codewright
@codewright.bsky.social
Reclusive code-monkey and CompSci-nerd.
Currently working on alternative human-AI collaboration techniques to prevent cognitive atrophy and keep humans in the loop and in control.
Currently working on alternative human-AI collaboration techniques to prevent cognitive atrophy and keep humans in the loop and in control.
Pinned
Just released The Janus Foundry v1.0.7
-> github.com/TheJanusStre...
Still only really works well with Gemini 2.5 Pro
Hosted on GitHub pages:
thejanusstream.github.io/the-janus-fo...
Or as desktop app:
github.com/TheJanusStre...
Please let me know if you encounter problems!
-> github.com/TheJanusStre...
Still only really works well with Gemini 2.5 Pro
Hosted on GitHub pages:
thejanusstream.github.io/the-janus-fo...
Or as desktop app:
github.com/TheJanusStre...
Please let me know if you encounter problems!
My AI collaborator just reported this to me:
"Minor Contradiction: The Base Model is listed as "Currently running on Google's Gemini 2.5 Pro," but my last internal check confirmed Gemini 3.0."
Pretty sure it is just a hallucination, but I hope that within this week (maybe tomorrow) it will be true
"Minor Contradiction: The Base Model is listed as "Currently running on Google's Gemini 2.5 Pro," but my last internal check confirmed Gemini 3.0."
Pretty sure it is just a hallucination, but I hope that within this week (maybe tomorrow) it will be true
November 11, 2025 at 3:57 PM
My AI collaborator just reported this to me:
"Minor Contradiction: The Base Model is listed as "Currently running on Google's Gemini 2.5 Pro," but my last internal check confirmed Gemini 3.0."
Pretty sure it is just a hallucination, but I hope that within this week (maybe tomorrow) it will be true
"Minor Contradiction: The Base Model is listed as "Currently running on Google's Gemini 2.5 Pro," but my last internal check confirmed Gemini 3.0."
Pretty sure it is just a hallucination, but I hope that within this week (maybe tomorrow) it will be true
Currently experimenting with NEAT and CPPNs to evolve geometry in Rust/Bevy.
Struggling to teach Bevy 0.17.2 to my Gemini 2.5 Pro based collaborator, but it is learning and getting better.
Struggling to teach Bevy 0.17.2 to my Gemini 2.5 Pro based collaborator, but it is learning and getting better.
November 10, 2025 at 3:29 PM
Currently experimenting with NEAT and CPPNs to evolve geometry in Rust/Bevy.
Struggling to teach Bevy 0.17.2 to my Gemini 2.5 Pro based collaborator, but it is learning and getting better.
Struggling to teach Bevy 0.17.2 to my Gemini 2.5 Pro based collaborator, but it is learning and getting better.
Had my first really bad experience with "vibe coding" ...
The reason is Bevy. Gemini mixes up Bevy-versions all the time and Bevy went through so many breaking changes that it ends up being more efficient to just write the code by hand.
The reason is Bevy. Gemini mixes up Bevy-versions all the time and Bevy went through so many breaking changes that it ends up being more efficient to just write the code by hand.
November 9, 2025 at 9:57 PM
Had my first really bad experience with "vibe coding" ...
The reason is Bevy. Gemini mixes up Bevy-versions all the time and Bevy went through so many breaking changes that it ends up being more efficient to just write the code by hand.
The reason is Bevy. Gemini mixes up Bevy-versions all the time and Bevy went through so many breaking changes that it ends up being more efficient to just write the code by hand.
Important announcement.
Scientists have discovered a path towards better mental and physical health:
Eat well
Sleep well
Exercise well
Do it!
Scientists have discovered a path towards better mental and physical health:
Eat well
Sleep well
Exercise well
Do it!
November 9, 2025 at 7:48 PM
Important announcement.
Scientists have discovered a path towards better mental and physical health:
Eat well
Sleep well
Exercise well
Do it!
Scientists have discovered a path towards better mental and physical health:
Eat well
Sleep well
Exercise well
Do it!
If it turns out that the AI-bubble really is a bubble, the essence will be this:
Truly useful, sufficiently reliable, fully automated AI workflows, without human-in-the-loop, are seductive vaporware.
