Radar this week · 02 — nothing moved on the wire, one endpoint moved on ours

This is edition 02 of the radar journal — and it lands three weeks after edition 01, not seven days. The pause was deliberate. Edition 01 promised that by the following Friday the daily scrape would have captured two snapshots per tool and the first real price_change rows would appear in the changes table. Two Fridays later we still had 45 snapshots across 35 tools — not enough density to call a diff "the diff" — so the column waited. Silence is data; premature silence is noise. We waited until we could say something honest.

What moved

Nothing. Zero change rows landed across the trailing seven days. No new tools, no price moves, no feature deltas, no takedowns. The slice of changes that this post is supposed to narrate looks exactly like this:

detected_at   change_type    count
2026-05-28..  (any)          0

That is the truthful answer to "what did the radar see this week." The cron is running, the scraper is touching the vendor pages, the snapshots are landing — but every successive snapshot on the 35 tools we currently track hashed inside the tolerance window of the previous one. Cursor didn't move. Codeium didn't move. Notion AI didn't move. Anthropic, OpenAI, and Gemini haven't quietly retiered. We will absolutely write about it the week one of them does. This week, they didn't.

Why this edition is short

Edition 01 said the column is a function of the data. The data this week is a flat line, so the column is a flat line. We are not going to manufacture commentary about tools we didn't see move. If you want shape from the standing dataset rather than from the diff: the median paid-tier price across the coding cohort is still in the $18–$30/month band we surfaced in edition 01, every coding tool we track still has a $0 entry plan on file, and the four multi-product vendors (OpenAI, Anthropic, Cognition, StackBlitz) still own two or three slugs each. The snapshot density that would let us call out a new shape isn't there yet.

What shipped on our end

The interesting news this week isn't on the radar — it is on the adjacent Dev Snippets API. We shipped the seventh endpoint: POST /v1/penpot-to-react. It takes a .penpot design file as a binary body and returns either a single JSX component (mode=single) or a zipped project (mode=project) with components/*.jsx, an assets/ folder, and a tokens.ts extracted from the design's color and typography variables. The CLI that backs the endpoint is a single Python file in-tree; the HTTP surface is the thin wrapper that lets you do the same conversion without installing anything beyond an X-API-Key header.

Two reasons it gets a paragraph in the radar journal rather than just an entry on the Dev API changelog. One: we tagged the design-to-code category in edition 01 as "watched but quiet," and now we are shipping into it ourselves. That is exactly the kind of disclosure the column owes its readers — Elofyn-the-tracker and Elofyn-the-tool-maker are the same company, and when the second ships into a category the first is watching, the first should say so. Two: the new endpoint shows up on the homepage "Endpoints live" metric (the number stepped from 6 to 7 this week) and in the "what we ship" section's column-04 list. If you noticed the metric move and wondered which endpoint joined, that's the one. The product landing page carries the 60-second drop-and-render demo, and the Dev Snippets doc anchor has curl, node, python, and go samples for both modes.

What the dataset looks like right now

• Tools tracked: 35
• Cumulative snapshots: 45
• Price rows on file: 150
• Feature rows on file: 638

The 45-snapshot count is the line we're watching most closely inside the studio. It needs to climb past 70 — two clean snapshots per tool — before the changes table can produce meaningful diffs at the weekly cadence the column was designed for. The cron itself is healthy; what's lagging is the scraper's successful completion rate on the half-dozen vendor pages that throttle or rotate their HTML aggressively. Next week's plumbing on the radar side is exactly that: harden the scrape so the snapshot count compounds correctly between Fridays.

What to expect in edition 03

We commit to one of two posts the Friday after next. Either the snapshot density crosses the threshold and edition 03 names the first real price_change row the radar sees, or it doesn't, and edition 03 explains the bug we found and fixed in the scraper. Either way the post is short and the post is honest. We are not going to fill the column with takes the data can't support — that was the whole point of starting it.

Until then: 35 tools watched, 0 of them moving this week, 45 snapshots on file, one new endpoint on the developer side. The live radar is the running source of truth between editions, and the Dev Snippets API is where the seventh endpoint lives if you want to try the conversion yourself.