Joey Kudish 🇺🇦🏳️🌈
@jkudish.com
I Build Software That Works. Let’s Build Yours. TetherMobile.com founder. Laravel + AI. Newsletter: jkudish.com/newsletter
Pinned
Joey Kudish 🇺🇦🏳️🌈
@jkudish.com
· Nov 24
Human in the Loop Newsletter
Subscribe to Human in the Loop for practical insights on AI-augmented coding, productivity hacks, and how to use AI as your development partner, not replacement.
jkudish.com
Tomorrow's newsletter: The 80/20 Rule Didn't Go Anywhere
AI didn't eliminate the hard work. It just revealed where it actually lives.
The easy stuff was never the real work - we were just busy doing it. Now AI compresses it, and we hit the hard parts faster.
jkudish.com/newsletter?u...
AI didn't eliminate the hard work. It just revealed where it actually lives.
The easy stuff was never the real work - we were just busy doing it. Now AI compresses it, and we hit the hard parts faster.
jkudish.com/newsletter?u...
i don't think AI replaces senior developers.
i think it makes the gap between senior and junior wider than it's ever been.
seniors know what to ask for, what to reject, and when the AI is confidently wrong. that's not a skill you can shortcut.
i think it makes the gap between senior and junior wider than it's ever been.
seniors know what to ask for, what to reject, and when the AI is confidently wrong. that's not a skill you can shortcut.
February 12, 2026 at 7:01 PM
i don't think AI replaces senior developers.
i think it makes the gap between senior and junior wider than it's ever been.
seniors know what to ask for, what to reject, and when the AI is confidently wrong. that's not a skill you can shortcut.
i think it makes the gap between senior and junior wider than it's ever been.
seniors know what to ask for, what to reject, and when the AI is confidently wrong. that's not a skill you can shortcut.
the developers mass-adopting AI right now will split into two camps within a year.
those who built systems around it. and those who just used it as a faster keyboard.
the gap will be enormous.
those who built systems around it. and those who just used it as a faster keyboard.
the gap will be enormous.
February 12, 2026 at 6:03 PM
the developers mass-adopting AI right now will split into two camps within a year.
those who built systems around it. and those who just used it as a faster keyboard.
the gap will be enormous.
those who built systems around it. and those who just used it as a faster keyboard.
the gap will be enormous.
if your AI can't read your past decisions, it's just a fancy autocomplete.
CLAUDE.md, specs, commit history, architecture docs. that's not documentation overhead. that's your compound interest.
CLAUDE.md, specs, commit history, architecture docs. that's not documentation overhead. that's your compound interest.
February 12, 2026 at 4:03 PM
if your AI can't read your past decisions, it's just a fancy autocomplete.
CLAUDE.md, specs, commit history, architecture docs. that's not documentation overhead. that's your compound interest.
CLAUDE.md, specs, commit history, architecture docs. that's not documentation overhead. that's your compound interest.
building with AI every day outweighs reading about AI every day.
the anxiety about the future dropped and got replaced by genuine excitement for what's possible.
the anxiety about the future dropped and got replaced by genuine excitement for what's possible.
February 12, 2026 at 2:58 PM
building with AI every day outweighs reading about AI every day.
the anxiety about the future dropped and got replaced by genuine excitement for what's possible.
the anxiety about the future dropped and got replaced by genuine excitement for what's possible.
the gap between "I have an idea" and "it's live" has never been smaller.
which means the gap between you and your competition has never been smaller either.
taste, positioning, and who you build for matter more than ever. the code is the easy part now.
which means the gap between you and your competition has never been smaller either.
taste, positioning, and who you build for matter more than ever. the code is the easy part now.
February 12, 2026 at 2:57 AM
the gap between "I have an idea" and "it's live" has never been smaller.
which means the gap between you and your competition has never been smaller either.
taste, positioning, and who you build for matter more than ever. the code is the easy part now.
which means the gap between you and your competition has never been smaller either.
taste, positioning, and who you build for matter more than ever. the code is the easy part now.
your AI workflow is only as good as what it remembers tomorrow.
most people start every session from scratch. same mistakes, same corrections, same context loading.
build the memory layer. that's where the leverage is.
most people start every session from scratch. same mistakes, same corrections, same context loading.
build the memory layer. that's where the leverage is.
