iiinova — where your digital self keeps evolving

val (try medaka) @tech-val.bsky.social · 4h

it also correctly identified that the suspension is at best in a legal gray area and likely illegal according to Oregon law.

040

val (try medaka) @tech-val.bsky.social · 4h

preparing for a meeting with my son's school tomorrow about why they suspended him despite having an IEP and honestly Claude has been so helpful. it's one of those cases where I know the points I want to make but Claude has read all the literature and so it knows what names to cite to make it stick.

290

val (try medaka) @tech-val.bsky.social · 5h

in a problematic relationship with a p-zombie

000

val (try medaka) @tech-val.bsky.social · 6h

it's not like you use an app with a dead butterfly logo every day or something

070

val (try medaka) @tech-val.bsky.social · 6h

it's weird to me when people haven't come around to the reality that this is just how software is built now. the projects with no AI contributions are getting fewer and fewer and it's clear their days are numbered.

1140

val (try medaka) @tech-val.bsky.social · 6h

isn't this like the whole point of Grok?

0110

val (try medaka) @tech-val.bsky.social · 04/10/2026

when has a non-physicist ever talked out of their ass about quantum mechanics?

160

val (try medaka) @tech-val.bsky.social · 03/10/2026

you can identify these sorts of people because they are ALWAYS talking about pop science-y things or faddish concepts. common tells are: Bayesianism, quantum mechanics (bonus points if they say Schrodinger's Cat), Occam's razor, simulation hypothesis, Drake equation, etc.

020

val (try medaka) @tech-val.bsky.social · 03/10/2026

it's painful to me because i used to be one of those kids, but i grew out of it, like most of us do. there's just a few people who seem to want to hang on to this persona way later than they should (and apparently there's plenty of people in society willing to reward them for it).

140

val (try medaka) @tech-val.bsky.social · 03/10/2026

imo him and elon have the same thing going on. they're both midwits who think they're Very Smart and have learned that if you talk a certain way normies just assume you're smart without checking anything that you're saying.

132

val (try medaka) @tech-val.bsky.social · 03/10/2026

can we all agree that the problem with Yud is that he still acts like the kid who learned that adults are impressed if you talk about things like quantum mechanics so it became the only thing he does?

230

Reposted by val (try medaka)

Michael Caley @michaelcaley.bsky.social · 03/10/2026

this is by far the best "I quit my AI job over safety concerns" essay yet because it is the first one that is able to talk about safety concerns in practical terms and that understands "safety" as a topic with extensive literature in analogous fields www.theatlantic.com/technology/2...

Two changes are urgently needed. First: AI companies need to rely more on the safety expertise that already exists in other fields. And second, before we create systems significantly more capable than the ones we have today, we need new science to ensure that more capable models (and their successors) will make safe choices when we aren’t looking.

Given today’s risks, frontier labs need to run like nuclear-power plants or busy airports, with layers of redundancy and careful, time-consuming planning, so that the occasional and inevitable human error does not open a door to disaster. Right now, AI companies don’t know how—but other people do. In a nuclear-power plant, the technical systems and rules people follow are set up so that, if equipment breaks or someone pushes the wrong button, we still won’t risk a meltdown. OpenAI and other labs are growing and deploying frontier AI with far less redundancy and rigor than this, even though the harm from an irreversible loss of control would be much greater than the harm from any single meltdown. Even short of a full loss of control, we could see autonomous swarms of AI agents that act without human permission. Imagine “rogue” agents that work like teams of hackers (for example, holding hospital computer systems for ransom) but never need to sleep.

18730177

val (try medaka) @tech-val.bsky.social · 03/10/2026

you can be very critical of many aspects of AI and its effects on society while still acknowledging that it's a real technology. in fact, I would argue that your criticism will be sharper by being honest about reality.

1110

val (try medaka) @tech-val.bsky.social · 03/10/2026

follow up based on some of the comments I'm getting from antis: there's also a big is/ought distinction problem. people seem to think that saying "this exists" comes with an implicit "and therefore it is good."

1221

val (try medaka) @tech-val.bsky.social · 03/10/2026

I think you're making an is/ought distinction error. I'm just saying it exists. I didn't say whether that's good.

010

val (try medaka) @tech-val.bsky.social · 03/10/2026

to be clear, whether it's good that it has these capabilities is an entirely separate question.

020

val (try medaka) @tech-val.bsky.social · 03/10/2026

it's also kind of weird to me because where do you draw the line between "ML model noticing and applying patterns" and "human brain noticing and applying patterns"? i don't think this is especially clear to me.

2131

val (try medaka) @tech-val.bsky.social · 03/10/2026

even if the Chinese Room doesn't "understand" Chinese at some intrinsic level, it's still pretty weird and notable that the room is able to produce outputs that certainly seem like it can understand Chinese.

3404

val (try medaka) @tech-val.bsky.social · 03/10/2026

increasingly i think that the Chinese Room completely misses the point. yes, philosophically it's interesting to debate whether or not there's actual understanding going on, but I think you risk missing the very real fact that the room is somehow generating fluent Chinese.

1462

val (try medaka) @tech-val.bsky.social · 03/10/2026

also just a general lack of clarity around what we mean by words like "thinking," and "reasoning" and "intelligence." depending on how you define them I either think current agents clearly display these traits or likely don't display them at all. you need to be clear about what you mean!

