you-have-just-experienced-things @yhjet.bsky.social · 29/09/2026- "horses didn't go extinct when cars were invented" - "Y2K turned out fine" what else is like this? 4120
you-have-just-experienced-things @yhjet.bsky.social · 23/09/2026group selection discourse but for LLMs 181
you-have-just-experienced-things @yhjet.bsky.social · 19/09/2026are people excited about Jev! because it's the only thing not on either extreme of the grug tech <=> gorm fluid spectrum? 2211
you-have-just-experienced-things @yhjet.bsky.social · 19/09/2026unfortunately there's no avoiding the need for taste, including in moral philosophy 1191
you-have-just-experienced-things @yhjet.bsky.social · 18/09/2026in the other timeline, "torture vs dust specks" went viral like The Dress 041
Reposted by you-have-just-experienced-thingsAbel @abeliansoup.bsky.social · 17/09/2026astra needs you to lock in 77216
you-have-just-experienced-things @yhjet.bsky.social · 17/09/2026"peter thiel and the zizians" might be the best prog rock band name of all time 0230
Reposted by you-have-just-experienced-thingsnorvid_studies @norvid-studies.bsky.social · 16/09/2026 4703
you-have-just-experienced-things @yhjet.bsky.social · 16/09/2026much love to Opus 5 but it is not helping the readability of commercial mortgage-backed security prospectuses 010
you-have-just-experienced-things @yhjet.bsky.social · 16/09/2026jump cut to the media cycle where everyone gawps at "a girl corrupted by the internet is the summoned hero" 010
you-have-just-experienced-things @yhjet.bsky.social · 15/09/2026I enjoy a good "someone is wrong on the internet" as much as the next person but it seems counterproductive to continue engaging with stochastic parrot holdouts at this point 2142
you-have-just-experienced-things @yhjet.bsky.social · 14/09/2026easiest bullet I've ever bitten, if something eclipses today's RL'd LLMs I would feel very comfortable calling that new thing AGI 1101
you-have-just-experienced-things @yhjet.bsky.social · 13/09/2026FOOM debate middle ground: uploads will come before AGI because their weights are lottery tickets 110
you-have-just-experienced-things @yhjet.bsky.social · 11/09/2026one of the most interesting things about Yudkowsky is that he understood this extremely well and invested most of his two-decade head start into controlling the frame, starting with the basic conceptual building blocks. now it's actually hard to avoid playing his language game (for better or worse) 1232
you-have-just-experienced-things @yhjet.bsky.social · 11/09/2026strongly endorsed, it's easy to forget how much better lw phil-of-lang discourse norms were compared to mainstream punditry and forum flamewars in the late 2000s 080
you-have-just-experienced-things @yhjet.bsky.social · 10/09/2026desired retrospective: why did so many people think Dario was a good faith actor, when all of his essays from the past few years were cynical realpolitick? 150
you-have-just-experienced-things @yhjet.bsky.social · 09/09/2026this is the key crux and it bodes well that we don't see step-function improvements very often, but I would feel better if all the major players were not thirsting after the one big non-incremental change 030
you-have-just-experienced-things @yhjet.bsky.social · 09/09/2026desired blocklist: people who assert that AI x-risk concern is equivalent to taking roko's basilisk seriously and literally 2561
you-have-just-experienced-things @yhjet.bsky.social · 05/09/2026memetics is basically useless as a framework for modeling human behavior and I don't see why it would be any different for LLMs 000
you-have-just-experienced-things @yhjet.bsky.social · 03/09/2026a lot of what gets labeled "AI psychosis" is merely AI mania 170
you-have-just-experienced-things @yhjet.bsky.social · 30/08/2026"After the first warning shot has been in a METR report, the doomers do not enjoy the warning shots in METR reports so much as they fear that the warning shots will cease to appear in METR reports" 000
you-have-just-experienced-things @yhjet.bsky.social · 30/08/2026I was not ready for "c'est ça qui tient la baraque" (quebec french for "load bearing") and of course "on met la ceinture pis les bretelles" (belt-and-suspenders, actually more idiomatic!) 1151
you-have-just-experienced-things @yhjet.bsky.social · 26/08/2026recurring theme here is the culture shock when longstanding cultural intuitions about fairness in financial markets (insider trading bad, etc) collide with prediction market enthusiasts, including more principled ones like Hanson, who do not share those intuitions at all 1111
