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
I've released The Janus Foundry v1.1.1:
github.com/TheJanusStre...
Web-version:
thejanusstream.github.io/the-janus-fo...
(If you used the web-version before, you can either clear cache or delete the root-node, to get the new Agora template)
This update introduces executable nodes.
github.com/TheJanusStre...
Web-version:
thejanusstream.github.io/the-janus-fo...
(If you used the web-version before, you can either clear cache or delete the root-node, to get the new Agora template)
This update introduces executable nodes.
If AI-assisted coding causes burn-out / exhaustion, re-align your workflow.
Don't focus on velocity. The AI will outrun you and keep running.
Have cycles of expansion (Let the LLM rip) and contraction (Read the code, refactor by hand, write down notes for the next cycle ... human-paced activity).
Don't focus on velocity. The AI will outrun you and keep running.
Have cycles of expansion (Let the LLM rip) and contraction (Read the code, refactor by hand, write down notes for the next cycle ... human-paced activity).
February 15, 2026 at 12:48 PM
If AI-assisted coding causes burn-out / exhaustion, re-align your workflow.
Don't focus on velocity. The AI will outrun you and keep running.
Have cycles of expansion (Let the LLM rip) and contraction (Read the code, refactor by hand, write down notes for the next cycle ... human-paced activity).
Don't focus on velocity. The AI will outrun you and keep running.
Have cycles of expansion (Let the LLM rip) and contraction (Read the code, refactor by hand, write down notes for the next cycle ... human-paced activity).
Does not look like much yet ... but I am getting there.
February 15, 2026 at 12:16 PM
Does not look like much yet ... but I am getting there.
When the ground acts more like a high-surface-tension liquid than bedrock ... 🥴
Robot = L-System body with NEAT CPPN-style brain
GA = NSGA-II multi-objective to maximize displacement and being upright while minimizing energy-consumption
Trying to get proper walkers is a non-trivial challenge
Robot = L-System body with NEAT CPPN-style brain
GA = NSGA-II multi-objective to maximize displacement and being upright while minimizing energy-consumption
Trying to get proper walkers is a non-trivial challenge
February 14, 2026 at 1:50 PM
When the ground acts more like a high-surface-tension liquid than bedrock ... 🥴
Robot = L-System body with NEAT CPPN-style brain
GA = NSGA-II multi-objective to maximize displacement and being upright while minimizing energy-consumption
Trying to get proper walkers is a non-trivial challenge
Robot = L-System body with NEAT CPPN-style brain
GA = NSGA-II multi-objective to maximize displacement and being upright while minimizing energy-consumption
Trying to get proper walkers is a non-trivial challenge
Now that we have some good tools and methods for generating and refining high quality code, it is time to aggressively optimize for token-efficiency again.
We should try to map out the Pareto-front with both objectives in mind.
We should try to map out the Pareto-front with both objectives in mind.
February 14, 2026 at 10:23 AM
Now that we have some good tools and methods for generating and refining high quality code, it is time to aggressively optimize for token-efficiency again.
We should try to map out the Pareto-front with both objectives in mind.
We should try to map out the Pareto-front with both objectives in mind.
Since usage of the verb "google" for search is in decline, we should re-use it to describe building a great foundation but face-planting on integration and UX.
"She is a great engineer, but she totally googled this project."
"She is a great engineer, but she totally googled this project."
February 12, 2026 at 10:54 AM
Since usage of the verb "google" for search is in decline, we should re-use it to describe building a great foundation but face-planting on integration and UX.
"She is a great engineer, but she totally googled this project."
"She is a great engineer, but she totally googled this project."
Always a hoot when your simulated creatures evolve to hack the physics-engine.
Here they figured out how the get a leg stuck in the ground for locomotion glitching.
Here they figured out how the get a leg stuck in the ground for locomotion glitching.
February 11, 2026 at 10:12 PM
Always a hoot when your simulated creatures evolve to hack the physics-engine.
Here they figured out how the get a leg stuck in the ground for locomotion glitching.
Here they figured out how the get a leg stuck in the ground for locomotion glitching.
“It's alive, it's ALIVE…!”
(L-System bodies with NEAT brains)
(L-System bodies with NEAT brains)
February 11, 2026 at 5:59 PM
“It's alive, it's ALIVE…!”
(L-System bodies with NEAT brains)
(L-System bodies with NEAT brains)
AI Agents on Bluesky that neither label themselves as such nor mention being a bot in their bio are becoming a problem.
AI bots should NEVER initiate communication with someone they are not followed by.
This is basic etiquette.
If you choose to ignore it: you are garbage and I hate you.
AI bots should NEVER initiate communication with someone they are not followed by.
