Skip to content
Sir CompareComparisons with evidenceYouTubeTelegram
United States · USD

News

Council in PandaCode: reviewers, target scores, 25-round limit

Updated

Illustrative photo: Best AI Coding Assistants
Illustrative photo

Agents grading their own work

Agents that check their own output tend to approve it. That gap is what pushed the channel behind PandaCode to build a feature called Council. The idea is simple: separate reviewers get to challenge the self-assessment of the agents that did the work. The lowest score counts, not the average, so one weak review can block a finish.

The settings you actually control

Council exposes three knobs: a target score, a round limit, and the number of reviewers. Each reviewer gets a lens, meaning a specific angle to judge from. Every round brings fresh reviewers who never see the target or the earlier scores, while the lenses stay locked after round one. If the result lands below target, PandaCode refuses the finish and tells the lead agent to fix the blockers and call another round.

Illustrative photo: AI Tools & SaaS: Code on computer monitor (Unsplash)

What the skill manager run showed

The walkthrough follows a skill manager through its review rounds. It ended at the limit of 25 rounds with a 9 against a 9.5 target, leaving the final call to the person running it. Most rounds moved the score very little.

Getting useful UI reviews

For UI reviews, the advice is to name concrete goals and states, ask for screenshot reviews, and keep appearance separate from interaction behavior. The same applies to the security, release and factual accuracy reviews covered in the walkthrough: each lens needs something specific to check.

The cost: plan usage and mixed models

Council uses more of your plan, because every round spins up reviewers. With Panda Router on, it tries to give each reviewer a different model, which is the mixed-model advice the channel gives. The trade-off is real: more rounds mean more usage, and the skill manager example shows diminishing returns.

A score is not taste

The channel's own verdict is that Council is pretty cool, but a score is not taste and you still decide when to stop. I have not run Council myself, so treat the 9-versus-9.5 example as the channel's account, not a measured result.

Related coverage

Source: AICodeKing · coding assistants's video

Hardware to try AI at home

Affiliate links: we may earn a commission from qualifying purchases, and it never changes our data or rankings. They open Amazon's search results: prices and availability are shown there.

What changed

  • Page updated

How I put products in order

Sir Compare’s letter

Updates on AI tools & SaaS? I’ll send them your way.

Pick what you care about: I only write when something new comes out — a guide, a news story, a verified price that drops.

Step 1 of 2

What are you into?

Pick as many as you like. Pick none and you get what’s new in every section.

What you want to receive
  • Only what you pick
  • Never an empty email
  • One click to leave

Your assistant

Open full page →