Why sharing your AI conversations matters more than sharing your code

For most of my career, sharing code was the primary way to teach. Show others your PRs; they learn from the changes. Read others’ PRs; you learn from theirs.

With AI tools, a new artifact emerges: the conversation. The AI conversation captures the thinking — the questions asked, the wrong attempts, the corrections, the eventual right answer.

I’ve started sharing my AI conversations. The teaching value exceeds my PRs.

What conversations show that PRs don’t

A PR shows the destination. Files changed. The final state of the code.

A conversation shows the journey:

What I was confused about
Where I was wrong
How I corrected
What I learned along the way

For someone learning, this is more valuable than the destination. The destination tells them what to type; the journey tells them how to think.

A specific example

A PR I merged recently was 40 lines of code change. From the diff, you’d see:

Some refactoring of the payment flow
A new error handling path
Updated tests

Reading the diff takes 5 minutes. The reader understands what changed. Useful but limited.

The conversation that produced the PR was 25 messages over 90 minutes. It included:

My initial misunderstanding of why the bug was happening
AI’s suggestion that turned out to be wrong
My realization of what was actually broken
A few false starts on the fix
The eventual approach
Discussion of edge cases
The decision to add specific tests

Reading the conversation takes 15 minutes. The reader understands not just what changed but why, including the false starts that shaped the final decision.

For a junior engineer, the conversation is far more valuable than the diff.

Sharing AI conversations requires:

Cleaning up the conversation (removing false starts that aren’t pedagogically useful)
Annotating where appropriate (“here’s where I realized X”)
Considering privacy (some prompts include code that shouldn’t be shared)
Choosing the right venue (Slack, internal docs, blog)

This is more work than sharing a PR. The PR exists; the conversation needs curation.

Worth doing for high-leverage moments — bug investigations that taught something, design decisions that involved tradeoffs, learning new technologies.

Privacy considerations

Some conversations include sensitive content:

Prompts that include code from the codebase
AI responses that reveal internal architecture
Personal frustrations or doubts expressed during work

For sharing internally with the team, most of this is fine. For external sharing (blog, conference talks), redact carefully.

A pattern: share the conversation with code references but not actual code. “I asked about implementing X in our payment service” rather than “[paste of payment service code].”

The categories I find worth sharing:

Investigations that taught something non-obvious. Bug hunts where the cause surprised me. The conversation captures the surprise; the diff just shows the fix.

Design decisions with non-obvious tradeoffs. When I considered three approaches and picked one, the reasoning is in the conversation. The PR just shows the chosen one.

Tools or techniques I learned. First time using a new library, framework, or pattern. The conversation captures the learning curve.

Failed AI interactions. When AI got something wrong and I corrected. Useful for showing AI’s failure modes.

Successful AI interactions. When AI was particularly helpful. Useful for showing what works.

Some categories I keep private:

Conversations with embedded credentials or secrets. Even with redaction, risk of leaving something.

Conversations about sensitive personnel or business matters. Just because the AI is a quasi-confidant doesn’t mean the content should be public.

Repetitive routine conversations. “How do I add a button?” doesn’t teach much.

Personal frustration content. When I was venting to AI about a frustrating bug. Doesn’t help others.

The team dynamic

Sharing AI conversations changes team dynamics:

More transparent thinking. Engineers see how their colleagues actually work, not just what they ship.

Faster knowledge transfer. New team members can read accumulated AI conversations to learn the codebase faster.

Mentor at scale. Senior engineers’ conversations teach implicitly. Their thinking becomes available without 1:1 time.

Less imposter syndrome. Watching senior engineers’ AI conversations shows they ask “dumb” questions too. The illusion of “they always know” dissolves.

These are genuinely positive effects. Teams that adopt sharing notice morale and learning improvements.

The format question

How to share matters:

Inline comments in PRs. “This was the discussion that led to this approach.” Quick context.

Internal blog posts. Longer-form sharing for important learnings.

Slack channels. A dedicated channel for “interesting AI moments.” Lower bar; higher volume.

Pair sessions. Sharing AI conversations live with a teammate. Highest bandwidth; highest value.

Different formats serve different purposes. Use what fits.

A call to action

For engineers using AI tools regularly:

When you have a conversation that taught you something, share it. Not every one — that’s overload. The ones that include real learning.

The teaching value compounds. Your team learns. Other teams learn. The community of engineers using AI tools learns from each other.

Pre-AI, learning happened mostly through code review and pair programming. Post-AI, conversations are a new artifact that’s underused for learning.

The artifact is sitting there. The teaching value is sitting there. The friction to share is small. The compound benefit across teams and communities is large.

Try sharing one conversation this week. See what reactions you get. The first time is awkward. The second is easier. By the tenth, it’s habit.

Closing

Sharing AI conversations is one of those things that’s clearly valuable but not yet common. It will become normal over the next few years.

Engineers who start now build a habit and reap the benefits. Engineers who wait will eventually do it too, but later. The compounding favors starting earlier.

Your AI conversations are teaching artifacts. Treat them as such. Curate them. Share them. The investment is small; the return compounds across your team and your career.