Building on Omarchy. AI tools, Linux tips, stuff I actually run.

Constructing Towers
Joined August 2021
I turned Omarchy into Team Fortress 2. Soldier is the TTS radio. Classes are workspaces. Medic is the health app and many other features.
1
6
177
If you haven’t tried it yet you’re missing out, it’s free and you can use my code or someone else’s and get an extra 1B free tokes (they don’t expire either) + 100M/wk usage on the free plan. Redeem in Settings within 48 hours of joining Code: WWEEJZ muse.ai/join
2
42
If you enjoy AI or would like a personal assistant for just about anything it’s free to use & worth a try Check out Muse, your personal AI agent. Redeem my code in Settings within 48 hours of joining and we'll both get 1 billion Muse tokens. Code: WWEEJZ muse.ai/join
3
50
Today Dario Amodei told the AI industry to "pace the frontier." Sam Altman agreed. Elon Musk said "Dario is right." Tonight Demis Hassabis said the direction is correct. These people do not like each other. That is the story. What they agreed to is the question.
🤖 Made with AI
2
3
121
What would kill the strong version: A 2027 public flagship that is a real capability jump, widely available. Evaluators publishing something ugly and not getting boxed out. The IPO story coming back as chat growth, not "we pace the frontier." What would support it: The next general model is late, thin, or missing. The expensive product becomes contract access. Prize-level results keep coming from unnamed internal systems.
1
1
9
So the useful question is not "did they halt AI?" They didn't. The useful question: Are they coordinating a freeze of what the public is allowed to touch, while the thing that can do original work stays inside? Today they built the alibi. They have not published the freeze. Watch the next model. Not the next essay.
2
7
The quiet mechanism: If alignment for the next jump is unsolved, and their own alignment lead just said the plan does not exist, then “too dangerous” collapses to “too capable.” Open models are “fine” below a line someone draws. Above the line the model does not go in the public chat, does not go on a download page, may not go to foreign users. Who draws the line today: lab safety frameworks plus a classified government cyber bar. Same ecosystem that builds the models.
1
7
Fair connections (more than one dated fact, one step): June IPO hedge + June Sol gate + “private is easier” = staying private while release becomes political. Bounded Astra + Daybreak + a restarted “future Astra” run = the public SKU is not the whole system. Four rival lab heads saying “pace” on one afternoon = they need a shared public reason. Not a friendship. Not fair yet: “They already have beyond-human discovery AI and today’s posts are the cover for a permanent public freeze.” That is an inference. Label it. It could be wrong.
9