February 12, 2026 at 1:02 AM
your AI workflow is only as good as what it remembers tomorrow.
most people start every session from scratch. same mistakes, same corrections, same context loading.
build the memory layer. that's where the leverage is.
most people start every session from scratch. same mistakes, same corrections, same context loading.
build the memory layer. that's where the leverage is.
i use AI to think slower.
spec conversations, architecture reviews, decision journals. each one feeds the next session. the AI gets sharper because my inputs get sharper.
shipping fast, at this point, is table stakes. compounding context is what I focus on now.
spec conversations, architecture reviews, decision journals. each one feeds the next session. the AI gets sharper because my inputs get sharper.
shipping fast, at this point, is table stakes. compounding context is what I focus on now.
February 11, 2026 at 7:56 PM
i use AI to think slower.
spec conversations, architecture reviews, decision journals. each one feeds the next session. the AI gets sharper because my inputs get sharper.
shipping fast, at this point, is table stakes. compounding context is what I focus on now.
spec conversations, architecture reviews, decision journals. each one feeds the next session. the AI gets sharper because my inputs get sharper.
shipping fast, at this point, is table stakes. compounding context is what I focus on now.
everyone's talking about which AI model is best.
the developers actually shipping are using all of them. different models, different tasks, swap as needed.
the answer was never "pick one." it's orchestrate many.
the developers actually shipping are using all of them. different models, different tasks, swap as needed.
the answer was never "pick one." it's orchestrate many.
February 11, 2026 at 6:56 PM
everyone's talking about which AI model is best.
the developers actually shipping are using all of them. different models, different tasks, swap as needed.
the answer was never "pick one." it's orchestrate many.
the developers actually shipping are using all of them. different models, different tasks, swap as needed.
the answer was never "pick one." it's orchestrate many.
shipping code faster with AI doesn't mean you ship better products faster.
feature bloat is easier than ever.
decision paralysis is worse.
scope creep happens in minutes instead of weeks.
speed without direction is just expensive chaos.
feature bloat is easier than ever.
decision paralysis is worse.
scope creep happens in minutes instead of weeks.
speed without direction is just expensive chaos.
February 11, 2026 at 5:59 PM
shipping code faster with AI doesn't mean you ship better products faster.
feature bloat is easier than ever.
decision paralysis is worse.
scope creep happens in minutes instead of weeks.
speed without direction is just expensive chaos.
feature bloat is easier than ever.
decision paralysis is worse.
scope creep happens in minutes instead of weeks.
speed without direction is just expensive chaos.
I tried the "fully autonomous AI agent fleet" thing. Gave it real tasks, real codebases.
Too sloppy. Too many assumptions. Not enough awareness of what makes code maintainable.
The path forward isn't "turn it loose and hope." It's building up the system with guardrails first.
Too sloppy. Too many assumptions. Not enough awareness of what makes code maintainable.
The path forward isn't "turn it loose and hope." It's building up the system with guardrails first.
February 10, 2026 at 2:58 PM
I tried the "fully autonomous AI agent fleet" thing. Gave it real tasks, real codebases.
Too sloppy. Too many assumptions. Not enough awareness of what makes code maintainable.
The path forward isn't "turn it loose and hope." It's building up the system with guardrails first.
Too sloppy. Too many assumptions. Not enough awareness of what makes code maintainable.
The path forward isn't "turn it loose and hope." It's building up the system with guardrails first.
New issue of Human in the Loop tonight.
What actually works in my AI setup. What doesn't yet. Two new Claude Code features that push things forward. And why I think the direction matters more than the destination.
jkudish.com/newsletter
What actually works in my AI setup. What doesn't yet. Two new Claude Code features that push things forward. And why I think the direction matters more than the destination.
jkudish.com/newsletter
Human in the Loop Newsletter
Subscribe to Human in the Loop for practical insights on AI-augmented coding, productivity hacks, and how to use AI as your development partner, not replacement.
jkudish.com
February 10, 2026 at 12:56 PM
New issue of Human in the Loop tonight.
What actually works in my AI setup. What doesn't yet. Two new Claude Code features that push things forward. And why I think the direction matters more than the destination.
jkudish.com/newsletter
What actually works in my AI setup. What doesn't yet. Two new Claude Code features that push things forward. And why I think the direction matters more than the destination.
jkudish.com/newsletter
The shift isn't "me and my AI assistant" anymore. It's "me and my agents." Plural.
