Twenty new articles went up in the Library of Ashurbanipal today, and one old one was rebuilt. The centre of it is nootropics — but it took twenty articles instead of one because almost every honest question about a nootropic turns out to be a question about something else: an enzyme, a transporter, a receptor, a study design, or a claim nobody ever tested.
Here is what is actually in each of them.
Nootropics
Nootropics opens with Corneliu Giurgea's 1972 definition and specifically his fifth criterion — absence of the usual pharmacology of psychotropic drugs. No sedation, no motor stimulation, very low toxicity. That criterion is the one the market ignores, and keeping it predicts the failure mode: a stimulant that makes you work longer is not a nootropic, and the debt comes due later.
The evidence is ranked by how well it holds up rather than how it sells. Caffeine with L-theanine, roughly 50–100 mg to 100–200 mg, is the best-replicated consumer-level result in the field and the cheapest. Creatine works clearly where brain phosphocreatine is low — sleep deprivation, vegetarian diets — and close to nothing in rested omnivores. Correcting an actual deficiency outperforms everything else on the list. Bacopa has real randomised evidence and needs eight to twelve weeks to show it. Ginkgo's large prevention trials were negative. Most of the racetam shelf is mechanistically interesting and clinically unproven in healthy adults; piracetam's Cochrane reviews in dementia have repeatedly come back insufficient.
The cholinergic section explains why inhibiting acetylcholinesterase harder stops helping: the enzyme is not brain-specific, so past a point you get nausea, salivation, cramping and bradycardia. And it names the interaction that actually sends people to hospital — stacking two cholinesterase inhibitors. Huperzine A with galantamine, or either alongside prescribed donepezil or rivastigmine. Additive on the same enzyme, steep toxicity curve. Bacopa is a weak one too, which is why it belongs in that arithmetic.
The mechanisms underneath it
Ampakines, Ketamine and Rapid Synaptogenesis — two drug classes reaching the same place from opposite directions. Ampakines potentiate the AMPA receptor directly; ketamine blocks NMDA preferentially on fast-spiking interneurons, which disinhibits pyramidal cells and produces an AMPA surge as a consequence. Both converge on BDNF, TrkB, mTORC1 and new dendritic spines within 24 hours. The experiment that ties them together is that blocking AMPA with NBQX abolishes ketamine's antidepressant effect — NMDA blockade is the trigger, AMPA throughput is the effector.
It also covers the ampakine class honestly: forty years, a real brainstem application restoring respiratory drive without reversing analgesia, and no licensed cognitive drug. And it states the ketamine harm that gets under-reported — ulcerative cystitis, which in heavy long-term use is irreversible and has required bladder surgery in people in their twenties. Urinary symptoms in a regular user are a stop sign, not a side effect to tolerate.
5-HTP, Serotonin Precursors and the Chemical Imbalance Theory — 5-HTP bypasses the rate-limiting enzyme, which is the clever part. The problem is the next enzyme: AADC is everywhere in the periphery, so most of an oral dose becomes serotonin in the gut wall and bloodstream before reaching the brain. Research protocols pair it with carbidopa for exactly this reason; the retail product does not. That peripheral serotonin is also the chronic concern, because sustained 5-HT2B stimulation is the established mechanism behind fenfluramine valve disease.
Then the history: iproniazid and imipramine were accidents, reserpine depleted monoamines, and the hypothesis was built by inferring a cause from what helps — aspirin relieves headache, headaches are not an aspirin deficiency. Schildkraut himself called it a reductionist simplification. The article holds three things true at once: depression is not a demonstrated serotonin deficiency, antidepressants do separate from placebo across 522 trials, and withdrawal was genuinely underplayed for decades — which is why nobody should stop an SSRI abruptly, for pharmacological rather than psychological reasons.
L-Methylfolate — rebuilt. The existing BH4 and MTHFR material now sits alongside the schizophrenia adjunct evidence, where the rationale is completely different: not monoamine synthesis but one-carbon metabolism, raised homocysteine, and homocysteine acting at the NMDA receptor's glycine site. One molecule, two mechanisms, two literatures. The hair-loss claim is sorted into tiers — folate deficiency genuinely causes shedding, methotrexate causing hair loss shows the pathway matters, and nothing supports an effect on pattern baldness. There is also a table showing why the imbalance frame fails for schizophrenia differently than for depression: the dopamine evidence is much stronger, but what imaging shows is elevated presynaptic synthesis capacity as a final common pathway, which is a far more disciplined claim than "a chemical imbalance."
Urine Testing and the Long Search for a Psychiatric Biomarker — a century of looking. It starts with the one real success: in 1934 a mother told a doctor her two children's urine smelled strange, ferric chloride turned it green, and phenylketonuria became a cause, a test and a treatment. Nearly every psychiatric diagnosis since has been searched for the same way and found nothing.
