smb06

The thing that could change is interpreting "the whole thing as generated by Claude"

I have to keep reminding it to not write a wall of text. It follows my instruction for a bit and then reverts to type. I have the instruction saved in agents.md but still have to keep reminding it.

My company pretty much exclusively uses Sol and Luna

Congrats to the CodeRabbit team. Key points:

1. GitHub is on pace for 14x more commits this year, and at companies in the 90th percentile of coding-agent adoption, agents now open 35% of PRs. The backlog has moved from tickets to PRs.

2. Diff-centric PR review breaks down at agent scale. Reviewing a 3,000-line agent PR line by line needs a whole new way of explainig what the agent wrote.

3. $10M committed over the next 12 months to keep AI code review free for open source projects and maintainers.

"Hi" isn't a "move"

Agree. There are some things I can do that I couldn't do before at all (graphics design, etc.) and loads of things I can do much faster than before (coding, data analytics, deep research etc.)

AI companies getting involved makes it newsworthy

Assuming this will allow more tool connectivity

I'm still on the $20 Anthropic plan but I really should move to the $100 plan. I guess I haven't been using Claude Cowork/Code as much in my personal subscription (on the Enterprise plan for work though)

Isn't a "spending commitment" just another form of debt?

Glad to see companies owning responsibility and putting out a message without corporate PR spin

As one of my colleagues said - "only took them 5 years"

Depends on how you define agents. I use CodeRabbit all the time but it has a swarm of dozens of sub-agents. So am i using 1 agent or 10+?

Not in my experience. Claude definitely gets annoyed when I am annoyed at it because it is losing context and starts going around in circles forcing me to start a new session.

Glad to hear it is not too terrible where you are

Especially so in a world where coding agents will carve up your project/product into smaller chunks of code without explicit guidance from you.

Why are we bringing up articles from a quarter century ago?

I do believe they were trying to use CodeRabbit to review their re-written code. https://www.coderabbit.ai/blog/ai-can-migrate-your-entire-co...

Discussed here: https://news.ycombinator.com/item?id=48947560

I feel like I've been hearing about atmosphere on Earth-like planets for a while. Or maybe I've missed some nuance.

It is a separation-of-duties argument for code written by AI agents and also reviewed by the same AI agents.

If you ask Agent1 to write code, the PR diff coming out of it should really be reviewed by Agent2 (or 3 or whatever), the reviewer should be a different system than the code gen agent.

The core claim is that a model is a poor judge of its own work — it shares the same priors and blind spots that produced a bug, so self-review trends toward a shallow review rather than catching blind spots.

The system that writes the code shouldn't also be the thing that signs off that it's safe to ship.

The mid-terms could really become about whether AI as a technology should be in the hands of the public/society or left to tech companies

My rent has gone up by 15% over the last two years and it is reflective of the increase in salaries in AI startups.

TL;DR -- more configurability with Claude Code, better out-of-the-box experience with Codex.

That tracks with the rest of their philosophy too

Their code reviewer was previously called Diamond

Do you use Copilot for coding and then also Copilot for reviewing? Or are you using some other coding agent and Copilot only for PR reviews?

Just tried it out. Very cool but would be helpful to also be able to upload .svg files

Part of Youtube premium or ad supported?

I read their pitch as trying out multiple agents to do the same task and then pick your favorite approach

On the flip side, our SaaS runs primarily on GCP so our users are fine. But our billing and subscription system runs on AWS so no one can pay us today.