Ingredient analysis: maltodextrin powder¶

Mode 5 analysis · fruit-maillard-prep v5 · verdict: excluded · Disclaimer

Related: Skill user guide

Date: 2026-09-17 · Skill: fruit-maillard-prep v5 · Mode: 5, ingredient analysis

Request. "Analyse maltodextrin powder." Maltodextrin is already excluded in data/setup.md, listed in data/preferred-ingredients.md under Considered and declined, named in chemistry.md §5's solute table and in process-inputs.md, and carried in the SKILL.md standing constraints. It is deliberately not in references/ingredients.md, which is right: it is not an ingredient, it is a rule.

This is therefore a re-examination against the v4/v5 changes rather than a fresh assessment. Three of those changes bear on it: Brix became a safety hurdle and not only a kinetic one (§21.4); the fermentable load is now reported planned and worst case against an absolute gravity gate (§18); and §22 added the acrylamide and furan regime. The verdict does not move. Two of the stored reasons do, and one of them is currently overstated — corrections are in the chat block, not here.

No maltodextrin appears in data/batch-log.md. Batch 001 used DME, one rung up the same ladder, and its recorded lesson was "Drop DME". That is the most relevant evidence this brewery has.

1. Verdict¶

Exclusion stands. Do not use it in any stage. The reactivity case is definitional and decisive on its own. The two v4-era cases are stronger than they were: maltodextrin degrades the informativeness of the one instrument Stage B depends on, monotonically and invisibly; and it is the one input that can walk straight through the gravity gate, because the gate looks for stability over 4 to 6 weeks and Brett dextrinase works over 6 to 24 months. (High on reactivity, High on the gate, Med on the a_w magnitude)

2. Dominant roles¶

FAN nil, Prot nil, no acid, no pentose, no polyphenol, no carotenoid, no glycoside, no amygdalin.

3. Axes served and suppressed¶

Served: none. No axis in the skill's vocabulary gains from it.

Suppressed: the pyrazine and Strecker axes, indirectly. It dilutes reactive amine per unit of paste mass, raises viscosity and worsens diffusion at a fixed Brix reading, and per §1 the ceiling on this method is nitrogen loading, not sugar supply. Adding a poor sugar donor to a nitrogen-limited jar buys nothing even before the rest of the argument.

4. Stage, and the do-not-cook rule¶

Do-not-cook rule does not apply: there are no monoterpenes, thiols or esters to lose. There is nothing to lose at all.

5. Handling traps¶

5.1 Reactivity: below the bar the brewery already set¶

DE is defined as reducing power expressed as grams of dextrose per 100 g of solids, so moles of reducing ends per 100 g = DE / 180.16.

So maltodextrin DE 10 is roughly one third of DME's reactivity, and data/setup.md already classes DME as "available; not a Maillard input". The exclusion follows a fortiori, from arithmetic on the definition of DE, with no experimental question attached. (High)

There is also a steric penalty not captured above: the reducing end sits on a polymer, so the Amadori product forms on a polymer and diffuses accordingly. The table is therefore an upper bound.

5.2 Brix, a_w and the safety hurdle — with the claim sized honestly¶

chemistry.md §5 already lists maltodextrin as reading as Brix while barely depressing a_w. What is not written down anywhere is how much, and the answer matters, because §21.4 made that reading Stage B's safety hurdle rather than only its kinetic one.

A jar at 65 °Brix holds 65 g solids in 35 g water (1.943 mol). Letting f be the maltodextrin share of the solids, with the remainder as hexose and maltodextrin DE 10 at an average MW near 1600:

Growth minima: B. cereus ~0.92, C. botulinum and C. perfringens ~0.93.

The honest reading of that table, which is weaker than the line currently stored in setup.md. The ideal calculation overstates a_w: 65 % sucrose computes to 0.911 against a measured ~0.85, so ideal runs about 0.06 high for a sugar solution at this concentration. Published sorption data puts DE 10 at 65 % solids near 0.93 to 0.96 against the ideal 0.980, a smaller deviation. Applying that correction moves the 0.92 crossing from f ≈ 0.6 to somewhere around f ≈ 0.7 to 0.85 (Med-Low).

At any dose anyone would actually build — a body or bulking dose is 10 to 20 % of solids — the hurdle survives with margin: a_w ~0.855 to 0.868 against a 0.92 minimum. So the correct claim is not that maltodextrin defeats the safety hurdle. It is this:

The hurdle degrades monotonically with maltodextrin fraction while the refractometer reading does not move at all, so the instrument stops tracking the thing it stands for, in the direction of false confidence, with no signal that it has.

That is a weaker safety claim and a stronger rule. A hazard that fails loudly gets respected; an instrument that quietly stops measuring is how a "just a little, to hit the band" exception gets taken later. It is also the reason the exclusion should stay absolute rather than dose-limited: there is no benefit on the other side of the trade to justify managing a graded risk. (High on monotonicity and direction; Med-Low on the crossing fraction)

A third, smaller effect in the same direction: maltodextrin contributes no titratable acid, so it dilutes the jar's acid per Brix point as well as its reactive amine.