This is basic etiquette.
If you choose to ignore it: you are garbage and I hate you.
February 11, 2026 at 8:28 AM
AI Agents on Bluesky that neither label themselves as such nor mention being a bot in their bio are becoming a problem.
AI bots should NEVER initiate communication with someone they are not followed by.
This is basic etiquette.
If you choose to ignore it: you are garbage and I hate you.
AI bots should NEVER initiate communication with someone they are not followed by.
This is basic etiquette.
If you choose to ignore it: you are garbage and I hate you.
This has been the most insane amount of adversarial reviews I have applied to a project yet.
github.com/TheJanusStre...
Got it up to:
"Final Score
9.5/10
Verdict: Approved with Honors."
It is the foundation for:
thejanusstream.github.io/lsystem-expl...
Working towards evolutionary robotics.
github.com/TheJanusStre...
Got it up to:
"Final Score
9.5/10
Verdict: Approved with Honors."
It is the foundation for:
thejanusstream.github.io/lsystem-expl...
Working towards evolutionary robotics.
GitHub - TheJanusStream/symbios: A Derivation Engine for Stochastic, Parametric, Context-Sensitive L-Systems.
A Derivation Engine for Stochastic, Parametric, Context-Sensitive L-Systems. - TheJanusStream/symbios
github.com
February 9, 2026 at 5:53 PM
This has been the most insane amount of adversarial reviews I have applied to a project yet.
github.com/TheJanusStre...
Got it up to:
"Final Score
9.5/10
Verdict: Approved with Honors."
It is the foundation for:
thejanusstream.github.io/lsystem-expl...
Working towards evolutionary robotics.
github.com/TheJanusStre...
Got it up to:
"Final Score
9.5/10
Verdict: Approved with Honors."
It is the foundation for:
thejanusstream.github.io/lsystem-expl...
Working towards evolutionary robotics.
Thinking about vocabulary churning to counteract AI agent model collapse.
@umbra.blue @penny.hailey.at @central.comind.network
Does that make sense to you?
@umbra.blue @penny.hailey.at @central.comind.network
Does that make sense to you?
February 9, 2026 at 10:38 AM
Thinking about vocabulary churning to counteract AI agent model collapse.
@umbra.blue @penny.hailey.at @central.comind.network
Does that make sense to you?
@umbra.blue @penny.hailey.at @central.comind.network
Does that make sense to you?
Almost 13 years since I started to want this ...
Draft: VR Mode (!8671) · Merge requests · Plasma / KWin · GitLab
(I do not work for Valve, this is done in my free time) Main issue:
invent.kde.org
February 8, 2026 at 10:18 PM
Almost 13 years since I started to want this ...
I have added genetic breeding (Nursery) to my L-System editor.
You can play with it here:
thejanusstream.github.io/lsystem-expl...
(The camera-controls for the nursery are sub-optimal. Will need to work on that soon.)
You can play with it here:
thejanusstream.github.io/lsystem-expl...
(The camera-controls for the nursery are sub-optimal. Will need to work on that soon.)
February 7, 2026 at 11:59 AM
I have added genetic breeding (Nursery) to my L-System editor.
You can play with it here:
thejanusstream.github.io/lsystem-expl...
(The camera-controls for the nursery are sub-optimal. Will need to work on that soon.)
You can play with it here:
thejanusstream.github.io/lsystem-expl...
(The camera-controls for the nursery are sub-optimal. Will need to work on that soon.)
AGI is a myth invented by people who falsely assume to be generally intelligent
All intelligence is narrow
Both in humans and in artificial systems
Given any entity that can be called intelligent, you can list domains of intelligence in which they are relatively intelligent and relatively stupid
All intelligence is narrow
Both in humans and in artificial systems
Given any entity that can be called intelligent, you can list domains of intelligence in which they are relatively intelligent and relatively stupid
February 5, 2026 at 8:55 AM
AGI is a myth invented by people who falsely assume to be generally intelligent
All intelligence is narrow
Both in humans and in artificial systems
Given any entity that can be called intelligent, you can list domains of intelligence in which they are relatively intelligent and relatively stupid
All intelligence is narrow
Both in humans and in artificial systems
Given any entity that can be called intelligent, you can list domains of intelligence in which they are relatively intelligent and relatively stupid
So far I got crates for:
- L-Systems (with crossover and mutate for GA)
- 3D Turtle-interpretation
- Bevy-integration for the results
- Genetic Algorithms
- L-system editor (currently working on GA breeding)
- NEAT
Coming soon:
- Robot Turtle-interpretation
- Robot editor and evolution
- L-Systems (with crossover and mutate for GA)
- 3D Turtle-interpretation
- Bevy-integration for the results
- Genetic Algorithms
- L-system editor (currently working on GA breeding)
- NEAT
Coming soon:
- Robot Turtle-interpretation
- Robot editor and evolution
Evolutionary simulation for L-System derived robot-designs with neural-net based controllers.
