I can see Apple also supporting this natively in the future. Their privacy-first stand and hardware lineup supports this as well with a iPad on the go and Mac Studio at home.
@dhh , @0xSero ,
Let’s show them how to do this properly on @omarchy ecosystem so they can copy us. 😎
Omarchy 4.6 is not just leaning into local AI, but will let you share any model running on any Omarchy machine on your tailnet with another. Get a DGX Spark, then run inference from your XPS13 netbook. Sharing is caring.
49
Local AI will be an undeniable force. It won’t replace frontier models running on hyperscalers, but it will be an important part of the ecosystem.
We’re already seeing local speech models get better and better, with integrations into operating systems and SaaS applications.
Open-weight models may trail frontier models by several months, but many business application features don’t need the latest frontier capabilities. Speech is a good example of where local models could be more than good enough.
I also see opportunities in cybersecurity and maintenance. AI-enabled operating systems like @omarchy could eventually diagnose and repair issues locally, even in air-gapped environments.
And there’s the “IKEA effect”: people place more value on things they help build. Local models give users more control over shaping their own AI experience. Local AI will multiply the IKEA effect and give even more joy to the users.
Power to @omarchy. Thanks to @ashxhart . Of course hats off to @dhh for bringing us all together around @omarchy
TensorFold is coming to Omarchy 🚀
TensorFold recipes are already in Omarchy's local AI catalogue for Qwen3.8-27B, Flash Next and Nemotron 3.5 Bolt.
Seeing that is pretty wild. 🤯
Huge thanks to @dhh and the Omarchy contributors for all the work that went into this.
I want local AI to be easier and faster to use on your own machine, and I'm really excited to contribute to this.
github.com/omacom/omarchy-lo…
tensorfold.dev
3
1
67
Ike Kavas retweeted
Assuming human-written = good and AI-written = bad is lazy, and wrong.
I've spent plenty of my career debugging handcrafted true slop. Both can produce good work and garbage.
Critique the work. Explain what's wrong. Calling everything “AI slop” just shows your own ignorance.
Ike Kavas retweeted
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
Ike Kavas retweeted
A huge amount of the Anti-AI code sentiment massively overestimates the quality of human code outside of a very small set of open source and high quality company codebases. Human Slop is everywhere and can trivially be improved on by any opus level model.
Yesterday a prospect described what he needs, and it was exactly why Cameron Orr and I started Time Machine.
He wants his team ramped before their January kickoff, not during it. Kickoffs burn days on basics. Experienced reps sit through sessions they don't need, and new hires still leave overwhelmed. His plan was to have practice leaders record their sessions and share their decks, so everyone learns, gets tested and practices on their own beforehand. Then the live time goes to advanced roundtables on deal-making and AI.
When he explained the why and the how, he was describing what we built.
I hear this again and again from sales leaders, founders and enablement teams. Companies hire experienced people, sometimes straight from a competitor, and nobody has time to teach them how this company sells and operates. Twelve months later the story is either "they never really ramped" or "it's the Wild West," with every rep and account manager running the playbook from their last job.
And if you think training people is a thing of the past and you just need to train your AI agents, look at the companies we all admire, AI companies included. They still compete hard to hire great people, and they pay top dollar for them. AI doesn't make people worth less. It makes the right people worth a lot more. So making sure every hire actually delivers what you're paying them for matters more now, not less.
Hiring good people is the expensive part. Investing in how they ramp is what keeps you from regretting the hire.
That's what Time Machine does: it turns your leaders' knowledge into practice your team does before it matters.
1
41
Thursday I went to the MasterMinds Startup Accelerator Workshop hosted by Scott Fox for @octechweek. The people and the conversations were great. Even better was Scott's coaching and feedback for first-time founders, which was top notch. As I told him, if the room takes away even 10% of what he teaches, they'll do very well.
One theme comes up again and again with first-time founders: they think all investors are the same, and that investors are there to give them money to build their company.
I couldn't help chiming in.
Yes, some VCs and incubators will give you money to start or build a product. But 99% of investors are looking for companies that can return 50x what they put in, especially at the early stage. That means a founder with a product that can scale, and a value proposition already proven with customers. Everything else is the exception. You can find those exceptions, but probably not by sending 100 cold applications through VC website forms.
By definition, a founder's job is to build something from nothing, to go from zero to one. An investor's job is to make money from your something. Each side needs the other, and when great founders and great investors work together they lift each other up. That's also the exception. Most people focus on their own side of the coin.
If you're a first-time founder, prove your product by getting real customers to pay for it. If you need money to build before you can prove it, raise it from friends, people you know, and future customers willing to prepay at 50 cents on the dollar.
Thanks to Scott Fox, @octechweek, and Michael Merchant at Codazen for hosting.
#OCTechWeek #Founders #Startups #Fundraising
3
115
Glance face unlock by @ayandexyz now runs on my M1 MacBook Pro (@omarchy 4, Aurora kernel by @AuroraSilicon).
Swapped MediaPipe for onnxruntime on ARM and fixed camera stalls and detection. Small tested patches are up for review. Ayan, would love your eyes on them. @OmarchyMac
The patches:
Apple Silicon series: github.com/ayandexyz/glance-…
Omarchy plugin README: github.com/ayandexyz/omarchy…












