It's surprisingly difficult, and the "obvious" techniques (just do embeddings) don't really work. I wrote about it and did benchmarks here: https://joecooper.me/blog/redundancy/
I need AI that scans every PR and issue and de-dupes
11–20 of 25 posts
Re: I need AI that scans every PR and issue and de-dupes
#12It's not clear to me: is he asking us to bluid this or is he using twitter to ask it to its clawd bot?
Re: I need AI that scans every PR and issue and de-dupes
#13> How's no startup working on this? Because there's no money in trying to filter out noise that costs next to nothing to generate. It's like asking why no startup is trying to bring forum moderation to the masses.
I think anti-spam providers might disagree with that take.
Re: I need AI that scans every PR and issue and de-dupes
#14> How's no startup working on this? Because there's no money in trying to filter out noise that costs next to nothing to generate. It's like asking why no startup is trying to bring forum moderation to the masses.
> Because there's no money in trying to filter out noise that costs next to nothing to generate.
Not yet, but when there's so much more noise than signal, it'll become valuable.
Re: I need AI that scans every PR and issue and de-dupes
#15It's not clear to me: is he asking us to bluid this or is he using twitter to ask it to its clawd bot?
Or more meta: is this message from the bot itself, controlling his twitter, who got fed up because it's also merging the MRs?
Re: I need AI that scans every PR and issue and de-dupes
#16People aren't even good at this task if Stack Overflow is any indication.
Re: I need AI that scans every PR and issue and de-dupes
#17Hire a staff
He already has dozens of Codex and Claude code accounts
Re: I need AI that scans every PR and issue and de-dupes
#18> Worked all day yesterday and got like 600 commits in. It was 2700; now it's over 3100. Why? There's no reason you need to actually handle that many in a day, right? Pace yourself.
The author of this Tweet has been making waves by talking about using AI to write code without reviewing it. He wasn’t actually reading and reviewing all of that code himself.
As for the pace: His project has become extremely popular and got him a very nice position at OpenAI just today.
Re: I need AI that scans every PR and issue and de-dupes
#19People aren't even good at this task if Stack Overflow is any indication.
True; way too many duplicates have gone unrecognized.
Re: I need AI that scans every PR and issue and de-dupes
#20> Worked all day yesterday and got like 600 commits in. It was 2700; now it's over 3100. Why? There's no reason you need to actually handle that many in a day, right? Pace yourself.
The reviews must be heavily AI assisted in order to get that sort of volume in. Either way, it doesn't surprise me that this number is so high. Productivity chasing is the name of the game for AI, regardless of how sustainable or helpful this extra work done actually is
Even with heavy AI assistance, how well-reviewed can this code be? 600 commits in a day is one commit every 2 minutes for 18 hours.