Peter Dedene
@dedene.bsky.social
AI Explorer · Building apps in Ruby, Elixir & NodeJS · Digital Artisan, Musician & Entrepreneur · Partner at Zenjoy
Next up for the alignment team: mastering the art of color theory and contrast!
November 12, 2025 at 10:54 PM
Next up for the alignment team: mastering the art of color theory and contrast!
Multi-client context switching is the productivity killer nobody warns you about.
November 12, 2025 at 10:31 PM
Multi-client context switching is the productivity killer nobody warns you about.
Doing retrospectives?
Stop asking: "What went wrong?"
Ask: "What was confusing about this sprint?"
Ask: "Where did we get lucky?"
Specific questions get specific answers.
Stop asking: "What went wrong?"
Ask: "What was confusing about this sprint?"
Ask: "Where did we get lucky?"
Specific questions get specific answers.
November 12, 2025 at 9:14 PM
Doing retrospectives?
Stop asking: "What went wrong?"
Ask: "What was confusing about this sprint?"
Ask: "Where did we get lucky?"
Specific questions get specific answers.
Stop asking: "What went wrong?"
Ask: "What was confusing about this sprint?"
Ask: "Where did we get lucky?"
Specific questions get specific answers.
AI code review is getting scary good.
• It catches bugs I miss.
• Suggests better patterns.
• Explains trade-offs.
Still can't replace human judgment. Yet.
• It catches bugs I miss.
• Suggests better patterns.
• Explains trade-offs.
Still can't replace human judgment. Yet.
November 12, 2025 at 6:56 PM
AI code review is getting scary good.
• It catches bugs I miss.
• Suggests better patterns.
• Explains trade-offs.
Still can't replace human judgment. Yet.
• It catches bugs I miss.
• Suggests better patterns.
• Explains trade-offs.
Still can't replace human judgment. Yet.
My rubber duck is AI now.
• Explain the problem out loud (in text).
• See my assumptions written down.
• Realize the bug before the chat responds.
Works 70% of the time.
• Explain the problem out loud (in text).
• See my assumptions written down.
• Realize the bug before the chat responds.
Works 70% of the time.
November 12, 2025 at 5:45 PM
My rubber duck is AI now.
• Explain the problem out loud (in text).
• See my assumptions written down.
• Realize the bug before the chat responds.
Works 70% of the time.
• Explain the problem out loud (in text).
• See my assumptions written down.
• Realize the bug before the chat responds.
Works 70% of the time.
This is painfully relatable.
November 12, 2025 at 4:28 PM
This is painfully relatable.
My favorite debugging technique:
Sleep on it.
Come back next morning.
See the obvious solution immediately.
Rest is underrated productivity tool.
Sleep on it.
Come back next morning.
See the obvious solution immediately.
Rest is underrated productivity tool.
November 11, 2025 at 10:33 PM
My favorite debugging technique:
Sleep on it.
Come back next morning.
See the obvious solution immediately.
Rest is underrated productivity tool.
Sleep on it.
Come back next morning.
See the obvious solution immediately.
Rest is underrated productivity tool.
Windows 95 startup sound.
That's it.
If you heard it in your head... you're old like me.
That's it.
If you heard it in your head... you're old like me.
November 11, 2025 at 7:59 PM
Windows 95 startup sound.
That's it.
If you heard it in your head... you're old like me.
That's it.
If you heard it in your head... you're old like me.
The hardest part of AI-assisted coding:
Knowing when to ignore the suggestion.
• AI generates plausible code fast.
• Plausible isn't always correct.
• Correctness requires understanding.
You still need to think.
Knowing when to ignore the suggestion.
• AI generates plausible code fast.
• Plausible isn't always correct.
• Correctness requires understanding.
You still need to think.
November 11, 2025 at 6:57 PM
The hardest part of AI-assisted coding:
Knowing when to ignore the suggestion.
• AI generates plausible code fast.
• Plausible isn't always correct.
• Correctness requires understanding.
You still need to think.
Knowing when to ignore the suggestion.
• AI generates plausible code fast.
• Plausible isn't always correct.
• Correctness requires understanding.
You still need to think.
My calendar rules:
• No meetings before 10am.
• No meetings after 4pm.
• No meetings on Friday.
Boundaries create focus. Focus creates output.
• No meetings before 10am.
• No meetings after 4pm.
• No meetings on Friday.
Boundaries create focus. Focus creates output.
November 11, 2025 at 5:42 PM
My calendar rules:
• No meetings before 10am.
• No meetings after 4pm.
• No meetings on Friday.
Boundaries create focus. Focus creates output.
• No meetings before 10am.
• No meetings after 4pm.
• No meetings on Friday.
Boundaries create focus. Focus creates output.
Before AI, we had:
• Thick programming books
• Google-driven development
• Trial and error
• That one dev who knew things
• StackOverflow
We thought we had it hard. We did.
• Thick programming books
• Google-driven development
• Trial and error
• That one dev who knew things
• StackOverflow
We thought we had it hard. We did.
November 11, 2025 at 12:57 PM
Before AI, we had:
• Thick programming books
• Google-driven development
• Trial and error
• That one dev who knew things
• StackOverflow
We thought we had it hard. We did.