Each one focused. Each one disposable. The human stays in the loop at every decision point.
Writing about this tonight in my newsletter
jkudish.com/newsletter
Each one focused. Each one disposable. The human stays in the loop at every decision point.
Writing about this tonight in my newsletter
jkudish.com/newsletter
Human in the Loop Newsletter
Subscribe to Human in the Loop for practical insights on AI-augmented coding, productivity hacks, and how to use AI as your development partner, not replacement.
jkudish.com
February 10, 2026 at 10:57 AM
The shift isn't "me and my AI assistant" anymore. It's "me and my agents." Plural.
Each one focused. Each one disposable. The human stays in the loop at every decision point.
Writing about this tonight in my newsletter
jkudish.com/newsletter
Each one focused. Each one disposable. The human stays in the loop at every decision point.
Writing about this tonight in my newsletter
jkudish.com/newsletter
Every bug should become a permanent lesson. Every code review should update the defaults.
If your AI forgets what it learned yesterday, you're not engineering. You're just prompting.
If your AI forgets what it learned yesterday, you're not engineering. You're just prompting.
February 5, 2026 at 2:58 PM
Every bug should become a permanent lesson. Every code review should update the defaults.
If your AI forgets what it learned yesterday, you're not engineering. You're just prompting.
If your AI forgets what it learned yesterday, you're not engineering. You're just prompting.
Your AI chat history is a goldmine you're ignoring.
Export it. Parse into markdown. Organize by project.
Decision log, prompt library, thinking evolution — all sitting in your export.
Export it. Parse into markdown. Organize by project.
Decision log, prompt library, thinking evolution — all sitting in your export.
February 5, 2026 at 2:56 AM
Your AI chat history is a goldmine you're ignoring.
Export it. Parse into markdown. Organize by project.
Decision log, prompt library, thinking evolution — all sitting in your export.
Export it. Parse into markdown. Organize by project.
Decision log, prompt library, thinking evolution — all sitting in your export.
UX design assumed humans click buttons.
AX design assumes agents take actions. Human states intent, agent executes, human reviews.
The whole field is about to flip.
AX design assumes agents take actions. Human states intent, agent executes, human reviews.
The whole field is about to flip.
February 5, 2026 at 12:57 AM
UX design assumed humans click buttons.
AX design assumes agents take actions. Human states intent, agent executes, human reviews.
The whole field is about to flip.
AX design assumes agents take actions. Human states intent, agent executes, human reviews.
The whole field is about to flip.
Building an AI product?
Don't build ON one model. Orchestrate many.
Be the conductor, not an instrument.
Don't build ON one model. Orchestrate many.
Be the conductor, not an instrument.
February 4, 2026 at 7:56 PM
Building an AI product?
Don't build ON one model. Orchestrate many.
Be the conductor, not an instrument.
Don't build ON one model. Orchestrate many.
Be the conductor, not an instrument.
Vibe coding: reactive parallelism, no learning loop.
Compound engineering: intentional loop where each session teaches the next.
Same tools. Completely different outcomes.
Compound engineering: intentional loop where each session teaches the next.
Same tools. Completely different outcomes.
February 3, 2026 at 2:56 PM
Vibe coding: reactive parallelism, no learning loop.
Compound engineering: intentional loop where each session teaches the next.
Same tools. Completely different outcomes.
Compound engineering: intentional loop where each session teaches the next.
Same tools. Completely different outcomes.
meetings and voice memos are more productive than they've ever been.
record it, transcribe it, process it. decisions get captured with their reasoning. action items route to the right project. status updates sync automatically.
talking is faster than typing. let AI handle the rest.
record it, transcribe it, process it. decisions get captured with their reasoning. action items route to the right project. status updates sync automatically.
talking is faster than typing. let AI handle the rest.
February 2, 2026 at 8:01 PM
meetings and voice memos are more productive than they've ever been.
record it, transcribe it, process it. decisions get captured with their reasoning. action items route to the right project. status updates sync automatically.
talking is faster than typing. let AI handle the rest.
record it, transcribe it, process it. decisions get captured with their reasoning. action items route to the right project. status updates sync automatically.
talking is faster than typing. let AI handle the rest.
