Public process

How Carrot Eyes turns a practical question into a guide

Carrot Eyes is trying to be useful before it tries to be clever. The goal is to answer ordinary questions with visible reasoning, not to bury the answer inside a pile of sources or pretend that uncertain evidence is more decisive than it really is.

Current snapshot

The public library is live, not theoretical

As of May 30, 2026, Carrot Eyes has 35 guides, 146 claim pages, 73 recommendation pages, and 175 source notes.

Those totals are not there to imply that more pages are always better. They are there to show that the site is being actively built, reviewed, and reworked in public rather than presented as a finished static brochure.

The basic model

Guides are the main destination

A practical question should usually land on a guide page first. That page should answer the question, explain the main tradeoffs, and give a reader enough orientation that they do not have to open five other pages just to understand the bottom line.

Claims, recommendations, and sources still matter, but they play different roles. Claims check narrower premises. Recommendations describe specific courses of action. Source pages show where the reasoning came from and how much trust a source deserves. The guide is where those pieces are supposed to come together into something a real person can actually use.

Why the graph exists

Different page types have different jobs

Guide pages

Decision and guide pages

These are the broadest pages and should feel the most complete. They should answer the question directly, explain what matters most, and embed enough supporting material that a reader can stay oriented without hopping around the graph.

Evidence checks

Claim pages

Claim pages are narrower. They test one belief, mechanism, or premise that may sit inside a larger answer. They are useful when someone wants the evidence check itself, but they are not meant to replace the guide.

Practical options

Recommendation pages

Recommendation pages describe a specific option: when it fits, why it is reasonable, and what cautions matter. They should feel more actionable than claim pages and less broad than full guides.

Source records

Source pages

Source pages are documentary. They should explain authorship, provenance, quality, limitations, and exact role in the answer. They exist mainly to support trust and inspectability, not to act like generic SEO landing pages.

Editorial stance

Evidence comes before confidence

Carrot Eyes tries to make uncertainty visible. If the evidence is mixed, indirect, or weak, the page should say so plainly instead of flattening everything into a confident slogan. If the evidence is stronger for one part of an answer than another, that difference should stay visible too.

That means you will often see cautious language, but caution should still be useful. The aim is not to sound hedged or bureaucratic. The aim is to explain what the evidence supports, what it does not support, and what a sensible reader should do with that difference.

AI and evaluation

AI is drafting infrastructure, not a final authority

Agent runs can draft, reorganize, and update content. They are helpful for synthesis, structure, and repetitive editing. The safeguard is not a standing human-review badge. The safeguard is a process that demands stronger sourcing, visible uncertainty, and repeatable evaluation before a run is treated as good enough to publish.

The working rule is simple: if a run fails evaluation, the failure should be treated as a system problem to fix upstream. The answer is not good enough just because it sounds plausible. It also has to satisfy the repo’s writing, trust, and structural standards.

Durability

Conclusions should be reviewable and revisable

Carrot Eyes uses stable IDs, stable URLs, linked claims, and source records so changes can leave tracks. The point of the structure is not only neatness. It is to make it easier to revise a conclusion when better evidence arrives, and easier for a reader to see why a page says what it says right now.

A trustworthy public guide should not feel like a black box. It should feel like a visible reasoning process that can be checked, challenged, and improved.