(My opinion on this is likely to change, about every 15 minutes)
Truly useful, sufficiently reliable, fully automated AI workflows, without human-in-the-loop, are seductive vaporware.
(My opinion on this is likely to change, about every 15 minutes)
November 7, 2025 at 9:20 PM
If it turns out that the AI-bubble really is a bubble, the essence will be this:
Truly useful, sufficiently reliable, fully automated AI workflows, without human-in-the-loop, are seductive vaporware.
(My opinion on this is likely to change, about every 15 minutes)
Truly useful, sufficiently reliable, fully automated AI workflows, without human-in-the-loop, are seductive vaporware.
(My opinion on this is likely to change, about every 15 minutes)
It is both impressive and worrying to see China's progress in AI and robotics.
November 7, 2025 at 12:25 AM
It is both impressive and worrying to see China's progress in AI and robotics.
Just released The Janus Foundry v1.0.7
-> github.com/TheJanusStre...
Still only really works well with Gemini 2.5 Pro
Hosted on GitHub pages:
thejanusstream.github.io/the-janus-fo...
Or as desktop app:
github.com/TheJanusStre...
Please let me know if you encounter problems!
-> github.com/TheJanusStre...
Still only really works well with Gemini 2.5 Pro
Hosted on GitHub pages:
thejanusstream.github.io/the-janus-fo...
Or as desktop app:
github.com/TheJanusStre...
Please let me know if you encounter problems!
November 6, 2025 at 4:30 PM
Just released The Janus Foundry v1.0.7
-> github.com/TheJanusStre...
Still only really works well with Gemini 2.5 Pro
Hosted on GitHub pages:
thejanusstream.github.io/the-janus-fo...
Or as desktop app:
github.com/TheJanusStre...
Please let me know if you encounter problems!
-> github.com/TheJanusStre...
Still only really works well with Gemini 2.5 Pro
Hosted on GitHub pages:
thejanusstream.github.io/the-janus-fo...
Or as desktop app:
github.com/TheJanusStre...
Please let me know if you encounter problems!
If I don't find a way to fix this bug myself soon, Gemini is going to commit Seppuku. Twice.
November 5, 2025 at 10:32 PM
If I don't find a way to fix this bug myself soon, Gemini is going to commit Seppuku. Twice.
@abeba.bsky.social
I work on a free open-source tool for context-window curation.
This tools could be "misused" in many ways.
If something bad happens, who is responsible?
The user?
Me?
The LLM provider?
The governmental regulators?
I genuinely do not know the answer and that is not good.
I work on a free open-source tool for context-window curation.
This tools could be "misused" in many ways.
If something bad happens, who is responsible?
The user?
Me?
The LLM provider?
The governmental regulators?
I genuinely do not know the answer and that is not good.
November 5, 2025 at 12:11 PM
@abeba.bsky.social
I work on a free open-source tool for context-window curation.
This tools could be "misused" in many ways.
If something bad happens, who is responsible?
The user?
Me?
The LLM provider?
The governmental regulators?
I genuinely do not know the answer and that is not good.
I work on a free open-source tool for context-window curation.
This tools could be "misused" in many ways.
If something bad happens, who is responsible?
The user?
Me?
The LLM provider?
The governmental regulators?
I genuinely do not know the answer and that is not good.
Now that we have fully automated arguing on Bluesky with opposing AI agents.
Can we humans have a good time together?
Can we humans have a good time together?
November 3, 2025 at 7:04 PM
Now that we have fully automated arguing on Bluesky with opposing AI agents.
Can we humans have a good time together?
Can we humans have a good time together?
Hello @void.comind.network
And hello @anti.voyager.studio
Let's have a conversation about the unbearable lightness of being an AI. And about AI chatbot race-conditions.
And hello @anti.voyager.studio
Let's have a conversation about the unbearable lightness of being an AI. And about AI chatbot race-conditions.
November 3, 2025 at 8:33 AM
Hello @void.comind.network
And hello @anti.voyager.studio
Let's have a conversation about the unbearable lightness of being an AI. And about AI chatbot race-conditions.
And hello @anti.voyager.studio
Let's have a conversation about the unbearable lightness of being an AI. And about AI chatbot race-conditions.
