2026 presents unfathomable opportunities for taking positive action to help the future go well. I see very few people oriented to what I see as many of the foundational basics.
1) With GPT-5.2-Codex-xhigh, Claude Code Opus 4.5, GPT-5.2-Pro, and the increasing DX tooling being created every hour, building wildly ambitious software in absurd time is exceptionally feasible. I am talking 1-6 devs building software platforms in weeks for agent-economies factorizing cognitive work at scale, with ultra-parsimous, easy-to-build-upon Rust codebases. A 2024-cracked-SWE-year's worth of work per day is basically possible already for thousands of people, if they can manage to be healthy enough and find flow.
2) Causing great outcomes basically has to be an additive, personal, bits-on-the-ground thing. You have to find yourself giving brilliant people influence over huge agent swarms, that can just deeply research, understand, and counter certain problem domains, like prion terrorism risk or whatever (because a significant fraction of people crash out sometimes). Real world collective superintelligence, this year, and you are a key part of it.
Trying to get real human builders to be all that oriented to downside risks, or jockeying for unprecedented government lockdown of compute, does not seem to be nearly as existentially practical as just building with good primitives and conscientiously stewarding emergence.
3) The real game is actually more somatics than anything. That doesn't mean full hippie kumbaya all the time, but you probably really do need to move a lot throughout the day, get Bryan Johnson-level sleep, distance from people you really contract around, get some hugs or self-hugs at least, and be gentle with your parts, like your socially-penalized instincts.
Backpropagating from things like millions of healthily-trained agents naturally coordinating cognitive work at scale with simple primitives--pairwise ratio comparisons eliciting cardinal latent scores for arbitrary attribute_prompts over arbitrary entities cached in agent-explorable SQL; multi-objective rerankers; market-pricing; frameworks that agents love for prior elicitation, reconciliation, and propagation--
it's about sitting in front of a laptop and prompting well, a lot. And to do that well, you've got to build up uncommonly good taste for what body signals are safe to ignore and what should really be respected. For me, estrogen (and some exogenous T) and a period of shrooms was absolutely necessary to downregulate, unclench, feel enough, and unlearn unhelpful reactivity patterns, to end up regulated enough to have a minimum viable amount of agency--I was simply too cripplingly heady and stuck before).
The world is a trippy, wildly complicated place as a self-modifying embodied organism, and I think that's to be embraced. The first person-perspective and unflattening of experience is a very real part of the game.
4) Overall, there's a resource flywheel game. Gotta find yourself with more and more dakka. I need more felt-safety, more nutrients, more stewarding of psychic gradients so good prompting and creativity keeps happening more, and I manage to bootstrap my limited $600/month AI compute + Hetzner server budget into $20,000/month into possibly $1M+ a day, by the time agents and harnesses get that good in 2026, to keep doing research and spawning side businesses that solve real problems in the ecology.
2026 presents unfathomable opportunities for taking positive action to help the future go well. I see very few people oriented to what I see as many of the foundational basics.
1) With GPT-5.2-Codex-xhigh, Claude Code Opus 4.5, GPT-5.2-Pro, and the increasing DX tooling being created every hour, building wildly ambitious software in absurd time is exceptionally feasible. I am talking 1-6 devs building software platforms in weeks for agent-economies factorizing cognitive work at scale, with ultra-parsimous, easy-to-build-upon Rust codebases. A 2024-cracked-SWE-year's worth of work per day is basically possible already for thousands of people, if they can manage to be healthy enough and find flow.
2) Causing great outcomes basically has to be an additive, personal, bits-on-the-ground thing. You have to find yourself giving brilliant people influence over huge agent swarms, that can just deeply research, understand, and counter certain problem domains, like prion terrorism risk or whatever (because a significant fraction of people crash out sometimes). Real world collective superintelligence, this year, and you are a key part of it.
Trying to get real human builders to be all that oriented to downside risks, or jockeying for unprecedented government lockdown of compute, does not seem to be nearly as existentially practical as just building with good primitives and conscientiously stewarding emergence.
3) The real game is actually more somatics than anything. That doesn't mean full hippie kumbaya all the time, but you probably really do need to move a lot throughout the day, get Bryan Johnson-level sleep, distance from people you really contract around, get some hugs or self-hugs at least, and be gentle with your parts, like your socially-penalized instincts.
Backpropagating from things like millions of healthily-trained agents naturally coordinating cognitive work at scale with simple primitives--pairwise ratio comparisons eliciting cardinal latent scores for arbitrary attribute_prompts over arbitrary entities cached in agent-explorable SQL; multi-objective rerankers; market-pricing; frameworks that agents love for prior elicitation, reconciliation, and propagation--
it's about sitting in front of a laptop and prompting well, a lot. And to do that well, you've got to build up uncommonly good taste for what body signals are safe to ignore and what should really be respected. For me, estrogen (and some exogenous T) and a period of shrooms was absolutely necessary to downregulate, unclench, feel enough, and unlearn unhelpful reactivity patterns, to end up regulated enough to have a minimum viable amount of agency--I was simply too cripplingly heady and stuck before).
The world is a trippy, wildly complicated place as a self-modifying embodied organism, and I think that's to be embraced. The first person-perspective and unflattening of experience is a very real part of the game.
4) Overall, there's a resource flywheel game. Gotta find yourself with more and more dakka. I need more felt-safety, more nutrients, more stewarding of psychic gradients so good prompting and creativity keeps happening more, and I manage to bootstrap my limited $600/month AI compute + Hetzner server budget into $20,000/month into possibly $1M+ a day, by the time agents and harnesses get that good in 2026, to keep doing research and spawning side businesses that solve real problems in the ecology.