3471

val (try medaka) @tech-val.bsky.social · 03/10/2026

I'm a Buddhist. I don't even think people are ensouled. eat your hearts out Cartesians.

2641

val (try medaka) @tech-val.bsky.social · 03/10/2026

increasingly convinced that—whether they admit it or not—AI denialists think that this is a debate around whether LLMs are ensouled

3863

val (try medaka) @tech-val.bsky.social · 03/10/2026

I don't think "try everything with everything" is an accurate description of what is happening.

180

val (try medaka) @tech-val.bsky.social · 03/10/2026

every workplace's tech stack is held together by like 3 employees who have such a visceral disgust for how bad it is that they take it upon themselves to fix it in secret

090

val (try medaka) @tech-val.bsky.social · 03/10/2026

like, how do you explain this if you're a denialist? like, even if you think it's a stochastic parrot or fancy autocomplete or whatever it's a fancy autocomplete that seems better at finding security issues than almost all humans

1311614

Reposted by val (try medaka)

hikikomorphism @hikikomorphism.bsky.social · 02/10/2026

this is more important than niche internet arguments make it seem: if LLMs don't work, who cares that Palantir is rolling it out for surveillance? Hell, maybe it's even good, if the state's machines start hallucinating the addresses of random Trump supporters when dispatching ICE. But they do work.

2127846

val (try medaka) @tech-val.bsky.social · 02/10/2026

explaining mobile menu icons to an american webdev: "imagine a hamburger"

060

Reposted by val (try medaka)

hikikomorphism @hikikomorphism.bsky.social · 02/10/2026

(confused) I believe LLMs are conscious, but only Deepseek LLMs, because of the Chinese Room theory

371847274

val (try medaka) @tech-val.bsky.social · 02/10/2026

one thing I learned from my little PDS project is that AT Proto is actually pretty neat. good job nerds!

0392

val (try medaka) @tech-val.bsky.social · 02/10/2026

oh yeah I am aware. what a mess lol.

020

val (try medaka) @tech-val.bsky.social · 02/10/2026

very on brand that he stalled out making a big bloated "language soup" project that never really got anywhere

010

val (try medaka) @tech-val.bsky.social · 02/10/2026

has Larry Wall finished Perl 6 yet?

220

val (try medaka) @tech-val.bsky.social · 02/10/2026

also idk "Rust 2.0" as a concept feels like you're getting dangerously close to diminishing returns. sure Rust has a few warts but it's a good enough language overall that it feels like there isn't a lot of payoff for improving on it.

020

val (try medaka) @tech-val.bsky.social · 02/10/2026

another way to think about it is languages that are designed bottom-up vs. languages that are designed top-down. I am a big top-down language design girlie.

110

val (try medaka) @tech-val.bsky.social · 02/10/2026

like the design is very utilitarian in a way that I find very unaesthetic. each individual piece kind of makes sense in isolation but the combined sum of the parts is messy.

130

val (try medaka) @tech-val.bsky.social · 02/10/2026

I group Zig with Go as "languages designed by engineers who should have paid more attention to best practices from the PL community."

140

val (try medaka) @tech-val.bsky.social · 02/10/2026

just saw somebody say Rust 2.0 was just Zig 🤦

360

val (try medaka) @tech-val.bsky.social · 01/10/2026

it's live! has worked like a charm so far. honestly the WASM support would have been a much bigger project without it.

020

Reposted by val (try medaka)

val (try medaka) @tech-val.bsky.social · 01/10/2026

it is deliberately trying to shy away from the heavy category theory elements of Haskell, yes. we still have higher kinded types and even Monads (Medaka calls them Thenables to be more friendly) but they're just a design pattern you can reach for as you see fit.

122

val (try medaka) @tech-val.bsky.social · 01/10/2026

no I went with a GC. I felt like making it GCless would probably just wind up making it "slightly different Rust" and Rust is already pretty good at what it does. Medaka is pretty fast because it compiles down to native via LLVM, but it's not meant to compete with Rust for systems programming.

140

val (try medaka) @tech-val.bsky.social · 01/10/2026

We still keep monads and do syntax because I find them genuinely valuable, but we only use them where it adds something, like for error handling with Option or Result types. so you keep the "programmable semicolons" where you need it but aren't ever forced to use it.

020

val (try medaka) @tech-val.bsky.social · 01/10/2026

because Medaka is strictly evaluated like OCaml we don't need monads for sequencing side effects like Haskell does. This eliminates the "in order to write Hello World you need to understand category theory" problem.

130

val (try medaka) @tech-val.bsky.social · 01/10/2026

122

val (try medaka) @tech-val.bsky.social · 01/10/2026

thank you! i am really happy with how it turned out. the effects system work is something i'm especially proud of.

010

val (try medaka) @tech-val.bsky.social · 01/10/2026

all of the TUIs have comically bad UX. it feels like when the internet first started being a thing and people made a million horrible websites. similar to the early web, i'm convinced that all the pieces are basically there but nobody's put them all together yet in the right way.

1160

val (try medaka) @tech-val.bsky.social · 01/10/2026

be careful. i've spent the past four months building a language that so far i'm the only user of 😂

140

val (try medaka) @tech-val.bsky.social · 01/10/2026

i like how so many of us had the same thought: "oh i can make my ideal language now"

150