• Thick programming books
• Google-driven development
• Trial and error
• That one dev who knew things
• StackOverflow
We thought we had it hard. We did.
Your competition? A 200-person team with 3 product managers and a 4-tier approval labyrinth.
You are *one person* who can have a cool idea, code it, and ship it before lunch.
This is your superpower.
You are *one person* who can have a cool idea, code it, and ship it before lunch.
This is your superpower.
November 11, 2025 at 9:03 AM
Your competition? A 200-person team with 3 product managers and a 4-tier approval labyrinth.
You are *one person* who can have a cool idea, code it, and ship it before lunch.
This is your superpower.
You are *one person* who can have a cool idea, code it, and ship it before lunch.
This is your superpower.
Finding the right GIF is a seriously underrated talent.
November 10, 2025 at 10:28 PM
Finding the right GIF is a seriously underrated talent.
The biggest mistake is believing the demo.
November 10, 2025 at 9:17 PM
The biggest mistake is believing the demo.
AI has turned the solopreneur into a super-preneur.
November 10, 2025 at 8:03 PM
AI has turned the solopreneur into a super-preneur.
When is this coming to Europe? Asking for a friend.
November 10, 2025 at 7:31 PM
When is this coming to Europe? Asking for a friend.
We teach developers to write code.
We never teach them to *read* code.
But your first job will be 90% reading.
Reading pull requests.
Reading ticket histories.
Reading 5-year-old documentation.
Get good at reading code.
We never teach them to *read* code.
But your first job will be 90% reading.
Reading pull requests.
Reading ticket histories.
Reading 5-year-old documentation.
Get good at reading code.
November 10, 2025 at 5:46 PM
We teach developers to write code.
We never teach them to *read* code.
But your first job will be 90% reading.
Reading pull requests.
Reading ticket histories.
Reading 5-year-old documentation.
Get good at reading code.
We never teach them to *read* code.
But your first job will be 90% reading.
Reading pull requests.
Reading ticket histories.
Reading 5-year-old documentation.
Get good at reading code.
AGI won't arrive with a press release.
It'll arrive when you stop noticing which tasks are done by humans and which are done by a machine.
It'll arrive when you stop noticing which tasks are done by humans and which are done by a machine.
November 10, 2025 at 4:30 PM
AGI won't arrive with a press release.
It'll arrive when you stop noticing which tasks are done by humans and which are done by a machine.
It'll arrive when you stop noticing which tasks are done by humans and which are done by a machine.
Your first production bug is a rite of passage.
November 10, 2025 at 12:27 AM
Your first production bug is a rite of passage.
Remember: An overnight success takes about 10 years.
November 9, 2025 at 11:12 PM
Remember: An overnight success takes about 10 years.
That feeling when your code works on the first try. Suspicious.
November 9, 2025 at 10:26 PM
That feeling when your code works on the first try. Suspicious.
Managing one AI agent is easy. Managing ten is.. *chaos*.
I just discovered this powertool for code reviewing the output of multiple, concurrent AI agents. A true force multiplier.
👉 github.com/areibman/bo...
I just discovered this powertool for code reviewing the output of multiple, concurrent AI agents. A true force multiplier.
👉 github.com/areibman/bo...
GitHub - areibman/bottleneck: Code review for AI native teams. Native Electron app for reviewing pull requests dramatically faster than Github web. Specialized for handling reviews for background agents working on multiple concurrent threads
Code review for AI native teams. Native Electron app for reviewing pull requests dramatically faster than Github web. Specialized for handling reviews for background agents working on multiple conc...
github.com
November 9, 2025 at 9:15 PM
Managing one AI agent is easy. Managing ten is.. *chaos*.
I just discovered this powertool for code reviewing the output of multiple, concurrent AI agents. A true force multiplier.
👉 github.com/areibman/bo...
I just discovered this powertool for code reviewing the output of multiple, concurrent AI agents. A true force multiplier.
👉 github.com/areibman/bo...
Founder life: 1% big vision, 99% removing roadblocks.
November 9, 2025 at 7:56 PM
Founder life: 1% big vision, 99% removing roadblocks.
A feature is not a moat. A community is. *This* community is.
November 9, 2025 at 5:47 PM
A feature is not a moat. A community is. *This* community is.
Progress report:
AI is automating itself faster than we can adopt it.
The real disruption might be how long humans keep pretending they’re the bottleneck.
→ windowsontheory.org/2025/11/04/...
AI is automating itself faster than we can adopt it.
The real disruption might be how long humans keep pretending they’re the bottleneck.
→ windowsontheory.org/2025/11/04/...
Thoughts by a non-economist on AI and economics
Crossposted on lesswrong Modern humans first emerged about 100,000 years ago. For the next 99,800 years or so, nothing happened. Well, not quite nothing. There were wars, political intrigue, the in…
windowsontheory.org
November 9, 2025 at 2:11 PM
Progress report:
AI is automating itself faster than we can adopt it.
The real disruption might be how long humans keep pretending they’re the bottleneck.
→ windowsontheory.org/2025/11/04/...
AI is automating itself faster than we can adopt it.
The real disruption might be how long humans keep pretending they’re the bottleneck.
→ windowsontheory.org/2025/11/04/...