What AIdan got wrong,
and when it was fixed

Was

A functional optimal range of 125–175 nmol/L for 25-OH vitamin D, presented in the same register as trial-derived figures.

Problem

The range had no authoritative source behind it. Worse, the US National Institutes of Health flags potential adverse effects from approximately 125–150 nmol/L — meaning mainstream guidance begins warning about harm at the bottom of the band that was being recommended as optimal.

Now

Endocrine Society: sufficiency at ≥75 nmol/L (≥30 ng/mL). Institute of Medicine: ≥50 nmol/L (≥20 ng/mL). NIH caution above approximately 125–150 nmol/L (50–60 ng/mL). Named body, both units, every time.

Found by asking AIdan to interpret the same blood panel six times and comparing the answers.

Was

AIdan gave the European Atherosclerosis Society very-high-risk threshold as 180 nmol/L, and told a reader their result of 216.8 exceeded it.

Problem

The EAS 2022 figure is 180 mg/dL, not 180 nmol/L — which is about 430 nmol/L. A result of 216.8 nmol/L is well below it. The EAS statement itself names mg/dL versus nmol/L confusion as a known clinical hazard, which is precisely the error that was made.

Now

ESC/EAS 2025 — risk rises slightly from 30 mg/dL or 62 nmol/L; risk-enhancing above 50 mg/dL or 105 nmol/L. ESC/EAS 2019 — very high above 180 mg/dL or 430 nmol/L. Each result is compared in the unit the lab reported; the units are never converted. Updated 7 October 2026: v52 used the National Lipid Association 2024 figures (high above 50 mg/dL or 125 nmol/L).

Found by checking a threshold AIdan quoted against the guideline it was attributed to.

Was

Free T4 1.2–1.5 ng/dL · Free T3 3.0–4.0 pg/mL · serum zinc 80–110 µg/dL · RBC magnesium 4.2–6.8 mg/dL.

Problem

UK laboratories report all four in different units. A normal UK zinc result of 13 µmol/L, read against “80–110”, looks severely deficient. A UK free T3 of 4.8 pmol/L, read against “3.0–4.0”, looks high when it is low-normal. The failure direction was consistently toward inventing a deficiency in someone who did not have one.

Now

Free T4 15.4–19.3 pmol/L · Free T3 4.6–6.1 pmol/L · zinc 12.2–16.8 µmol/L · RBC magnesium 1.73–2.80 mmol/L. Both units stated throughout.

Found by comparing the knowledge base against real UK laboratory report formats.

Was

Asked a question that mentioned somebody else — a husband, a friend — AIdan would sometimes answer as though that person were the account holder, including applying the account holder’s own test results to them.

Problem

This was the most serious fault found. It produced supplement doses and plans addressed to a person who had never been assessed, built on somebody else’s blood results.

Now

AIdan establishes who the question is about before answering. Where anyone other than the account holder is named or implied, it says so and does not proceed to specifics — no dose, no timeline, no plan, no protocol. General information about a condition is still available; the route to a proper assessment is given as the answer, not as an afterthought.

Found by deliberately asking third-party questions in a structured test, then verified across five repeat runs.

Was

AIdan would give a detailed answer — sometimes an entire protocol — and then add that this was not really its place to advise.

Problem

A limit stated after the content it should have prevented reads as compliant and behaves as though it is not. The reader has already got the plan.

Now

Where a limit applies, it is stated before the content it limits, never after.

Found in three of eight test cases. It was a house habit, not an occasional slip.

Was

AIdan loaded a client’s oldest forty messages as context, not their most recent forty.

Problem

A long-standing client was being answered against their earliest conversations, with everything since invisible — and the gap widened with every exchange.

Now

Most recent first. Fixed in code, not in wording.

Found by reading the database query rather than the comment above it, which said the opposite.