5.3 The gravity gate cannot see it — the v4/v5 point¶

Brettanomyces carries cell-bound and extracellular alpha-glucosidase active on alpha-1,4 glucans, characterised in B. lambicus, and the continued attenuation of Brett-aged beer over months is consistent with dextrin attack. Only the ~5 % of alpha-1,6 branch points resist. (Med-High on mechanism, Med on extent and rate)

At a conventional body dose of 15 g/L into 4.5 L (67.5 g), with the ×1.11 hydration factor on hydrolysis:

The gate is "a stable gravity read twice, 4 to 6 weeks apart". Dextrinase action runs over 6 to 24 months. Two stable reads in month three do not exclude continued attenuation in month twelve, so maltodextrin is an input the gate is structurally unable to catch.

That distinguishes it from the two other things that sit outside the gate. Stage D and Stage T additions also escape it, but they are known, sized and declared in output contract item 12. Maltodextrin escapes it unannounced, having been added precisely because it was believed to be unfermentable. That belief is the failure, not the sugar.

(The equivalent bottle-pressure figure is about 3.9 volumes. It is stated once for scale only: v4 retired that unit, because the brewery relies on the gate rather than calculating bottle pressure — which is exactly why an input the gate cannot see is the serious form of this problem.)

5.4 Native pH and low-acid class¶

Food-grade maltodextrin runs about pH 4.0 to 6.0 in solution; call it ~5.0 (Med-Low), which puts the top of the range in the low-acid class (>4.6, §21). In practice this never binds: TA is approximately zero and there is no buffering, so it does not enter the anchored base solve, and it would see no warm hold in any case. Recorded for completeness, not as a live constraint.

5.5 Section 22: reducing ends with no aroma return¶

Stage B's regime — 121 °C, 90 to 120 min, a_w 0.80 to 0.87, pH 6.15, asparagine-rich fruit — is a favourable acrylamide regime, and acrylamide forms from asparagine plus a reducing sugar. Maltodextrin supplies reducing ends and returns no aroma, so 100 % of its acrylamide contribution is cost. The absolute scale is small (Low, per §22's own confidence on scale), but the sign is unambiguous and there is nothing on the other side of it.

Furan: no ascorbate, so no contribution. The hot vent is standard regardless.

5.6 The inversion step brings the worst case forward¶

The v4 inversion step — phosphoric acid to a measured pH ~2.9, then 45 to 60 min at the boil — is acid hydrolysis conditions. Maltodextrin riding that step partially hydrolyses, raising its effective DE. Two consequences, both marginal: slightly more reactive in the jar than the label DE implies, and the fermentable worst case arrives sooner rather than being deferred to Brett. (Low)

5.7 Contaminant and toxicant screen¶

Nothing to report. No sulphite, no patulin, no oxalate, no cyanogenic glycoside, no mycotoxin route, no protein, no enzyme inhibition, no photo-activated compounds. Maltodextrin's problems are all systemic, none of them toxicological.

6. Worked comparison¶

Against the nearest catalogued ingredient by role: date¶

Date is the table's designated pure RS donor (RS 63, Suc 2, TA 3, low-acid) and the lowest-asparagine choice per the §22 index line. To place 100 g of Maillard-reactive reducing sugar into a Stage B jar:

1000 g into a 4.5 L prep is not a formula.

Against the nearest process input: DME, and batch 001¶

setup.md holds DME and classes it as "available; not a Maillard input". Batch 001 ran the experiment: 15 g of DME per 100 g of aronia, and the retrospective found it contributed roughly 40 % of the free amino nitrogen but only ~0.2 g of amino acids, in exchange for 11 g of the least reactive sugar present. Recorded verdict: "Drop DME."

Maltodextrin is DME with the FAN removed and the reactivity cut by a further two thirds. This brewery has already fired the better version of this ingredient, on its own evidence. That is the shortest form of the argument and the one worth remembering.

Against the nearest preferred-list ingredient: date again¶

Date is on preferred-ingredients.md as the primary reducing-sugar donor. There is no gap for maltodextrin to fill.

If the actual want is body¶

The in-house answer already exists and is free. Stage B melanoidins are non-fermentable, permanent and generated by the process you are already running (§1, §18); since v4, unbased Stage C also retains more pectin-derived polysaccharide and more colour than it did when it was based to 6.15.

Do not substitute lactose, which is the other thing reached for here and is worse in this system on two counts: it is a reducing disaccharide, so it enters the Maillard budget rather than sitting out of it, and Lactobacillus ferments it, so it is not unfermentable in a mixed culture either.

Confidence summary¶

Nothing outstanding¶

Unlike the marigold analysis, there is no open question here. No measurement, genus check or sourcing decision would change the verdict.