Paul
@PaulSangleF
Followers
1K
Following
15K
Media
63
Statuses
602
Founder at https://t.co/UUeydPyIis (YC X25) AI code reviews for complex codebases
San Francisco
Joined July 2013
currently working on a complex #rsyslog feature request. Have a look at a set of local pre-checks. New is @cubic_dev_, which runs as a final local validator. Note how it caught issues even after all the other checks. Only code that passes local checks goes to PR stage. #aicode
1
2
2
So great to see people starting to build on top of cubic!
@usebrief @helicone_ai @langfuse Day 3 of release week: Bring Your Own MCP. Brief ships new integrations every week. But you have internal tools and more great apps pop up every day. Now you can register your own MCP servers. Customer data APIs. Analytics pipelines. Internal dashboards. Whatever your team
1
0
3
We have amazing list of judges! if you're in sf for feb 7 register now 🫡
Introducing Better Hack At Better Auth, our goal is to democratize access to high quality software that enables others to build category defining products and companies. We’re finding software is often weighed down by legacy decisions, bloated features, and workflows that no
7
2
73
This really is a great tool. We currently use it for core #rsyslog, but I plan to extend it's use to close siblings like #librelp or #liblognorm. (If only there were a bit more time...).
@cubic_dev_ went from nothing to one of the most popular AI code reviewers for open source in less than a year. We're coming for BugBot next. If you have an open source project, sign up and you'll automatically be on the open source plan.
0
1
1
Just cancelled Greptile for @cubic_dev — why is no one talking about this? → Local CLI that runs BEFORE you push (my AI agents use it automatically now) → Auto-generates custom review agents from YOUR actual code patterns → Background agents that auto-fix issues via Claude
2
3
5
@thepanta82 @greptile @Cursor @Claude @cubic_dev_ nobody talks about @cubic_dev_, but: - the bot is way more polished then greptile. Greptile has tons of small annoying glitches. - the review quality is much better than greptile and slightly better then bugbot. ofc that's very subjective, but that's the sentiment our team
1
1
3
Did I already tell you we evaluated @greptile @cursor bot, @claude code GH action and @cubic_dev_, and the later is a clear winner.
1
2
7
0
2
5
"After the first couple of PRs, we knew that we didn't want to remove cubic." - Bu, CTO at @resend Really proud to share this one. Resend processes billions of emails every month. Their bar for quality is incredibly high, and their product craft shows it. The fact that teams
0
1
13
"After the first couple of PRs, we knew that we didn't want to remove cubic." - Bu, CTO at @resend Really proud to share this one. Resend processes billions of emails every month. Their bar for quality is incredibly high, and their product craft shows it. The fact that teams
0
1
13
@cubic_dev_ is so much better than any other ai code review out there and it's not even close. in a word: tasteful.
1
2
3
That's pretty cool, you can ask @cubic_dev_ questions right from your PR and get a great response. Here in action with a pretty complex #rsyslog PR with a weird threading issue.
1
2
3
1
1
2
That's how it looks if @cubic_dev_ cli works on a WiP #rsyslog PR within another AI that iterates based on cubic findings. :-)
1
2
2
Day 5 of Launch Week: Today we’re launching the cubic CLI. `cubic review` runs on your local changes: - catches issues before you push - syncs your repo rules - generates fix prompts for Cursor/Copilot Install: pnpm add -g @cubic-dev-ai/cli && cubic
5
2
15
Day 4 of Launch Week: Your AI review settings should be version-controlled. Introducing cubic.yaml: → All settings in one file at repo root → Every change goes through code review → IDE validation in VS Code, Cursor, JetBrains Already using the UI? Download your
1
1
13
Right now playing with @cubic_dev_ #CLI on a pretty complex #rsyslog PR for #wolfssl integration. The tools looks very promising. It also saves a lot of CI cycles as I do not need to start a full CI run just to do the check.
1
2
3