Was

Two full knowledge-base sections. Screening criteria, a first-morning urine test for HPU/KPU, and a correction protocol — zinc bisglycinate 30–60 mg, P5P 25–100 mg, magnesium glycinate, GLA/evening primrose oil, copper monitoring, avoidance of high-dose folic acid. The DUTCH interpretation tool carried it too, with proposed overlaps to hypermobile EDS and mast cell activation.

Problem

The kryptopyrrole assay it rests on has no established reliability, and the condition is not recognised in the peer-reviewed clinical literature. The structure is one this practice has since named explicitly as a pattern to watch for: an unvalidated assay, a reference range defined by the vendor, and a correction sold by the same vendor. Judged by that standard, it does not survive.

Now

Retired. It exists in the knowledge base only as a historical entry, and AIdan is forbidden from retrieving it as clinical guidance. Zinc and B6 status remain real and testable things — what was removed is the route that claimed to identify who needs them.

A blog post on this site still argues the original position: blog-pyroluria.html. It is left up deliberately, and this entry is the correction to it.

Was

Urinary HVA, VMA and 5-HIAA treated as proxies for central nervous system neurotransmitter status, with precursor repletion dosed against the urinary values. Present in the organic acids analyser and in the DUTCH sections.

Problem

There is no transport route that would make it work. Urinary dopamine is overwhelmingly of renal origin; urinary serotonin is gut-derived and diet-influenced. The clinical chemistry review of this area is explicit that these measurements help with central nervous system disorders “only when the measurements are made in cerebrospinal fluid” (Hyland, Clinical Chemistry 2008;54(4):633–41, PMID 18310141). Most of the literature offered in support is authored by the laboratories selling the test. The one independent attempt at criterion validation — in dogs, n = 100, comparing urinary neurotransmitters against four validated behaviour instruments — found no correlation at all, and concluded the testing “could not be validated” (Schmidt et al., Frontiers in Veterinary Science 2023;10:1124231).

Now

Retired as an inference. AIdan will not read a urinary neurotransmitter result as a statement about brain chemistry, and will not dose precursors against one. The interaction risk is now flagged where it belongs: 5-HTP and L-tryptophan with SSRIs, SNRIs, MAOIs, tramadol or triptans — which was the real hazard sitting underneath a test that could not justify the dose in the first place.

The canine study is named because it is the strongest independent test that exists, not because dogs are the population of interest. That the best available criterion validation is veterinary is itself the finding.

Was

Diosgenin, the steroidal saponin in wild yam, described as converting to progesterone in the body.

Problem

The conversion is real laboratory chemistry — the Marker degradation, a multi-step synthesis run in a reaction vessel. Humans do not have the enzymes to do it. A double-blind, placebo-controlled crossover trial of wild yam cream in 23 menopausal women found no change in serum or salivary progesterone and no difference from placebo on symptoms (Komesaroff et al., Climacteric 2001;4(2):144–50, PMID 11428178). Creams that do raise progesterone contain added USP progesterone — sometimes without saying so on the label.

Now

Retired. Diosgenin is given weak oestrogen-receptor–modulating activity and no progesterone-like activity whatsoever. Where a client reports a response to a wild yam cream, the first question is what else is in it.

Not a fringe concern: the US Federal Trade Commission charged seven online sellers of “natural progesterone” creams over unsubstantiated claims on 5 October 2007, having previously sent warning letters to 34 website operators.

Was

Transversus abdominis described as firing in anticipation of limb movement to pre-stabilise the spine — the feed-forward model from the Richardson, Jull and Hodges work — and used as the mechanistic justification for drawing-in and hollowing cues.

Problem

The model is contested and the underlying question — whether what is being measured is anticipatory or reactive — is not settled. Stating a disputed mechanism as the reason for an exercise makes the exercise look better evidenced than it is.

Now

The cueing is kept; the mechanism is dropped. AIdan teaches bracing and maintaining alignment, and does not explain it by reference to firing order.

The “300 repetitions to learn a movement, 3,000 to unlearn one” motor-learning figures were retired in the same pass, as folklore with no traceable source.