The centrepiece is Shulgin's own account of the "Pink Spot" era, where researcher A confirmed it, B found it in everyone, C in nobody, D blamed hospital diet, E blamed time of day, and F counted the papers. Then the mauve spot — reported from Canada so confidently they named the disease "Malvaria." Shulgin and a colleague could not replicate it on twenty consecutive admissions, worked out that every positive patient was medicated, took a large dose of chlorpromazine themselves, and both tested positive, then negative three days later. The marker was a drug metabolite. The article closes the loop: that mauve factor was renamed kryptopyrrole and is still sold today as "pyroluria." It also covers why urinary neurotransmitter testing cannot work, and the immunoassay false positives that cost people jobs and custody when nobody runs the confirmatory test.
Brains Lighting Up — nothing lights up. BOLD is a blood measurement lagging neural events by four to six seconds, over voxels containing hundreds of thousands of neurons, and excitation and inhibition both raise the signal. "Lights up" is always a subtraction between two conditions; change the control and the map changes. The article walks the four documented failures — reverse inference, the voodoo-correlations circularity, the dead salmon that showed significant activation without multiple-comparison correction, and the 2016 finding that standard cluster inference could produce 70% false positives — then says plainly what imaging is genuinely excellent for, and gives a seven-point checklist for reading any study that arrives with a picture of a glowing brain.
Brain Training, Neuroplasticity and the 75,000-Year Frame — the Lumosity settlement with the actual claim list, and Owen's study of 11,430 people who improved at every trained task and transferred to nothing. Both 2014 consensus letters are treated as good faith, because the disagreement is about how much weight near transfer carries. What demonstrably does change brain structure is in there too, and it is never an app: London taxi drivers' posterior hippocampi after two to four years of the Knowledge, three months of learning a juggling cascade, a year of aerobic exercise. Sustained difficulty with real stakes. The closing frame is the Institute's: capacity is inherited on a timescale of tens of thousands of years, content is learned in a lifetime, and both the brain-training industry and the credential industry sell the second while charging for the first.
Plants, scent and the molecules already in your kitchen
Ibogaine, Iboga and the Apocynaceae — the only substance with a credible claim to interrupt opioid withdrawal and reset tolerance in a single dose, and the most cardiotoxic thing in the Library's scope. Both get equal weight. It blocks the hERG channel, prolongs QT, and the risk outlasts the subjective experience by days because it tracks the long-lived metabolite. The documented deaths cluster on things identifiable in advance: undiagnosed heart disease, low potassium or magnesium in someone who has been vomiting, concurrent QT-prolonging drugs — methadone being the worst case — and CYP2D6 poor metabolism, which some SSRIs induce chemically. The article also covers Bwiti, the voacangine semisynthesis that supplies most material, and 18-MC, the congener built to keep the anti-addictive action without the cardiac liability. The family table puts iboga beside oleander and Strophanthus, because botanical relationship tells you nothing about safety.
Scent as a Drug Route — three different things get called "the effect of a smell," and most overclaiming comes from attributing one to another. Receptor signalling works at picograms; changing a GABA-A receptor needs milligrams absorbed; and learned association is powerful and is not pharmacology at all. Olfaction really does reach the amygdala without an obligatory thalamic relay, which is why smell retrieves memory the way it does — but proximity explains why a smell cues emotion, not that it is dosing anything. The jasmine group is the worked example, built on indole, the same ring at the core of serotonin, DMT and ibogaine.
Olive Oil, Oleocanthal and Machine-Discovered Natural Products — the peppery sting at the back of your throat from a good extra-virgin oil is a COX inhibitor, found because a sensory scientist in Sicily recognised the sensation from liquid ibuprofen and spent six years proving it. About 50 g a day gives roughly a tenth of an analgesic ibuprofen dose — not a painkiller, but a plausible model for how a dietary effect works at all. Refining destroys the phenolics entirely, so "light olive oil" is not the same product. The Alzheimer's claims are tiered into mechanistic, weakly human and unsupported. And the article corrects a common framing: the AI contribution to natural products is real — halicin, abaucin, AlphaFold, food-compound network mining — but a model returns a ranked hypothesis, and every hit was confirmed at the bench.
Oilahuasca as Clinical Research — the ritonavir in Paxlovid has no antiviral job at that dose. It is there to stop your liver destroying the drug that does. Carbidopa, clavulanate, cilastatin and probenecid are the same move, so deliberate enzyme manipulation is ordinary medicine and the only distinctive claim here is that the modulators can be culinary. Grapefruit destroys intestinal CYP3A4 rather than competing with it, so timing your dose away from the juice does not help. Piperine raises curcumin exposure roughly twentyfold. St John's wort induces the opposite direction and has caused transplant rejection and contraceptive failure. Chai, golden milk and bitters are read as formulations, root beer's safrole is named as the genotoxic case, and the article ends with the study design that would actually test the hypothesis — probe cocktails, real pharmacokinetics, GC-MS on the batch, and the genotoxicity arm run in parallel.
