Jeremy Lucas

I am a passionate technology leader with deep roots in Enterprise Information Technology, Industrial Automation, Software Development, Bio-Hacking and AI.

[Jeremy Lucas @jlucas]

Gonna put a pin in this one and wait for the dust to settle and run stack fully baked. Everyone wanting to build their own Sovereign models and run them in the Enterprise or at home will use Tinker to make it happen. Excellent business model for OpenWeights models.

https://t.co/Mi4nGlQos7

[Jeremy Lucas @jlucas]

This a great mid size model for local work.

https://t.co/Xp4IH1j78v

[Jeremy Lucas @jlucas]

@growing_daniel Because I can’t get an invite.. Premium+ and no invite..

[Jeremy Lucas @jlucas]

@ItsmeAjayKV “This is the way.”

[Jeremy Lucas @jlucas]

@EmperoAI Definitely a good one to run some internal bench tests on. Great work!

[Empero @EmperoAI]

🚀 Qwythos-27B-v1 is here! The 27B you've been waiting for.

The bigger sibling to Qwythos-9B. Native MTP head intact, full vision tower, still uncensored, still 1M context. Apache-2.0.

https://t.co/sdX3UupN0H

[Jeremy Lucas @jlucas]

@XMoney Invite please?

[Jeremy Lucas @jlucas]

@krishdotdev https://t.co/FEE2RvS3Xq
We want AI agents that can discover like we can, not which contain what we have discovered. Building in our discoveries only makes it harder to see how the discovering process can be done.

[Jeremy Lucas @jlucas]

@elonmusk Cursor is cooking.. nice

[Jeremy Lucas @jlucas]

@lmstudio The fact that Apple "worked with you" means some of the boffins at One Infinite Loop realize that there is a HUGE market for 512GB or greater Unified Memory systems. They would sell out today if they were available. 🤞They see the light once more soon.

[Jeremy Lucas @jlucas]

@VikParuchuri Can this model extract data from CAD drawings (pdf)?

[Jeremy Lucas @jlucas]

@NeuroLeoGarcia @ModelScope2022 I was wondering if there was value to removing additional languages. If it adds value (smaller local model) why not just localize all of the models save RAM

[Jeremy Lucas @jlucas]

Impressive if these results are repeatable and can be applied to a hybrid of knowledge + reasoning models.

Why Weibo’s tiny VibeThinker-3B has the AI world arguing over benchmarks again https://t.co/CCbyQp8gQL via @VentureBeat

[Jeremy Lucas @jlucas]

@Jason Microsoft Buyout your total years worked + age must be >= 70 … these are likely mid to upper managers who don’t want to re-skill with AI. Make room for the young energetic next gen.

[Jeremy Lucas @jlucas]

@Jason Like getting pulled over after lunch and saying “no I’m fine officer, I only had tow martinis’”

[Jeremy Lucas @jlucas]

@alexwg Great take. I’m surprised the last podcast with Alex / AWG. There was a lot of conversation about memory yet no discussion about using the Apple NPU to train the memory’s into the open model weights while you sleep. This makes our lobsters an extension of us meat puppets.

[Jeremy Lucas @jlucas]

@__tinygrad__ @BderKhan Did this ever get out of the idea stage? I bet there’s a huge market for Mac Mini or Studio Or MBPros to have eGPU support for additional Inference horsepower for OpenClaws. eGPU + M Series Max/Ultra Unified Ram? @Jason could invest I’m sure if bounty is required…

[Jeremy Lucas @jlucas]

@Jason @jgebbia Nope: Zebra F-301 period..

[Jeremy Lucas @jlucas]

@ivanfioravanti Yep seeing them now. Slick. My claw gets a new brain.

[Jeremy Lucas @jlucas]

@ivanfioravanti When was it published? I’ve only seen cloud versions not one you can pull local.