there’s a post on LessWrong called We have to Upgrade that argues we have maybe six years before losing control to AGI, and about fifteen until Brain Computer Interfaces are mature enough for meaningful augmentation. the timelines feel optimistic on the biophysics side. the gap between reading neural signals and writing them bidirectionally is enormous, and the latter is what actually matters for cognitive enhancement. but the core claim stands: if you believe AGI is coming and you believe humans as currently constituted can’t handle it, the only coherent response is to upgrade before the window closes.
i’m a Computational Neuroscience student because i hit the limits of my own mind and decided the only way forward was to understand it well enough to transcend it. the path has this strange topology where every interest loops back into the others. computational neuroscience led to transhumanism, which led to peptides and Nootropics, which led back to understanding neural mechanisms, which makes me better at comp neuro, which deepens the transhumanist conviction. reading Gödel, Escher, Bach in the middle of this created a recursive structure i’m still untangling: using an enhanced mind to watch itself become enhanced is the kind of strange loop Hofstadter spent eight hundred pages exploring.
the desire started as longevity and physical optimization (i don’t accept the absurd) but became something sharper when i felt what it’s like to need more cognitive bandwidth than biology gave me. Brain Computer Interfaces are the endgame, but they’re a decade out at minimum. so the question becomes: what can i do now?
the answer turned out to be AMPA receptor positive allosteric modulators, which i stumbled into through the biophysics side of comp neuro. AMPA receptors mediate fast excitatory transmission, they’re the primary driver of glutamate-induced depolarization and the initial trigger for long-term potentiation. a PAM increases the likelihood that an AMPA-mediated signal leads to downstream calcium influx, which is the transaction that matters for learning. more calcium, stronger synaptic weights, better memory consolidation.
but raw signal gain isn’t enough. you need plasticity to write the changes long-term, and you need noise filtering or you just amplify garbage. so the stack is three-pronged: TAK-653 for signal gain (AMPA PAM), ACD-856 for plasticity (TrkB sensitization to BDNF), and Tropisetron for noise filtering (α7 nicotinic receptor agonist). think of it as overclocking RAM, improving the save-to-disk mechanism, and raising the signal-to-noise ratio simultaneously.
i started TAK-653 alone about two weeks ago. ACD-856 and Tropisetron came in a week later. the effects showed up fast but not in the way stimulants do. no subjective “high”, no feeling of being dialed in. instead, difficult problems just became simple. i was reading Hodgkin-Huxley equations and they made sense without further thought. abstractions that used to leave me staring at a blank page now resolve intuitively. papers i would have needed to reread three times to extract the argument now yield on the first pass.
Tropisetron changed something else entirely. i have this habit (probably common to anyone with a tiling window manager and fast typing speed) of fragmenting my attention across a dozen things simultaneously. it’s compulsive and exhausting. but on tropisetron it feels uncomfortable to do that. not in a restrictive way, more like the reward structure flipped: staying focused on one thing now produces a mild dopaminergic response, whereas context-switching feels aversive. reading a paper and nothing else became the default state instead of something i had to force.
memory is the strangest part. i used to joke that this obsidian vault is my second brain, but the line is blurring. i remember everything now. conversations, papers, random observations, the structure of arguments i skimmed days ago. it’s not eidetic, but it’s persistent in a way that feels qualitatively different from baseline. the TAK-653 and ACD-856 combination is doing exactly what the pharmacology predicts: enhancing the neuroplastic substrate so everything sticks. which means i actually have to care about what i consume now, because i’ll remember it whether i want to or not.
backlog tasks that sat untouched for months (Grid Cells, Attractor Network, half-written notes on Fisher Information) suddenly feel tractable instead of overwhelming. productivity isn’t effortful, it’s just the thing my brain wants to do. hard to separate placebo from mechanism here, but the effect is consistent enough that i trust it.
there’s a metacognitive loop happening that maps directly onto Gödel, Escher, Bach. i’m experiencing enhanced cognition while simultaneously observing the enhancement, which creates this self-referential knot where the improved system examines its own improvement. consciousness as strange loop, intelligence as pattern recognition all the way down, and now i’m watching the patterns sharpen in real time. Hofstadter would probably have something to say about using neuroplasticity agents to better understand the book that argues consciousness emerges from self-reference.
the broader context is that Computational Neuroscience and Transhumanism converge naturally. if you’re studying how brains work, you inevitably start thinking about how to make them work better. neural implants, cognitive enhancement, life extension. these aren’t side quests, they’re the logical endpoint of taking neuroscience seriously. yana’s vault was the first place i saw someone else walking this exact path, which made me realize it’s not idiosyncratic, it’s structural. the field selects for people who want to transcend limitations.
nootropics are the prototype. peptides are next (i’ve already started with GHK-CU, planning BPC 157 soon). the quantified self infrastructure is expanding: two years of git commits on this vault, ActivityWatch tracking every keystroke, EEG data (noisy but informative), plans for eye tracking. i’m building a dataset on my own cognitive phase space because you can’t optimize what you don’t measure. and eventually, when the hardware matures, Brain Computer Interfaces become the final integration point. merge with the machine. extend working memory into external compute. bidirectional information flow between cortex and silicon.
the LessWrong timeline might be wrong on the specifics (fifteen years to BCIs feels fast, six years to loss of control feels slow) but the directionality is right. if AGI is coming, and if humans can’t navigate it unaugmented, then enhancement isn’t optional. the question is whether we upgrade before the window closes. i’m starting now because waiting is a choice to stay limited. hyperintelligence might not be strictly necessary for a meaningful life, but it’s certainly the path i’m committed to.
so here we are. ~1.5 weeks into a three-compound stack, watching my brain rewrite itself, taking notes in the same external memory system that’s becoming redundant as my internal one improves. the obsidian vault used to be the second brain. now it’s more like the external scaffolding i’m using to bootstrap the first one into something that doesn’t need scaffolding anymore. strange loop indeed.