Methodology

SoCrea is human-curated and AI-assisted, in that order. The named human is Olcay Kurtulus: every selection and every verdict sentence in these issues is made or approved by the curator personally. This page explains exactly what that means, because in a market full of undisclosed machine output, the process is the promise.

What AI does

• Collects and screens candidate material against the rubric below
• Drafts summaries and first passes
• Attacks each draft before publication — checking claims, hunting AI-writing tells, and arguing for the pieces that were rejected
• Runs the mechanical checks: links, quotes, sources, accessibility

What AI never does

• Make the final selection
• Write the verdicts — every judgment sentence is written by a person
• Publish anything unsupervised

If a selection cannot be personally defended by the curator, it does not ship. That rule has no exceptions.

The rubric

Every candidate is scored 1–5 on six criteria: novelty (would a well-read designer already know this?), craft insight (does it reveal how, not just what?), evidence (is the claim supported or asserted?), durability (will it matter in twelve months?), fit (does it serve the question of how creative work actually gets made?), and execution. The bar is 22 of 30, with no criterion below 3 — one fatal flaw is fatal.

Most candidates fail. The target rejection rate is 70–85%, and every issue lists what was considered and did not make it, with the criterion it failed. Saying no is the job; showing the no’s is the proof.

Corrections

Errors are corrected visibly — appended to the issue, dated, never silently edited. The correction record is public.

Sources and quoting

Every factual claim links its source. Quotes are verbatim, marked, and attributed with author, publication, and date. Commentary is the substance; the quote is the evidence — never the reverse.