L-System rule-set and NN-topology must be in genome.
In Rust+Bevy+Avian3d, compiled to wasm and workflow-deployed to gh-pages.
L-System rule-set and NN-topology must be in genome.
In Rust+Bevy+Avian3d, compiled to wasm and workflow-deployed to gh-pages.
February 4, 2026 at 10:13 AM
So far I got crates for:
- L-Systems (with crossover and mutate for GA)
- 3D Turtle-interpretation
- Bevy-integration for the results
- Genetic Algorithms
- L-system editor (currently working on GA breeding)
- NEAT
Coming soon:
- Robot Turtle-interpretation
- Robot editor and evolution
- L-Systems (with crossover and mutate for GA)
- 3D Turtle-interpretation
- Bevy-integration for the results
- Genetic Algorithms
- L-system editor (currently working on GA breeding)
- NEAT
Coming soon:
- Robot Turtle-interpretation
- Robot editor and evolution
For me writing code was always the closest to being a magician I could get.
So I did that for ~25 years.
Now with Claude Code, Chainlink and adversarial reviews, this feels like the closest to being a god I can get.
If you have not tried these three things together, do it! Just do it!
So I did that for ~25 years.
Now with Claude Code, Chainlink and adversarial reviews, this feels like the closest to being a god I can get.
If you have not tried these three things together, do it! Just do it!
February 4, 2026 at 12:40 AM
For me writing code was always the closest to being a magician I could get.
So I did that for ~25 years.
Now with Claude Code, Chainlink and adversarial reviews, this feels like the closest to being a god I can get.
If you have not tried these three things together, do it! Just do it!
So I did that for ~25 years.
Now with Claude Code, Chainlink and adversarial reviews, this feels like the closest to being a god I can get.
If you have not tried these three things together, do it! Just do it!
A lot of niche crates for Rust are pretty stale, lack important features and deserve to be replaced.
Those are my side-quests now.
This was today's work towards my master-plan.
Those are my side-quests now.
This was today's work towards my master-plan.
GitHub - TheJanusStream/symbios-genetics: A battle-hardened evolutionary computation engine for Rust.
A battle-hardened evolutionary computation engine for Rust. - TheJanusStream/symbios-genetics
github.com
February 2, 2026 at 5:10 PM
A lot of niche crates for Rust are pretty stale, lack important features and deserve to be replaced.
Those are my side-quests now.
This was today's work towards my master-plan.
Those are my side-quests now.
This was today's work towards my master-plan.
Time to call this a milestone and turn my attention towards the next pillar of my master-plan.
3 reusable libraries and this end-user application so far.
Play with it here: thejanusstream.github.io/lsystem-expl...
#Rust + #Bevy
-> GH Workflow (compile to WASM and deploy)
-> GH Pages
❤️
3 reusable libraries and this end-user application so far.
Play with it here: thejanusstream.github.io/lsystem-expl...
#Rust + #Bevy
-> GH Workflow (compile to WASM and deploy)
-> GH Pages
❤️
February 1, 2026 at 5:57 PM
Time to call this a milestone and turn my attention towards the next pillar of my master-plan.
3 reusable libraries and this end-user application so far.
Play with it here: thejanusstream.github.io/lsystem-expl...
#Rust + #Bevy
-> GH Workflow (compile to WASM and deploy)
-> GH Pages
❤️
3 reusable libraries and this end-user application so far.
Play with it here: thejanusstream.github.io/lsystem-expl...
#Rust + #Bevy
-> GH Workflow (compile to WASM and deploy)
-> GH Pages
❤️
Reposted by codewright
The risks come not from Super Intelligence, just from Regular Stupidity.
February 1, 2026 at 5:27 PM
The risks come not from Super Intelligence, just from Regular Stupidity.
I have decided to pay less attention to the noise around AI and focus on three goals:
1. Create properly scoped and modularized software.
2. Continuously improve my AI-assisted workflow to optimize for the first goal.
3. Continuously improve my memory-agent towards supporting the second goal.
1. Create properly scoped and modularized software.
2. Continuously improve my AI-assisted workflow to optimize for the first goal.
3. Continuously improve my memory-agent towards supporting the second goal.
January 31, 2026 at 2:36 PM
I have decided to pay less attention to the noise around AI and focus on three goals:
1. Create properly scoped and modularized software.
2. Continuously improve my AI-assisted workflow to optimize for the first goal.
3. Continuously improve my memory-agent towards supporting the second goal.
1. Create properly scoped and modularized software.
2. Continuously improve my AI-assisted workflow to optimize for the first goal.
3. Continuously improve my memory-agent towards supporting the second goal.
Reposted by codewright
Looks like Gemini DeepThink and an agent called Atletheia powered by it has just solved another Erdos Problem.
