verify pH 8.5
Beany off-note → reduce soy iso to 30g, increase tempeh pwd to 50g
No umami depth → verify shiitake powder ≥12g (NOT shiitake flakes — must be dried-and-ground)

§ 03 · VARIANT V5V5 KOJI-MAX — overnight ferment

V5 doubles down on the koji enzymatic path. Identical pre-treatment + identical hydrocolloid stack as V1, but the marinade is heavier on shio-koji + soy isolate + tempeh, and gets a 12-hour preferment at room temperature before steam. Deeper umami, longer sear, higher cost.

L2-019 · KOJI-MAX

V5 — max ferment depth

12h preferment unlocks shio-koji enzymes × 2 vs V1. Worth the time only if V1 passed at ≥8.5/10.

8.72/10

predicted composite

COGS RM 19-24 · 37-41% margin

Total 15h 45min · Active 55min

§ 1 · Same pre-treatment as V1

Alkali soak 90min pH 8.5 → vinegar bath 60min → salt cure 15min. No changes from V1.

§ 2 · Marinade — koji-max stack

Shio-koji paste: 80g (× 2 V1) · max enzymatic Glu release
Tempeh powder: 100g (× 2.5 V1) · Cys-rich protein pool
Miso: 0g
Amazake: 0g (omitted — koji is doing the sugar work)
Soy protein isolate: 80g (× 2 V1)
Vital wheat gluten: 40g · same as V1
Pea isolate: 0g
TGase: 0.25% · same as V1

§ 3 · Same hydrocolloids + umami as V1

Psyllium 8g · konjac 12g · methylcellulose 6g · κ-carrageenan 4g · kombu 8g · shiitake pwd 12g · nutri yeast 32g.

§ 4 · The 12-hour preferment (the V5 signature)

9

PREFERMENT — V5 unique step12h @ 20-22°C

After step 8 (vacuum-tumble), rest the marinated substrate at room temperature 20-22°C for 12 hours in vacuum bag. Shio-koji enzymes deepen Glu liberation, soy/tempeh proteins begin pre-disulfide bond shuffling.

Aroma shifts from raw-bean to soy-sauce / mushroom-stock depth · slight bag-bloat is normal (CO₂ from koji metabolism)

Preferment safety 12h @ 22°C is enzymatic, not microbial-spoilage range — koji/salt/acid environment is naturally hostile to pathogens. BUT: if room exceeds 26°C or bag-bloat is excessive (>30% volume), abort + score as failed trial.

§ 5 · Cook — longer sear

Steam: 85°C × 10min · same as V1
Sear: 180°C × 90sec (V1 was 60sec) · the deeper ferment substrate browns slower

§ 6 · V5 scoring card

Specifically benchmark V5 vs V1 on N1 + N2 (umami nodes). V5 must beat V1 by ≥ 0.3 here to justify the 12h cost.

Node	V1 score (reference)	V5 target	V5 actual
N1 Free Glu	V1 ___	V1 + 0.5	___ /10
N2 Ribonucleotide	V1 ___	V1 + 0.3	___ /10
N3 Aroma volatiles	V1 ___	V1 + 0.5 (koji perfume)	___ /10
N4 Maillard volatiles	V1 ___	≥ V1 (90s sear)	___ /10
X1-X5 (texture)	V1 ___ avg	≥ V1 (no regression)	___ /10
V5 wins iff: umami sum-of-(N1+N2+N3) ≥ V1 sum + 1.0			PASS / FAIL

If V5 fails — most common cause Preferment temp drifted above 24°C → enzymes overshoot → bitter aftertaste. Use a fridge thermometer in the preferment room before next attempt.

§ 04 · VARIANT V2V2 TEMPEH-MAX — steam-only · aquafaba surface

V2 is the no-sear profile. Useful for bao filling, kimbap, sushi roll, or any application where Maillard surface isn't wanted. Aquafaba wash gives a golden surface from steam alone. Longer alkali soak (120min vs 90min) compensates for the missing sear's textural punch.

L3-013 · TEMPEH-MAX

V2 — steam-only profile

For applications where you DON'T want a crust. Tempeh-forward, aquafaba surface, NO sear step.