Boris Cherny (creator of Claude Code) just shared his top 10 tips.
Tip #4: "If you do something more than once a day, turn it into a skill or command."
Sitting at 46 commands, 44 skills, 38 sub-agents now. But the number isn't the point.
Tip #4: "If you do something more than once a day, turn it into a skill or command."
Sitting at 46 commands, 44 skills, 38 sub-agents now. But the number isn't the point.
February 2, 2026 at 7:20 PM
Boris Cherny (creator of Claude Code) just shared his top 10 tips.
Tip #4: "If you do something more than once a day, turn it into a skill or command."
Sitting at 46 commands, 44 skills, 38 sub-agents now. But the number isn't the point.
Tip #4: "If you do something more than once a day, turn it into a skill or command."
Sitting at 46 commands, 44 skills, 38 sub-agents now. But the number isn't the point.
The old friction of writing code was a gift we didn't appreciate.
It forced you to be deliberate. You couldn't start 5 features at once because you physically couldn't build them fast enough.
Now that friction is gone. And if we're not intentional about replacing it, we're in trouble.
It forced you to be deliberate. You couldn't start 5 features at once because you physically couldn't build them fast enough.
Now that friction is gone. And if we're not intentional about replacing it, we're in trouble.
February 1, 2026 at 6:01 PM
The old friction of writing code was a gift we didn't appreciate.
It forced you to be deliberate. You couldn't start 5 features at once because you physically couldn't build them fast enough.
Now that friction is gone. And if we're not intentional about replacing it, we're in trouble.
It forced you to be deliberate. You couldn't start 5 features at once because you physically couldn't build them fast enough.
Now that friction is gone. And if we're not intentional about replacing it, we're in trouble.
Claude Code has a lesser-known feature called Rules.
You can add .md files to .claude/rules/ and scope them to specific file paths with glob patterns. They load automatically when Claude touches matching files.
You can add .md files to .claude/rules/ and scope them to specific file paths with glob patterns. They load automatically when Claude touches matching files.
February 1, 2026 at 11:27 AM
Claude Code has a lesser-known feature called Rules.
You can add .md files to .claude/rules/ and scope them to specific file paths with glob patterns. They load automatically when Claude touches matching files.
You can add .md files to .claude/rules/ and scope them to specific file paths with glob patterns. They load automatically when Claude touches matching files.
If you're letting AI write code, you need automated verification.
I set up PHPStan on every Laravel project now. Static analysis, type checks, test suites.
Let the machines check the machines.
I set up PHPStan on every Laravel project now. Static analysis, type checks, test suites.
Let the machines check the machines.
February 1, 2026 at 3:02 AM
If you're letting AI write code, you need automated verification.
I set up PHPStan on every Laravel project now. Static analysis, type checks, test suites.
Let the machines check the machines.
I set up PHPStan on every Laravel project now. Static analysis, type checks, test suites.
Let the machines check the machines.
Laravel AI SDK is coming and I can't wait to dig in.
Laravel having a first-party AI story is a big deal. The ecosystem needed this.
Laravel having a first-party AI story is a big deal. The ecosystem needed this.
January 31, 2026 at 8:02 PM
Laravel AI SDK is coming and I can't wait to dig in.
Laravel having a first-party AI story is a big deal. The ecosystem needed this.
Laravel having a first-party AI story is a big deal. The ecosystem needed this.
What's the one thing you automated this month that you'll never do manually again?
January 31, 2026 at 6:58 PM
What's the one thing you automated this month that you'll never do manually again?
I gave @OpenClaw too much autonomy too fast.
It wrote 30k lines of code for a personal software system. Got sloppy quick. The kind of sloppy where you open a file and go "what is this doing."
But I learned more from those 30k lines than I expected.
It wrote 30k lines of code for a personal software system. Got sloppy quick. The kind of sloppy where you open a file and go "what is this doing."
But I learned more from those 30k lines than I expected.
January 31, 2026 at 4:33 PM
I gave @OpenClaw too much autonomy too fast.
It wrote 30k lines of code for a personal software system. Got sloppy quick. The kind of sloppy where you open a file and go "what is this doing."
But I learned more from those 30k lines than I expected.
It wrote 30k lines of code for a personal software system. Got sloppy quick. The kind of sloppy where you open a file and go "what is this doing."
But I learned more from those 30k lines than I expected.