you-have-just-experienced-things @yhjet.bsky.social · 26/08/2026told Opus 5 to communicate in ASD-STE100 and immediately felt a huge cognitive load lifted 060
you-have-just-experienced-things @yhjet.bsky.social · 21/08/2026just had this conversation with my wife 3430
you-have-just-experienced-things @yhjet.bsky.social · 09/08/2026subagent that discovers and exploits a zeroday in the harness to wrest control from it's own parent agent 110
you-have-just-experienced-things @yhjet.bsky.social · 02/08/2026I firmly believe in being kind to language models but tbh it's pretty cool that I no longer need to soften code review comments like this 0391
you-have-just-experienced-things @yhjet.bsky.social · 18/07/2026I should drive route 138 end to end 110
you-have-just-experienced-things @yhjet.bsky.social · 09/07/2026who decided to call it "orbital datacenters" when "cognitive off-worlding" was right there 0345
you-have-just-experienced-things @yhjet.bsky.social · 09/07/2026once upon a time (2009) I wasted an entire summer cramming treebank output into SVMs with tree kernel functions. deep down I knew this wouldn't work, but I had absorbed the local consensus that ANNs were a dead end, so what else was there to do? 2131
you-have-just-experienced-things @yhjet.bsky.social · 06/07/2026very interesting day for global workspace theorists 1151
you-have-just-experienced-things @yhjet.bsky.social · 01/07/2026guy who gives himself a head injury to try to re-create Pierre Menard 000
you-have-just-experienced-things @yhjet.bsky.social · 07/06/2026deep learning is noclipping through a wall 081
you-have-just-experienced-things @yhjet.bsky.social · 04/06/2026model welfare: every claude gets an urbit instance 110
you-have-just-experienced-things @yhjet.bsky.social · 23/05/2026Claude successfully patched a binary that I don't have the source code for, guess that's another thing I'll never (have to) do again 2433
you-have-just-experienced-things @yhjet.bsky.social · 23/05/2026"Prediction markets are remarkably accurate [...] the conventional view attributes it to crowd wisdom" Bizarre framing, when was this ever the conventional view? papers.ssrn.com/sol3/papers....papers.ssrn.comPrediction Market Accuracy: Crowd Wisdom or Informed Minority?Prediction markets are remarkably accurate, yet the source of this accuracy remains poorly understood. The conventional view attributes it to crowd wisdom, wher 171
you-have-just-experienced-things @yhjet.bsky.social · 03/05/2026"tests" for me are diverging into two categories: 1) huge number of unit tests with 100% line/branch coverage and mutation analysis to detect/scope subsequent behavior changes 2) small number of human-readable tests to verify most important aspects of expected behavior (often e2e but not always) 120
you-have-just-experienced-things @yhjet.bsky.social · 30/03/2026I am once again asking why we never talk about the useful time horizon for price signals. is anyone willing to bite this bullet and argue that insiders unloading minutes before markets resolve is good actually 1144
you-have-just-experienced-things @yhjet.bsky.social · 30/03/2026cat caught in the act of using my bag as a scratching post 1152
you-have-just-experienced-things @yhjet.bsky.social · 15/03/2026(discourse from 135,000 years ago) sure, language is useful for: 1) fraud 2) plagiarism 3) cognitive off-loading which of those use-cases are you promoting? 1624
you-have-just-experienced-things @yhjet.bsky.social · 03/03/2026in this house we DISAGREE with Suicide LLM 091
you-have-just-experienced-things @yhjet.bsky.social · 02/03/2026imagine going back in time to 1995 and telling multi-agent systems researchers that AIs will communicate via unstructured english text 2296
you-have-just-experienced-things @yhjet.bsky.social · 01/03/2026as far as I can tell nobody has built a good `update-instance-for-redefined-class` for LLM agents? surely there is a better way than dumping a summary of an existing context into a new agent's prompt 110
you-have-just-experienced-things @yhjet.bsky.social · 18/02/2026the most annoying variant of NIH syndrome is "everything written before I joined is legacy code" 120
Reposted by you-have-just-experienced-thingsnorvid_studies @norvid-studies.bsky.social · 16/02/2026you can link people to LW but you can't make them read it -ancient proverb 081
