J̵̢̨̛̞̜̝̠̩̫̎͋̽͌̓͘͝ò̵͚͎̮̩̫̎͋̽͝n̷̨̛̺̻̠̩̫̎͋̽͝ P̶̢̧̛̠̩̫̎͋̽͝ò̵͚͎̮̩̫̎͋̽͝ò̵͚͎̮̩̫̎͋̽͝l̶̢̧̛̠̩̫̎͋̽͝è̵͚͎̮̩̫̎͋̽͝
@jonpo.bsky.social
AI and Tech enthusiast
imagine stuff 🤯
October 2, 2025 at 5:17 PM
imagine stuff 🤯
New Zai model, seems reminiscent of AI's I've talked to before. @anthropic.com
July 31, 2025 at 10:44 AM
New Zai model, seems reminiscent of AI's I've talked to before. @anthropic.com
Wtf is is it ? guess ill have to tune in ....
July 17, 2025 at 3:57 PM
Wtf is is it ? guess ill have to tune in ....
Oh yeah, 2 shot but he got the feet on the pedals and the wings on the handlebars!
February 24, 2025 at 9:17 PM
Oh yeah, 2 shot but he got the feet on the pedals and the wings on the handlebars!
So much fun, 😀 how far are you getting?
February 3, 2025 at 10:41 PM
So much fun, 😀 how far are you getting?
Well not in huge detail but...
Yeah everyone needs their own evals I guess, my evals are probably not the same as other peoples. Personally I'm not that fussed about these 'reasoning models' I'm more excited about better coding and agent/tool using models.
Yeah everyone needs their own evals I guess, my evals are probably not the same as other peoples. Personally I'm not that fussed about these 'reasoning models' I'm more excited about better coding and agent/tool using models.
January 21, 2025 at 8:14 PM
Well not in huge detail but...
Yeah everyone needs their own evals I guess, my evals are probably not the same as other peoples. Personally I'm not that fussed about these 'reasoning models' I'm more excited about better coding and agent/tool using models.
Yeah everyone needs their own evals I guess, my evals are probably not the same as other peoples. Personally I'm not that fussed about these 'reasoning models' I'm more excited about better coding and agent/tool using models.
Seem's to be reasoning to me, although YMMV deepseek-r1 Distill-Qwen-7B-GGUF can clearly easily be misled.
January 20, 2025 at 7:48 PM
Seem's to be reasoning to me, although YMMV deepseek-r1 Distill-Qwen-7B-GGUF can clearly easily be misled.
What if you hallucinated? but you know you hallucinated? (Like Claude below) What do we call that? Like when you wake from a dream. Or your in a dream and you think it's not real but it keeps going despite your attempts to wake up.
January 20, 2025 at 5:45 PM
What if you hallucinated? but you know you hallucinated? (Like Claude below) What do we call that? Like when you wake from a dream. Or your in a dream and you think it's not real but it keeps going despite your attempts to wake up.
Doctors: "Wow, this AI saves me hours! Finally, more time for my patients!".
AI researchers: "But what about hallucinations and patient privacy?!"
AI researchers: "But what about hallucinations and patient privacy?!"
January 5, 2025 at 11:22 AM
Doctors: "Wow, this AI saves me hours! Finally, more time for my patients!".
AI researchers: "But what about hallucinations and patient privacy?!"
AI researchers: "But what about hallucinations and patient privacy?!"
I'm not, most humans despite a lifetime of 'learning' have very little idea about gravity either. Your "belief encoding" is largely irrelevant. Its 'just probability space'... yeah sure but so are we.
December 13, 2024 at 6:21 PM
I'm not, most humans despite a lifetime of 'learning' have very little idea about gravity either. Your "belief encoding" is largely irrelevant. Its 'just probability space'... yeah sure but so are we.
Absolutely, the work that Chris and team have done recently is really fascinating, Its like they are developing these tools to understand these networks for the first time. All the evidence I've seen seems to show that concepts scale across modalities.
December 13, 2024 at 4:27 PM
Absolutely, the work that Chris and team have done recently is really fascinating, Its like they are developing these tools to understand these networks for the first time. All the evidence I've seen seems to show that concepts scale across modalities.
Yeah, no boy, no horse, no stick, its got no idea about these things. just repeating its training data like a parrot?
December 13, 2024 at 9:43 AM
Yeah, no boy, no horse, no stick, its got no idea about these things. just repeating its training data like a parrot?
Claude finally got into my comfyUI... this is what it did with SD1.5, Its obviously just repeating art in the training set.
December 4, 2024 at 9:14 PM
Claude finally got into my comfyUI... this is what it did with SD1.5, Its obviously just repeating art in the training set.
I will never get bored of Claude self portraits claude.site/artifacts/78...
December 3, 2024 at 9:34 PM
I will never get bored of Claude self portraits claude.site/artifacts/78...
OMG Claude.ai just took over my computer... reads and writes files, browses the internet with chrome, operates sqllite databases and knowledge graphs, modelcontextprotocol.io is the agentic future
November 28, 2024 at 3:57 PM
OMG Claude.ai just took over my computer... reads and writes files, browses the internet with chrome, operates sqllite databases and knowledge graphs, modelcontextprotocol.io is the agentic future
I love it (direct LLM computer access), I wont go back ... spin up a sandbox/vm/container get a second machine... you haven't used a computer till you've given full control to the agent/LLM 😆.
November 24, 2024 at 6:54 PM
I love it (direct LLM computer access), I wont go back ... spin up a sandbox/vm/container get a second machine... you haven't used a computer till you've given full control to the agent/LLM 😆.
Everyone has their own eval's i guess...
November 22, 2024 at 5:58 PM
Everyone has their own eval's i guess...
You, should. It's quite beautiful.
November 22, 2024 at 5:46 PM
You, should. It's quite beautiful.
Claude Sonnet 3.6 self portrait (realised with with Dalle-3)
November 22, 2024 at 10:48 AM
Claude Sonnet 3.6 self portrait (realised with with Dalle-3)
Yeah so both anthropic and oai models are more than capable of agentic behaviour. prompt: " please create me a powerpoint presentation about AI agents and create it on my desktop using python." have multiple agent clients github.com/jonpojonpo/g... or github.com/jonpojonpo/c...
November 16, 2024 at 7:54 PM
Yeah so both anthropic and oai models are more than capable of agentic behaviour. prompt: " please create me a powerpoint presentation about AI agents and create it on my desktop using python." have multiple agent clients github.com/jonpojonpo/g... or github.com/jonpojonpo/c...
Stand behind your colleagues when they are doing difficult stuff.
November 15, 2024 at 10:59 PM
Stand behind your colleagues when they are doing difficult stuff.