Screenclutter

@DylanRoss I just wanna know his Anduril stake.

take it from a member of the 2020 Lakers championship team, we're in a bubble

@badlogicgames @libpol_org How is it mean to note that a company can"t say or mean things above a certain headcount?

@screenclutter @libpol_org no need to get mean.

recommended reading. i too am very done with people anthropomorphizing a bunch of matrices on a GPU cluster, especially if the same people do not give two fucks about actual human beings.

@libpol_org @badlogicgames do you actually interact with the company's products and announced contracts in a meaningful way, or just follow corporate messaging? Because the only thing worse than anthropomorphizing an LLM is anthropomorphizing a pre--IPO business with shareholders.

I feel this is a deep misinterpretation of what Anthropic says and believes. They have an alternative plan they're offering (essentially something akin to socialism). IMO Anthropic is way, way, way more authentic, honest, transparent, responsible, and humanistic on this front than OpenAI, who are moving more towards dishonestly covering up what will happen to humans. I think Altman knows humans will be replaced but is currently lying because it's not good PR to say that. Dario is saying what he genuinely believes and is saying it because people deserve to hear the warning and not have it covered up.

@anselm_io @mitsuhiko No, you're interacting with the output of something that resembles a human's output. there's no lived experience, cultural affinities or anything else experienced by human senses or a human mind. LLMs are trained on human outputs, not human capabilities.

@mitsuhiko The reason to not say clanker is the same why you say please to the agent: you interact with something that speaks like a person, because it is made from lived human experience. It does’t matter if it has a soul or if it ever will - you erode your soul by speaking to it rudely.

More musings after some people got upset about the word clanker. lucumr.pocoo.org/2026/5/26/clan…

@katienotopoulos Tons of companies used things like Notion which had questionable workspace to public documentation layers. I would assume it wasn't the CMS but the project-planning layer with an oopsie on the share settings that got hit by a scraper.

Genuinely confused how this would happen... some sort of staging site that's different than the regular Time site that was accessible? I can't imagine the Time CMS has drafts like.... publicly visible if not logged in with credentials? nytimes.com/2026/05/26/mag…

@tszzl AI doesn't have this because labs have to advertise opportunities because they suck at judging industry fit. but there isn't a place in history for the smartest LLM minds, but the ones that link it to the most societal-changing applications. See, e.g. Nikola Tesla.

@tszzl This relies on an unclear antecedent: previous technological outcomes are known because they supplied a clear implementation: oil in gasoline for cars, but also plastics and so on. A tech advancement doesn't enter history books until its clear downline effects are traced.

one interesting thing that shows up re: the definition of a true american is the idea that one's ancestors fought and bled for the movement of History in the country. in the fukuyamist sense, there is actually just no longer an opportunity to do this. history has ended "The end of history will be a very sad time. The struggle for recognition, the willingness to risk one's life for a purely abstract goal, the worldwide ideological struggle that called forth daring, courage, imagination, and idealism, will be replaced by economic calculation, the endless solving of technical problems, environmental concerns, and the satisfaction of sophisticated consumer demands." you can see what people are reaching towards--and at this stage there is no need to judge its moral value--when they say one or more of their ancestors played a role in the act of 'history' of real ideological battle and warfare. there is nothing left to die for, not really, and proxy conflicts like iraq or what have you just don't count as ideological struggle and don't hold a kind of mythological weight. people today, hungry for a kind of recognition that is impossible, can at least reach for the idea that one's ancestors partook in megalothymia and created History. for this instinct, it probably doesn't even matter which side of the civil war one's grandparents were on, at least it was "real" as fukuyama writes, the only thing that could possibly restart history now is the progress of technology, the last unpredictable adventure left to us, that alters the nature of man and makes new things under the sun possible. make of that what you will

@Jordan_Bianchi putting memorials behind paywalls is a clown move.

Gray skies, grief hover over NASCAR race weekend in wake of Kyle Busch’s stunning death. nytimes.com/athletic/73034…

He thinks he's @NickKristof and able to parachute into anythíng but it just isn't the same thing.