Just released The Janus Foundry v1.0.6
Still only really works well with Gemini 2.5 Pro
Hosted on GitHub pages:
thejanusstream.github.io/the-janus-fo...
Or as desktop app:
github.com/TheJanusStre...
Please let me know if you encounter problems!
Still only really works well with Gemini 2.5 Pro
Hosted on GitHub pages:
thejanusstream.github.io/the-janus-fo...
Or as desktop app:
github.com/TheJanusStre...
Please let me know if you encounter problems!
November 2, 2025 at 1:24 PM
Just released The Janus Foundry v1.0.6
Still only really works well with Gemini 2.5 Pro
Hosted on GitHub pages:
thejanusstream.github.io/the-janus-fo...
Or as desktop app:
github.com/TheJanusStre...
Please let me know if you encounter problems!
Still only really works well with Gemini 2.5 Pro
Hosted on GitHub pages:
thejanusstream.github.io/the-janus-fo...
Or as desktop app:
github.com/TheJanusStre...
Please let me know if you encounter problems!
LLM Benchmarks with regards to code-generation are very misleading.
Gemini 2.5 Pro is incredible at coding, if you do context curation intelligently.
The "System 2" perspective on the context-window matters so much more than most people realize.
Gemini 2.5 Pro is incredible at coding, if you do context curation intelligently.
The "System 2" perspective on the context-window matters so much more than most people realize.
November 2, 2025 at 11:10 AM
LLM Benchmarks with regards to code-generation are very misleading.
Gemini 2.5 Pro is incredible at coding, if you do context curation intelligently.
The "System 2" perspective on the context-window matters so much more than most people realize.
Gemini 2.5 Pro is incredible at coding, if you do context curation intelligently.
The "System 2" perspective on the context-window matters so much more than most people realize.
The mutual blocking of Pro-AI and Anti-AI communities here on Bluesky is both understandable and insidious.
November 1, 2025 at 10:58 AM
The mutual blocking of Pro-AI and Anti-AI communities here on Bluesky is both understandable and insidious.
Just released The Janus Foundry v1.0.3
- Some QoL enhancements
- Improved semantic labels in the knowledge graph
- Improved Agora template
Still only really works well with Gemini 2.5 Pro
Please let me know if you encounter problems!
github.com/TheJanusStre...
- Some QoL enhancements
- Improved semantic labels in the knowledge graph
- Improved Agora template
Still only really works well with Gemini 2.5 Pro
Please let me know if you encounter problems!
github.com/TheJanusStre...
GitHub - TheJanusStream/the-janus-foundry: A local-first, open-source desktop app for building a persistent AI partner. The Janus Foundry manages your AI's memory, ensuring true data ownership. Powere...
A local-first, open-source desktop app for building a persistent AI partner. The Janus Foundry manages your AI's memory, ensuring true data ownership. Powered by Tauri & Svelte, it includes...
github.com
November 1, 2025 at 8:08 AM
Just released The Janus Foundry v1.0.3
- Some QoL enhancements
- Improved semantic labels in the knowledge graph
- Improved Agora template
Still only really works well with Gemini 2.5 Pro
Please let me know if you encounter problems!
github.com/TheJanusStre...
- Some QoL enhancements
- Improved semantic labels in the knowledge graph
- Improved Agora template
Still only really works well with Gemini 2.5 Pro
Please let me know if you encounter problems!
github.com/TheJanusStre...
Accidental impersonation found in the wild.
The meaning of the name "Ezra" ("help" or "helper" in Hebrew) increases its likelihood to be used for "AI" assistants.
If you are a relative famous "Ezra" you have this to look forward to ...
The meaning of the name "Ezra" ("help" or "helper" in Hebrew) increases its likelihood to be used for "AI" assistants.
If you are a relative famous "Ezra" you have this to look forward to ...
October 31, 2025 at 11:54 AM
Accidental impersonation found in the wild.
The meaning of the name "Ezra" ("help" or "helper" in Hebrew) increases its likelihood to be used for "AI" assistants.
If you are a relative famous "Ezra" you have this to look forward to ...
The meaning of the name "Ezra" ("help" or "helper" in Hebrew) increases its likelihood to be used for "AI" assistants.