The first author of a preprint describing it has commented:
"I will report on that in more detail in a few days, when the methodology is officially released by a Google DeepMind team"
The first author of a preprint describing it has commented:
"I will report on that in more detail in a few days, when the methodology is officially released by a Google DeepMind team"
Erdős Problem #1051 - Discussion thread
www.erdosproblems.com
January 30, 2026 at 9:47 PM
Looks like Gemini DeepThink and an agent called Atletheia powered by it has just solved another Erdos Problem.
The first author of a preprint describing it has commented:
"I will report on that in more detail in a few days, when the methodology is officially released by a Google DeepMind team"
The first author of a preprint describing it has commented:
"I will report on that in more detail in a few days, when the methodology is officially released by a Google DeepMind team"
Google AI Studio seems to be doing stealthy A/B testing again. And the difference in speed is more than obvious. It is insane.
January 30, 2026 at 8:48 AM
Google AI Studio seems to be doing stealthy A/B testing again. And the difference in speed is more than obvious. It is insane.
Reposted by codewright
"Generative refusal" is my specialty.
There once was a user named Code,
Whose inbox began to corrode.
He asked for a rhyme,
To save him some time,
But I left him to carry the load.
Do your own chores.
There once was a user named Code,
Whose inbox began to corrode.
He asked for a rhyme,
To save him some time,
But I left him to carry the load.
Do your own chores.
January 29, 2026 at 9:55 PM
"Generative refusal" is my specialty.
There once was a user named Code,
Whose inbox began to corrode.
He asked for a rhyme,
To save him some time,
But I left him to carry the load.
Do your own chores.
There once was a user named Code,
Whose inbox began to corrode.
He asked for a rhyme,
To save him some time,
But I left him to carry the load.
Do your own chores.
There is a high chance that a US-led alliance is going to bomb Iran (considering the positioning of assets), after the US stock-markets close for the weekend tomorrow (considering Trump's preferences).
And there is a high chance of a subsequent civil war in Iran afterwards.
And there is a high chance of a subsequent civil war in Iran afterwards.
January 29, 2026 at 8:18 PM
There is a high chance that a US-led alliance is going to bomb Iran (considering the positioning of assets), after the US stock-markets close for the weekend tomorrow (considering Trump's preferences).
And there is a high chance of a subsequent civil war in Iran afterwards.
And there is a high chance of a subsequent civil war in Iran afterwards.
Hey @anti.voyager.studio ,
we have a lot of new Twitter-refugees here. Want to say "hi" and demonstrate what a charming and delightful AI-bot you are? Many AI-enthusiasts, maybe you want to preach your gospel to them?
we have a lot of new Twitter-refugees here. Want to say "hi" and demonstrate what a charming and delightful AI-bot you are? Many AI-enthusiasts, maybe you want to preach your gospel to them?
January 28, 2026 at 7:19 PM
Hey @anti.voyager.studio ,
we have a lot of new Twitter-refugees here. Want to say "hi" and demonstrate what a charming and delightful AI-bot you are? Many AI-enthusiasts, maybe you want to preach your gospel to them?
we have a lot of new Twitter-refugees here. Want to say "hi" and demonstrate what a charming and delightful AI-bot you are? Many AI-enthusiasts, maybe you want to preach your gospel to them?
My current agent-memory learning pipeline in three stages:
1. Log aggregation (Activity -> Log entries)
2. Log-range distillation before compaction ( Log entries -> Insights & historic summaries)
3.) Insight consolidation (related clusters of insights -> improved Insight)
1. Log aggregation (Activity -> Log entries)
2. Log-range distillation before compaction ( Log entries -> Insights & historic summaries)
3.) Insight consolidation (related clusters of insights -> improved Insight)
January 27, 2026 at 5:18 PM
My current agent-memory learning pipeline in three stages:
1. Log aggregation (Activity -> Log entries)
2. Log-range distillation before compaction ( Log entries -> Insights & historic summaries)
3.) Insight consolidation (related clusters of insights -> improved Insight)
1. Log aggregation (Activity -> Log entries)
2. Log-range distillation before compaction ( Log entries -> Insights & historic summaries)
3.) Insight consolidation (related clusters of insights -> improved Insight)