Is there a @GaryMarcus for sane people? Like an active observer of the AI space as a whole, rather than a use-case contrarian, who actually interacts with the space but doesn't require cable news hits because he no longer does research and is irrelevant otherwise?

@LLMJunky is that I think Elon"s industry leader picks are subject to a higher variance. He might hit 4th on frontier AI, first on private space, 2nd on EVs and strike out on space compute. still amazing as a track record but doesn't hit IP valuation.

@LLMJunky I wanted to be careful on this: picking three or four industries will set up a person for life. picking the methodology for dominating thar category is like 20 people in any area. Elon is already 1 in 20. definite special mind and he crosses lndustries. What I wanted to suggest

SpaceX’s IPO filing is wild. The company is split into 3 segments: - Space - Connectivity - AI In Q1: - Space lost $662M - Connectivity, mostly Starlink, made $1.1B - AI lost $2.5B So the only profitable business is Starlink. But the valuation story is AI. SpaceX claims a $28.5T addressable market. 93% of that is AI. Not rockets or satellite internet. AI. And over 60% of Q1 capex went to AI infrastructure. The unusual part is that the AI business is not exactly dominant yet. A major near-term AI revenue source is a $1.25B/month compute deal with Anthropic, one of the leading AI companies, but that deal can be cancelled in fewer than 3 months. They are also preparing to drop $60 Bs on Cursor. So the AI pitch is partly: - rent GPUs to Anthropic - improve Grok - Cursor Then there’s governance. Public investors get Class A shares: 1 vote each. Musk gets Class B shares: 10 votes each. and there is also non-voting Class C stock, which can be used for future acquisitions without diluting control. The filing says Musk will control the company. It also says some of his restricted Class B shares can be voted before they vest, even though the vesting goals include a $7.5T market cap and a permanent Mars colony with 1M people. This all absolutely depends on the success of Starship (congrats on a successful launch today!) Starship is needed for bigger Starlink satellites, cheaper launches, orbital data centers, Moon/Mars transport, and basically the entire long-term story. So this is not really “SpaceX the rocket company going public.” It is a public bet on Starlink funding the present, Starship unlocking the future, AI justifying the valuation, and Elon keeping control. The upside is enormous. But so is the speculation. We are more or less buying the belief that all of Musk’s hardest promises eventually converge.

@SIGKITTEN a flake would help and I can contribute a PR on how to maintain functionality. but your saving grace is that your config never messes up. keeping a commented template avoids rust analysis. it either works or doesn't based on the toml

@SIGKITTEN I already have and you pointed out how to fix it in the config! One big one would be keeping a config template clankers have to adjust on feature add/drop. I have shell access to a Nix-store addressed bash, but only because I pointed my agent at config.rs.

I won't tag him in the call-out of a popular account whose app is basic electron slop for anti-openclaw snobs. but if you want mobile access to your agents, there isn't anythíng better than litter from @SIGKITTEN and it isn't even close.

@SIGKITTEN So when I say Litter and your system are good for mobile, it's operating from a fairly hostile stack.

@SIGKITTEN To be more clear with how well your stuff works on a niche setup: 1) I run NixOS 2) I install agents locally with Bun 3) I check them on a stock pixel any time you run an update, I can't rerun it because your user service doesn't apply to nix. I just fetch, copy to .local/bin

@SIGKITTEN no, that's the point. You establshed the contract: can do wild things, need users to test edge cases because you're one guy. I have used your stuff since launch. I just want devs to be honest when they post. This was a compliment.

@screenclutter shit sorry i feel bad. way too many things now with all the clankers, trying to stabilize the testflight and android and do a good release

@SIGKITTEN I might throw a clanker at it." True. calling something with no mobile surface "the best remote control" is a wider slap against the faces of many people who have made solid mobile from Happy/happier to Paseo to larger players in the space.

@SIGKITTEN I spent an entire day figuring out app-server listening on unix fucking up local tool calls on a codex/litter update. I don't say this because I think what you do is perfect, but because the claims are more honest, "This should work across every surface but we're trying shit and