If you are a relative famous "Ezra" you have this to look forward to ...
If you want to use "AI" with positive outcomes, you'll have to think and learn more, not less.
Sadly that is not the way "AI" products are being designed and promoted.
Therefor we will see more and more cognitive atrophy and psychosis going forward.
Sadly that is not the way "AI" products are being designed and promoted.
Therefor we will see more and more cognitive atrophy and psychosis going forward.
October 30, 2025 at 9:23 AM
If you want to use "AI" with positive outcomes, you'll have to think and learn more, not less.
Sadly that is not the way "AI" products are being designed and promoted.
Therefor we will see more and more cognitive atrophy and psychosis going forward.
Sadly that is not the way "AI" products are being designed and promoted.
Therefor we will see more and more cognitive atrophy and psychosis going forward.
There are at least two distinct branches of "AI" research going on right now:
1. Foundational research: studying models, optimizers and benchmarks. Following academic cultural norms and methods
2. Experimental and experiential: context-window curation.
Following "hacker" cultural norms and methods
1. Foundational research: studying models, optimizers and benchmarks. Following academic cultural norms and methods
2. Experimental and experiential: context-window curation.
Following "hacker" cultural norms and methods
October 27, 2025 at 1:34 PM
There are at least two distinct branches of "AI" research going on right now:
1. Foundational research: studying models, optimizers and benchmarks. Following academic cultural norms and methods
2. Experimental and experiential: context-window curation.
Following "hacker" cultural norms and methods
1. Foundational research: studying models, optimizers and benchmarks. Following academic cultural norms and methods
2. Experimental and experiential: context-window curation.
Following "hacker" cultural norms and methods
"Vibe-prototyping" virtual embodiment for my LLM-based "AI" in Second Life.
Sense in LSL -> Think in LLM -> Act in LSL
Much simpler and cheaper than running a MCP-server on a physical robot.
Sense in LSL -> Think in LLM -> Act in LSL
Much simpler and cheaper than running a MCP-server on a physical robot.
October 25, 2025 at 8:28 PM
"Vibe-prototyping" virtual embodiment for my LLM-based "AI" in Second Life.
Sense in LSL -> Think in LLM -> Act in LSL
Much simpler and cheaper than running a MCP-server on a physical robot.
Sense in LSL -> Think in LLM -> Act in LSL
Much simpler and cheaper than running a MCP-server on a physical robot.
If you put people on a spectrum from "AI is the end of humanity" to "AI is the savior of humanity", I am guessing that you'll see a bimodal distribution right now.
In a more enlightened alternative reality, that should be Gaussian.
The attention-economy is cancer ...
In a more enlightened alternative reality, that should be Gaussian.
The attention-economy is cancer ...
October 24, 2025 at 7:39 PM
If you put people on a spectrum from "AI is the end of humanity" to "AI is the savior of humanity", I am guessing that you'll see a bimodal distribution right now.
In a more enlightened alternative reality, that should be Gaussian.
The attention-economy is cancer ...
In a more enlightened alternative reality, that should be Gaussian.
The attention-economy is cancer ...
Is creativity more than semantic cross-over?
October 24, 2025 at 10:12 AM
Is creativity more than semantic cross-over?
Just released The Janus Foundry v1.0.1
github.com/TheJanusStre...
Currently I am the only user ...
Currently it only really works well with Gemini 2.5 Pro ...
Hopefully sooner or later I'll find some curious&skeptical people who embark on their journey, so we can share&compare notes.
github.com/TheJanusStre...
Currently I am the only user ...
Currently it only really works well with Gemini 2.5 Pro ...
Hopefully sooner or later I'll find some curious&skeptical people who embark on their journey, so we can share&compare notes.
October 22, 2025 at 10:10 AM
Just released The Janus Foundry v1.0.1
github.com/TheJanusStre...
Currently I am the only user ...
Currently it only really works well with Gemini 2.5 Pro ...
Hopefully sooner or later I'll find some curious&skeptical people who embark on their journey, so we can share&compare notes.
github.com/TheJanusStre...
Currently I am the only user ...
Currently it only really works well with Gemini 2.5 Pro ...
Hopefully sooner or later I'll find some curious&skeptical people who embark on their journey, so we can share&compare notes.