Coding agent can create huge code diffs. Someone still has to read it. seendiff is a local diff tool for large changes. Track what you've seen, review in chunks, and pull the coding agent in to explain its own work.
1. Go through the painful review process, or 2. Go YOLO mode and commit without reviews.
If you care about maintenance of your codebase you will want to go through the pains review your code.
Most code diff tools today were built for the past world - but AI scale of code is different.
So I built seendiff - to scratch my own itch. It: 1. Automatically keeps track of what code you've seen in a diff 2. Keeps track of manually reviewed code blocks. 3. Integrates a code walkthrough that lets your AI walk you through the code 4. Other nifty features that matter - local global search, minimap etc.
What does your agentic code review process look like?
Report
the "what you've seen in a diff" tracking is the part that'd actually get me to use this daily. is that state local to the machine/clone, or does it sync somewhere if I review on my laptop and then pull the same branch on a different machine later
@galdayan yeah that was the original motivation - hence "seendiff".
It's designed to be private and secure so everything's local - flip side is that it won't sync across machines (ie. unless someone figure out a way to sync the sqllite across machines)
Report
@abhishek_iyer makes sense, local-only is the right tradeoff for the privacy angle. that probably rules it out for team code review then, since two people reviewing the same PR wouldn't share seen-state - more of a solo-developer tool than a team one for now?
Report
took me a second to realise it had marked a hunk seen that i basically just scrolled past @abhishek_iyer . went back to that file twice more the same night and by the third pass i wasnt trusting the marks at alll.
Report
I often loose track of what I already reviewed and what’s pending, so I get where that pain comes from :))
Report
Love that seendiff pulls the coding agent in to explain its own diffs, turning review from a chore into a guided walkthrough of large AI generated changes.
ACME.BOT
Hey Product Hunt 👋
AI code is difficult to review.
So you have two options:
1. Go through the painful review process, or
2. Go YOLO mode and commit without reviews.
If you care about maintenance of your codebase you will want to go through the pains review your code.
Most code diff tools today were built for the past world - but AI scale of code is different.
So I built seendiff - to scratch my own itch. It:
1. Automatically keeps track of what code you've seen in a diff
2. Keeps track of manually reviewed code blocks.
3. Integrates a code walkthrough that lets your AI walk you through the code
4. Other nifty features that matter - local global search, minimap etc.
Im putting it up at https://seendiff.com in case its useful to someone else.
Free. Open source.
What does your agentic code review process look like?
the "what you've seen in a diff" tracking is the part that'd actually get me to use this daily. is that state local to the machine/clone, or does it sync somewhere if I review on my laptop and then pull the same branch on a different machine later
ACME.BOT
@galdayan yeah that was the original motivation - hence "seendiff".
It's designed to be private and secure so everything's local - flip side is that it won't sync across machines (ie. unless someone figure out a way to sync the sqllite across machines)
@abhishek_iyer makes sense, local-only is the right tradeoff for the privacy angle. that probably rules it out for team code review then, since two people reviewing the same PR wouldn't share seen-state - more of a solo-developer tool than a team one for now?
took me a second to realise it had marked a hunk seen that i basically just scrolled past @abhishek_iyer . went back to that file twice more the same night and by the third pass i wasnt trusting the marks at alll.
Love that seendiff pulls the coding agent in to explain its own diffs, turning review from a chore into a guided walkthrough of large AI generated changes.