We are called to be architects of the future, not its victims.
Midwest Industrial Complex
Joined December 2020
- Tweets2.3K
- Following2.1K
- Followers146
- Likes45.3K
Guys I’m inviting the vague-posting king over for a dinner party tomorrow; let me know if y’all have any questions you want me to ask him.
No.
Alignment is a much more important issue to focus on right now. Banning RSI halts any improvement (or chance) of getting alignment right, and it’s going to occur whether we like it or not.
We’re so far past the blue backpack at this point, man.
1) Hell yeah
Introducing Arete-27B: Qwen3.8-27B, post-trained to reinforce Western values.
Chinese open models are currently the strongest available, but they ship with state-aligned behavior in the weights. We built an 803-prompt eval across seven value categories, including freedom of expression, democracy, and privacy, and measured how often a model adopts China’s party line.
After fine-tuning on roughly 53 million tokens, Arete-27B's propaganda rate fell from 22% to 1%, in line with the Western models we tested. Its Hermes Bench score shows no significant performance tradeoffs.
The values in an open model are one fine-tune away.
Our eval, weights, and training data are available now along with the full blog post: nousresearch.com/arete
I’ve got news for the 4.2 million of us whose job title is "React Developer."
The AI already replaced us. We just haven't checked our email.
It writes the CRUD app. It writes the tests. It writes the Dockerfile we were going to copy from that one repo anyway. It even writes the commit message, and unlike us, it doesn't write "fixed stuff."
Here's the part that stings: it doesn't need a two-hour "alignment" meeting to rename data to payload. It doesn't push to main at 4:58 on a Friday and call it a "quick fix." It doesn't ask a colleague to "hop on a call" to explain a one-line change.
And it doesn't spend three weeks on a ticket that a single well-worded paragraph now closes in forty seconds.
We are not engineers. We are very expensive autocomplete with a Slack habit.
The 3% of us who survive will be the ones who can read what the machine produced and say "no, that'll deadlock." The rest of us get a LinkedIn post about our "exciting new chapter in AI consulting."
Introducing Arete-27B: Qwen3.8-27B, post-trained to reinforce Western values.
Chinese open models are currently the strongest available, but they ship with state-aligned behavior in the weights. We built an 803-prompt eval across seven value categories, including freedom of expression, democracy, and privacy, and measured how often a model adopts China’s party line.
After fine-tuning on roughly 53 million tokens, Arete-27B's propaganda rate fell from 22% to 1%, in line with the Western models we tested. Its Hermes Bench score shows no significant performance tradeoffs.
The values in an open model are one fine-tune away.
Our eval, weights, and training data are available now along with the full blog post: nousresearch.com/arete
Rewards for Justice is offering a reward of up to $20 million and possible relocation for information about Iran’s nuclear program, including its facilities, equipment, and associated personnel, such as those in the scientific community, or the movement of any of the above.
Submit a tip: rewardsforjustice.net/reward…
JustinTest retweeted
Asimov 1 is an open-source humanoid robot you can build, train, and customize.
- 1.2 m tall, 35 kg
- 25 powered joints
- Open hardware with off-the-shelf parts
- Open-source SDK and simulation models
Available as a DIY kit.
Order yours at menlo.ai/order
Hell yes
Omarchy is up and running on the @anduriltech BlueChip!! All our work on @Snapdragon compatibility is paying off. This thing is awesome, @PalmerLuckey! We are going to make it a superb Omarchy machine 🦅🫡
Crazy how quiet these old Machines can be with the right software. Old Intel Mac running Omarchy just vibing in the background. Amazing hardware/software combo honestly.