8.85/10

predicted composite

COGS RM 16-21 · 38-43% margin

Total 4h 00min · Active 50min

§ 1 · Pre-treatment — longer alkali, shorter vinegar

Alkali soak: 120min (× 1.33 V1) @ 22°C · pH 8.5 (same recipe)
Vinegar bath: 15min (¼ V1) · less acid because no sear to balance
Salt cure: 15min · same as V1

§ 2 · Marinade — tempeh-forward stack

Shio-koji paste: 40g · same as V1
Tempeh powder: 100g (× 2.5 V1) · the V2 signature — drives F18 disulfide via Cys-rich tempeh protein
Miso (shiro white): 40g (V1 was 0) · adds Glu + Maillard precursor for steam-only browning
Amazake: 32g · same as V1
Soy protein isolate: 60g (× 1.5 V1)
Vital wheat gluten: 40g · same as V1
Pea isolate: 60g (V1 was 0) · F22 co-extruded protein matrix
TGase: 0.25% · same as V1
Aquafaba wash: 120ml · brushed on surface pre-steam (the V2 signature)

§ 3 · Hydrocolloids — adds alginate

Psyllium: 8g · same as V1
Konjac: 12g · same as V1
Methylcellulose: 6g · same as V1 (do NOT omit)
κ-Carrageenan: 4g · same as V1
Sodium alginate: 2g (V1 was 0) · F10 Ca²⁺ ion-bridge for surface gel from aquafaba wash

§ 4 · Cook — steam only, longer

Surface wash: Brush 120ml aquafaba over substrate PRE-steam. F2 sugar-binder + F19 protein gel → golden glaze without Maillard.
Steam: 85°C × 12min (V1 was 10min) · longer cook compensates for missing sear-temp gradient
Sear: NONE — V2 is steam-only

§ 5 · V2 scoring card

V2 wins iff: N1 + X1 + X5 ≥ V1 baseline, AND surface visually golden/glossy (aquafaba working). N4 will be lower than V1 (expected, no sear).

Node	What you're scoring	Target	Score
N1 Free Glu	Heavier tempeh + miso should boost this	≥ V1 + 0.2	___ /10
N4 Maillard	Aquafaba glaze ONLY — no sear	5-6 acceptable	___ /10
X1 Chitin softness	120min alkali should soften more than V1	≥ V1	___ /10
X5 Network density	Pea isolate + alginate should add springback	≥ V1	___ /10
Surface visual	Golden, glossy, no dull spots	qualitative	PASS / FAIL

Best applications for V2 Bao filling · kimbap · sushi roll · cold sandwich · poke bowl · congee topping · anywhere a hard crust would fight the dish texture.

§ 05 · INGREDIENTSBulk-order quick list — one trial batch

Order everything below before the first trial. Most items have 1-3 day lead time except pea isolate, TGase, methylcellulose (~7d via Modernist Pantry). One trial batch yields 4-6 portions.

Ingredient	Amount	Supplier	Cost (RM)	Halal
Substrate & pre-treatment
Dried Hericium erinaceus	60g → 200g rehydrated	Ganofarm (Tg Sepat) / Pinxin supply	8-12	✅
Baking soda (NaHCO₃)	8g per soak	Mercato / Lazada	4 / 500g	✅
Sea salt (coarse)	12g soak + 3g cure	any	6 / 500g	✅
Apple cider vinegar 5-8%	150ml per bath	Bioway / Costco	15-22 / 500ml	✅ NOT wine
Protein binders (the egg-cure chemistry replacement)
Shio-koji paste	40-80g	Hello Tempayy KL / Don Don Donki	28-45 / 250g	✅
Tempeh powder	40-100g	Hello Tempayy custom dry-grind	30-40 / 200g	✅
Miso (white shiro)	0-40g	Bioway / Marukome	38-55 / 500g	✅
Amazake (rice ferment)	0-32g	Hello Tempayy / Don Don Donki	32-42 / 250g	✅
Soy protein isolate 90%	40-80g	Bioway / Modernist Pantry	48-65 / 1kg	✅ cert
Vital wheat gluten	40g	Bioway / Bake With Yen	25-35 / 1kg	✅
Pea protein isolate 85%	0-60g	Modernist Pantry / Sapro MY	95-130 / 1kg	✅
Transglutaminase (TGase)	0.25% protein	Modernist Pantry (Activa RM)	180-220 / 100g	⚠️ verify cert
Hydrocolloids
Psyllium husk	8g	Bioway / Pharmacy	18-26 / 250g	✅
Konjac glucomannan	12g	Modernist Pantry / Lazada	38-55 / 250g	✅
Methylcellulose A4M	6g (V1/V5/V2)	Modernist Pantry	65-90 / 250g	✅
κ-Carrageenan	4g	Modernist Pantry / Bioway	48-65 / 250g	✅
Sodium alginate	0-2g (V2 only)	Modernist Pantry	55-75 / 250g	✅
Umami stack
Dried kombu (kelp)	8g	Don Don Donki / Bioway	28-42 / 50g	✅
Dried shiitake powder	12g	Ganofarm / Bioway	32-48 / 100g	✅
Nutritional yeast	32g	Bioway / Mercato	38-52 / 200g	✅