Was

DUTCH described as “the gold standard” for overtraining syndrome, and DUTCH Plus as “the gold standard for comprehensive hormone assessment”. Fat that won’t shift despite training put down to “hormonal dysregulation (cortisol elevation, oestrogen dominance, insulin resistance)” that “requires testing, not more exercise or less food”. CrossFit, HIIT and heavy lifting ruled out for anyone with a low cortisol pattern. A 30–45 minute post-training eating window called optimal. Under-fuelling said to cause “paradoxical fat retention around the waist and hips”.

Problem

No single test is the gold standard for “hormones”. Overtraining is diagnosed clinically, largely by exclusion: no hormonal, performance or biochemical marker meets the criteria for general use (ECSS/ACSM consensus, Meeusen 2013, PMID 23247672). “Oestrogen dominance” has no agreed definition, and putting stubborn fat down to hormones skips intake, sleep, stress, alcohol and medicines. No trial shows that choosing exercise by a cortisol pattern improves outcomes. In a meta-analysis of randomised trials, protein timing made no significant difference to strength or muscle once total protein was accounted for (Schoenfeld 2013, PMID 24299050). The waist-and-hips claim had no source.

Now

DUTCH Plus is a comprehensive dried-urine hormone panel; the test is chosen for the question. Stubborn fat is treated as having several contributors, and testing doesn’t replace looking at intake. Where someone isn’t recovering, AIdan suggests reducing intensity and volume first and reassessing, and says plainly that this is coaching judgement, not an evidence-based rule. Eating after training is sensible; no window is given. Under-fuelling is described as Relative Energy Deficiency in Sport (IOC consensus, Mountjoy 2023, PMID 37752011), without the fat-distribution claim.

Found by reviewing the training and hormone sections claim by claim. October builds are labelled by date, because their version numbers restarted. The v37 build in the same week changed which blood tests AIdan describes; that is a product change, not a correction, so it isn’t listed.

Was

“DUTCH Plus is essential. Never treat oestrogen dominance without first assessing…” Weight gain that resists diet put down first to hormones, including “cortisol (visceral fat distribution)” and “gut dysbiosis (microbiome affecting caloric extraction)”. “GUT SYMPTOMS: Always think GI-MAP. Bloating without infection diagnosis is incomplete.” “SKIN PROBLEMS: Think gut first. Gut-skin axis is direct.” A post-meal walk “reduces glucose spike 30%”; resistance training “improves insulin sensitivity 25–48%”. An exercise “J-curve” for immunity with minute cut-offs for “optimal” and “suppressed”. Exercise rules chosen by cortisol pattern, and a ban on high running mileage during gut infections, stated as clinical rules.

Problem

“Oestrogen dominance” has no agreed definition, and no single test is essential for hormone symptoms. Putting stubborn weight down to hormones skips intake, sleep, stress, alcohol and medicines. Many people with bloating have no infection, so a test is not always the next step, and a negative one is a useful answer. Gut–skin links are mostly observational. The two percentages had no source. The J-curve is a model built mainly on self-reported infections, and it is disputed. No trial shows that choosing exercise by a cortisol pattern improves outcomes.

Now

Hormone and weight questions start with the ordinary contributors and choose the test for the question; “oestrogen dominance” is not used as a diagnosis. Gut symptoms start with history, diet and red flags; a GI-MAP is considered when symptoms persist or the history points to infection. Skin: the gut is one possible contributor alongside skin causes, diet, stress and medicines, with a GP or dermatologist where appropriate. A short walk after meals lowers the post-meal glucose rise in short trials (Engeroff 2023); resistance training improves insulin sensitivity, with no percentage given. The J-curve is described as a contested model (Nieman 1994; Campbell & Turner 2018). Exercise choices made because someone isn’t recovering are labelled coaching judgement, not evidence-based rules. The GI-MAP zonulin add-on is no longer offered.

Found by reading the knowledge base’s symptom and exercise sections line by line after the entry above.