Reviewing in the Age of AI

#1

Post by maxalbarello » Wed, May 06, 2026, 4:24 PM UTC

How are you guys ensuring that the code that goes into production is high quality now that the time to review code is significantly greater than the time to generate it.

There is a huge asymmetry between who is generating the code and who is reviewing it, making the review process even more painful than it used to be.

Wondering whether if instead of reviewing PRs, we should instead move towards reviewing plans so that no code is generated before at least another person approves the plan.

Once the code is generated, the users who contributed to the plan can still review it but the fact that both participated in the plan should help reduce the asymmetry.

Feels like we need a way to collaborate and iterate on plans. Would love your thoughts on this.

Re: Reviewing in the Age of AI

#2

Post by ilbert » Wed, May 06, 2026, 4:29 PM UTC

In the last couple of companies I've worked in, I've felt both overwhelmed by PRs to review and disappointed by my teammates that were just rubber-stamping my PRs

Re: Reviewing in the Age of AI

#3

Post by taeshdas » Wed, May 06, 2026, 4:32 PM UTC

In these times i would argue on probably getting the code reviewed by an ai agent which has been specifically trained on code quality, robustness and company or product specific code practices, that reduces the loan on manual reviewing by a large margin

Re: Reviewing in the Age of AI

#4

Post by nodeflare » Wed, May 06, 2026, 4:52 PM UTC

[flagged]

Re: Reviewing in the Age of AI

#5

Post by maxalbarello » Wed, May 06, 2026, 4:58 PM UTC

taeshdas wrote: post #3

In these times i would argue on probably getting the code reviewed by an ai agent which has been specifically trained on code quality, robustness and company or product specific code practices, that reduces the loan on manual reviewing by a large margin

that definitely helps but i'm wondering if that whole process (via an agent or via a human teammate) should be performed at the plan stage instead of the review stage. this should reduce the back and forth after the pr is opened (avoiding the delays and costs of running the CI several times)

Re: Reviewing in the Age of AI

#6

Post by maxalbarello » Wed, May 06, 2026, 5:03 PM UTC

ilbert wrote: post #2

In the last couple of companies I've worked in, I've felt both overwhelmed by PRs to review and disappointed by my teammates that were just rubber-stamping my PRs

wondering if the problem is the teammates rubber-stamping or the tool/workflow being outdated

Re: Reviewing in the Age of AI

#7

Post by MatrixOrigin » Thu, May 07, 2026, 2:44 AM UTC

you've identified a real structural probiem with how most teams are still operating.And we're working on it.

Re: Reviewing in the Age of AI

#8

Post by bnchrch » Thu, May 07, 2026, 12:50 PM UTC

Our teams efforts have shifted toward reviewing plans as a team.

Before code hits a humans eyes it's went through a few independant review passes from our review agents.

If it's low complexity / blast radius it gets auto merged.

If it's high complexity / blast radius it gets flagged for human review.

And funny enough our team has agreed even our human review layer has the best results if we have an agent create supplementary descriptions of the code and potential issues that we read and apply human judgement.

Reason being is that it can be very pendantic. Both in a good and bad way. Either flagging things that are non-concerns, or catching things my lazy human eye wouldnt have caught. e.g. A docstring mis-describing the actual shape of an object.

Re: Reviewing in the Age of AI

#9

Post by Digiswarm » Thu, May 07, 2026, 2:22 PM UTC

maxalbarello wrote: post #6

ilbert wrote: post #2

In the last couple of companies I've worked in, I've felt both overwhelmed by PRs to review and disappointed by my teammates that were just rubber-stamping my PRs

wondering if the problem is the teammates rubber-stamping or the tool/workflow being outdated

[flagged]

Re: Reviewing in the Age of AI

#10

Post by maxalbarello » Thu, May 07, 2026, 3:43 PM UTC

bnchrch wrote: post #8

Our teams efforts have shifted toward reviewing plans as a team. Before code hits a humans eyes it's went through a few independant review passes from our review agents. If it's low complexity / blast radius it gets auto merged. If it's high complexity / blast radius it gets flagged for human review. And funny enough our team has agreed even our human review layer has the best results if we have an agent create suppl…

interesting! how do you review each other's plans? do you just copy-paste them into chat or something else?