→ Full Ingredient Vault (verified Shopee MY links + tiered suppliers + halal) → Homemade Ferments (Shio-koji · Amazake · Tempeh Powder)

§ 06 · COMPLIANCEHalal · plant-only · kill-zones

Every variant on this page is engineered to be 100% plant-based AND halal-compatible. The rules below are non-negotiable — if in doubt, stop and ask via @Gaiachef_bot /ask before opening the package.

Plant-only audit

Zero animal-derived ingredients. No eggs, no dairy, no honey, no fish sauce, no oyster sauce, no bonito, no gelatin, no lanolin.
Verify every package label before opening — "vegan" or "100% plant" mark required. If unclear, escalate via bot.
Replacement chemistry is exact: vital wheat gluten + TGase + κ-carrageenan replace egg-white's disulfide-bond function (PMR1 — disulfide universal rule).

Halal substitution audit (JAKIM-facing)

Conventional ingredient	Plant + halal substitution	Reason
Rice wine vinegar	Apple cider vinegar 5-8%	Wine vinegar carries alcohol trace — JAKIM kill-zone
Sake / mirin	Amazake (rice ferment, 0% alcohol)	Mirin is fermented with alcohol — kill-zone
Egg-white (egg cure)	Aquafaba + vital wheat gluten + TGase	Same disulfide chemistry, plant-derived
Bonito flakes	Dried shiitake powder + kombu	PMR4 — plant Glu beats animal Glu on chemistry
Fish sauce	Tamari + shio-koji blend	Glu/IMP analog without anchovy
Standard TGase	Microbial Streptomyces TGase (cert required)	Some TGases are mammalian-derived — kill-zone

Kill-zones — never use Rice wine vinegar · Chinese black vinegar (alcohol-fermented) · sake · mirin · cooking wine · animal-derived rennet · standard gelatin · standard TGase without microbial cert · "natural flavour" without origin disclosure.

§ 07 · FEEDBACK LOOPHow your feedback shapes the next iteration

Every trial you log via @Gaiachef_bot feeds back into Zennith OS overnight. If 3+ trials surface the same issue, the system auto-promotes an AVOID rule that bends the next R&D iteration. Your kitchen judgement directly trains the chemistry engine.

The trial loop

Open @Gaiachef_bot on Telegram. Run /v1 (or /v5 · /v2) to pull the spec.
Execute the 13-step process. Tick off each verify-check as you go.
Fill the sensory scoring card after step 13 (or step 12 for V2).
Take the 5 photos: overhead · cross-section · sear-crust · interior-tear · bite-test.
Log feedback: /feedback V1 trial 1 — texture rubbery, X1=6.5, suspect konjac too high, will retry at 10g
For questions to R&D lead (Jenn): /ask Can we substitute kappa-carrageenan with iota for V2 if we run out of kappa?
Bot acknowledges; entry is logged to feedback.jsonl; bridge cron picks it up at 02:55 MYT.

What happens to your feedback (the compound loop)

Texture issues (rubbery, falls-apart, cloth-mouth) — 3+ matching reports → Zennith OS writes an AVOID rule. The next autoresearch run weights AGAINST whatever combination caused it.
Cost breaches (above target margin) — surfaces in marketing's next campaign brief (caps premium price).
Halal concerns — escalates to JAKIM audit trail + may trigger ingredient swap.
Process clarifications — folded back into doc 19 (the source recipe sheet).

Cross-OS bridge — what you're plugged into GAIA OS runs the kitchen (this page, the bot). Zennith OS runs the marketing + creative + persona-sim (your feedback teaches it what the kitchen reality looks like). Single shared memory, two domains. See doc 20 if you want the full architecture.