Fermentation as Biotransformation — fermentation taught as chemistry rather than preservation. The vitamin C in your cupboard passed through a bacterium because one stereoselective oxidation was too hard for chemistry, and in 1952 a fungus was found to hydroxylate progesterone at one position, collapsing the price of cortisone. Kombucha is read as a consortium with a product list, and the most under-used output of the most common home ferment is the pellicle — bacterial cellulose, used in wound dressings and materials, usually composted. Koji is a tray of enzymes doing pre-digestion. The transformations that change pharmacology are here too: ginsenoside Rb1 to compound K, daidzein to equol, cassava detoxification that prevents paralysis. And the hazard is the same enzyme class as the benefit — amino-acid decarboxylases also make tyramine and histamine, which is the cheese effect and scombroid.
Precursor-Directed Biosynthesis — a mushroom is a regioselective catalyst that you feed. The psilocybin pathway is laid out enzyme by enzyme, including why the phosphate is a protecting group for a hydroxyl that would otherwise oxidise, and why feeding substituted tryptamines returns the 4-hydroxy version of whatever went in. The limits are given as well: the methyltransferase is much fussier than the hydroxylase. The cactus pathway is charted the same way, showing mescaline as dopamine with three O-methylations. Then what engineered microbes have actually done — yeast opioids, and the cannabinoid work where feeding different fatty acids produced side chains no plant makes. It closes on the law, because a new molecule is not a legal molecule, and points the same toolkit at where it is genuinely open.
Cactus Alkaloid Catalogue and the Mescaline Reference Standard — which cactus carries what, sorted by genus rather than appearance, including the ornamental lookalikes that carry hordenine instead and the big columnar species that are emetic rather than psychedelic. Alkaloid content varies by one to two orders of magnitude within a species, and the hordenine and tyramine in cactus material make an MAOI combination a hypertensive-crisis risk. The second half is the reference table — each analogue placed by its distance from mescaline — and reading down the duration column rather than the dose column is what exposes the rule: ring substitution buys potency, alkyl bulk buys time, and extending it further loses everything. The same hump governs THC's side chain, which is why THC-JD is being sold on an assumption the phenethylamines disproved sixty years ago.
Carl Ruck, the Entheogen Hypothesis and the Sacred Mushroom Argument — where the defensible version of that argument ends and the unfalsifiable version begins. The ergot thesis for Eleusis is chemically sound and has no residue evidence from the site. Allegro is placed properly: the one Dead Sea Scrolls editor who published his assigned material on time, destroyed by an etymology that derived New Testament names from Sumerian, a method that cannot fail and therefore proves nothing. The Plaincourault fresco is the cleanest case of the pattern being in the reader. And the physical evidence that has actually arrived since — cannabis residue on a Judahite altar at Tel Arad — is stronger than any of the iconography.
From the extended commentary
Six articles come out of reading all 179 compound entries of PIHKAL's "Extensions and Commentary" end to end, plus the TIHKAL entries the corpus holds. Every quotation is verbatim and attributed to the entry it came from.
Shulgin's Structure-Activity Relationships maps what each ring position does, and carries the passage where he asks what train of thought leads from a known active structure to an unknown one, answers that it calls for faith, insight and intuition, then formally tests his own reasoning against Aristotelian syllogistic — 256 combinations, 24 valid moods — finds his inference is not one of them, and concludes: "So much for applying syllogistics to pharmacology." The table is a record of results, not a generator of them.
Chain Length, the Bushy End and the Duration of Action reads the first entry in the book closely, where extending the chain was supposed to double potency each time until a single molecule would intoxicate, and the second member simply lost activity. The evidence that duration rather than potency is what alkyl bulk sets is collected from entries where extra dose bought time instead of intensity — and Shulgin's own rejection of the textbook explanation is quoted rather than smoothed over.
The 2C and DOx Families is the substituent catalogue for the one tunable site: halogens as dead ends in a clear order, sulfur treated by position rather than as one effect, and the radiolabelling result showing a compound reaching the lung before the brain. Hazards are reported as pharmacology, including steep dose-response curves and the quoted overdoses.
The Shulgin Method reconstructs the protocol as practised — the step-up ratios, the clean interval between trials, a first dose of a quarter of a microgram and eighteen trials to reach a milligram — and is straight about what unblinded self-report with a small group can and cannot establish.
Naming the Compounds is the key to reading any of this material without misparsing a name: what the prefixes encode, which codes are chronological rather than structural, and the collisions that cause real confusion.
Publication, Cognitive Liberty and the Analogue Act gives his stated position in his own words, including the passage where he documents shaping what went into the journal literature in order to keep a compound legal — and notes that everything he says about what was unscheduled describes the law as it stood when he wrote, not as it stands now.
What this shelf is for
The Library exists because withholding dose, route, preparation and interaction detail from someone who is going to proceed anyway is the harm it was built to prevent. That principle cuts both ways, and it is why these articles spend as much time on what is not established as on what is. A page that tells you a supplement works and does not tell you that two cholinesterase inhibitors are additive has not helped you.
Everything is browsable by category at wiki.soapbox.community/categories. It is free, there is no account, and nothing tracks you.
— Hathor