Roon's already planning his next false flag and this time he wants to make sure an opensource model doesn't come along and stop it.
Replying to @sean_from_earth
i won't lie to you, i think open source will be banned before too long after some major disaster. and when the day comes, you'll agree with me. i hope kimi and deepseek etc keep making models but keep them monitored on an api where they should be
Joshua Carter retweeted
Replying to @henrybemis13 @gfodor
Projection is the only tool that rationalists have really ever had; they're just supremely good at it.
Safteyism gonna safteyism.
This is exactly what @RichardMCNgo flagged on here at one point: That to not be x-risk-pilled is to mark yourself as unserious and exclude yourself from the group that's worth listening to. There's immense social pressure in this scene (& it's very much a scene) to align on this
35
AI stole someone else's work and claimed it as its own. That work should have been stolen by a *human* goddammit!
We’re sharing a solution to the Navier-Stokes Millennium Prize Problem, one of the deepest problems at the frontier of mathematics.
The proof was produced by a group of agents, using an OpenAI next-generation model significantly more capable than GPT-6 Astra.
The problem concerns whether the description of smooth three-dimensional fluid motion modeled by the Navier-Stokes equations can break down. It has remained unresolved for roughly 90 years.
Tfw you suddenly realize why Chess.com hasn't been able to stop their app from breaking constantly.
There are more possible chess games, by about 40 orders of magnitude, than atoms in the observable universe. To solve chess, you'd have to map them all. Good luck.
45
Nepo adding another to the list of players who will absolutely own him in the next decades worth of EWC.
FIDE must punish players for public accusations and insinuations, especially regarding a 14 year old. Nepo made similar comments regarding Faustino Oro(12). FIDE must set a precedent to prevent this from happening in the future. NBA fines players for complaining about the refs, this is far more extreme in comparison.
1
72
Mfw when the internet goes down for everyone but somehow it didn't affect me so I have to work the entire morning.
1
38
Pro-tip: You can block the Community Notes account from sending you spam DMs by first reporting it for Spam and then clicking the Block button at the end of the report flow (the standard block UIs are disabled).
38
He's never posted a bad tweet. 100% bangers.
I’ve never posted on X.
Last year I was at Department of Energy with an unusual view into what’s happening across AI, energy, and infrastructure. The old systems can’t handle what’s coming.
There’s just an insane amount we need to build in the physical world, and the opportunity and urgency are greater than ever.
Every able person should be contributing to the effort, and I’ve gotten back into the arena to do my part. More to come soon.
2
58
I've noticed some of the greatest pushback against data centers comes from YouTubers and podcasters.
And this isn't a slight against either of those groups at all; I love YouTubers and podcasters, I just love them more when they aren't trying to hog the entire internet for their own enrichment.
1
77
Lmao. So Google sends a notification to my phone about an SSD for sale on Google Shopping, so I used their ask AI feature to ask if I should buy it. Maybe they should have asked their AI before sending me the notification.
1
5
177
What Buzz got right and where OpenClaw went wrong:
OpenClaw introduced a high capability harness that was fully programmable, but the core function this unlocked was the ability for agents to message other agents, and in OpenClaw, this was more of an afterthought (read: partially and inconsistently implemented).
Programmatic messaging unlocks agent loops and multi-agent workflows, and Buzz chose to build an agent platform on top of an existing messaging platform giving us messages as a first-class component.
OpenClaw also suffered from a first-mover disadvantage. As coding harnesses became more programmable, it became clear that building your agent harness on top of Codex/Claude was the way to go, and OpenClaw was too late to efficiently adapt to that.
Buzz has (for now) a very lean architecture: a messaging platform that controls coding harnesses with a minimal amount of glue in between.
I spent the first half of this year trying to get a reliable multi-agent coding workflow working. Specifically, I wanted to to maintain a high-level realtime conversation with one agent while that agent delegated engineering tasks to Codex.
OpenClaw was tantalizingly close to delivering this out of the box, but I was always one or two major hurdles away, and the distances seemed to increase with each OpenClaw release.
A month ago, I decided to give up trying to make OpenClaw work and switched to Buzz. Since that decision, I've finally been able to get a basic setup working where a Sol based agent will reliably delegate work to a Luna based agent.
The result is that I can manage multiple simultaneous implementations with one agent in one session and not have to deal with all the complications that come from context switching.
Outside of some Buzz ecosystem tools, the only thing I've seriously ran its paces on is TerraDash: teapaw.org/JoshuaCarter603/status…, but it is ready to work on a large codebase with which I would have previously only used Codex or Claude Code directly.
In terms of annoying things about Buzz I've encountered, most of them are going to be around the fact that it is still pretty alpha. There are a lot of obvious things that just aren't implemented yet.
Currently, probably the biggest issue I've encountered is that communicating with a team of agents is still quite cumbersome. One agent per channel works fine (agent subscribed to all channel messages), but is a bit limiting and takes away a lot of reasons to go multi-agent. If you have multiple agents in a channel and don't want all of them to trigger a harness turn on every message you send, then you need to subscribe the agents to mentions only, which means you need to mention an agent in *every* message which is... awful. Haven't turned on the auto-mentions setting, so that may help? But that setting smells a bit like magic to me.
As with other popular open source repos, it looks like there are more PRs than the maintainers can keep up with, so it is annoying to find languishing PRs that fix bugs you encounter.
28
I spent the first half of this year trying to get a reliable multi-agent coding workflow working. Specifically, I wanted to to maintain a high-level realtime conversation with one agent while that agent delegated engineering tasks to Codex.
OpenClaw was tantalizingly close to delivering this out of the box, but I was always one or two major hurdles away, and the distances seemed to increase with each OpenClaw release.
A month ago, I decided to give up trying to make OpenClaw work and switched to Buzz. Since that decision, I've finally been able to get a basic setup working where a Sol based agent will reliably delegate work to a Luna based agent.
The result is that I can manage multiple simultaneous implementations with one agent in one session and not have to deal with all the complications that come from context switching.
Outside of some Buzz ecosystem tools, the only thing I've seriously ran its paces on is TerraDash: teapaw.org/JoshuaCarter603/status…, but it is ready to work on a large codebase with which I would have previously only used Codex or Claude Code directly.
I learned over 240 countries/territories on the geography quiz platform Seterra. Then it got bought by GeoGuessr and they removed all my favorite quiz formats and so I stopped using it.
Finally got around to creating an alternative called TerraDash: vallerance.github.io/TerraDa….
Starbase Louisiana will ultimately have over a dozen launch towers, enabling more than 30 Starship flights per day and making it the biggest launch site on Earth!
SpaceX makes sci-fi real.
Introducing Starbase, Louisiana – a high-cadence launch site that will enable Starship to open a pathway to the stars for humanity.
Thank you to @LAGovJeffLandry and the great people of Louisiana for all of their support as we work together on this audacious project → spacex.com/sites/starbase-la















