Radar this week · 01 — twenty-one tools added, zero price moves

This is the first edition of radar this week, a Friday journal driven directly by the AI Tool Radar SQLite file. The rule for the column is short: every Friday we open the radar's changes table, take everything detected in the trailing seven days, and write down what actually moved. No vendor announcements we didn't catch in the scrape. No "5 hot AI tools" listicles. Just the diffs the crawler saw.

Week one is, by construction, a slightly strange edition. The radar itself only started recording change rows on the 15th of May — the day we flipped from "we have a tools table" to "we have a changes table fed by a daily diff." Anything that was already in the dataset before then doesn't have a "new_tool" event attached, even though it absolutely was new to the radar at some point. So the seven days ending 2026-05-16 look, from the changelog's point of view, like this:

detected_at   change_type    count   first row
2026-05-15    new_tool       21      05:06:04Z
                (every other change_type)  0

Twenty-one new tools, zero price moves, zero feature deltas, zero takedowns. That's the news. The rest of the post is the context behind the numbers.

What twenty-one new tools actually look like

The seed batch on the 15th was the radar admitting it had been under-counting. We were watching fourteen tools — Cursor, Aider, Cline, Continue, Lovable, v0, Bolt, StackBlitz, Replit Agent, Tabnine, Midjourney, Sora, Kling, Jasper — and then on Friday we folded in the rest of the credible field. The full list of new_tool events this week, joined to the tools table for category:

• llm-api (5) — Anthropic, Gemini, Groq, Mistral, OpenAI
• coding (5) — Claude Code, Codeium, Cody, GitHub Copilot, Windsurf
• image (4) — DALL·E 3, FLUX, Leonardo.Ai, Recraft
• video (3) — Hedra, Pika, Runway
• productivity (2) — Notion AI, Perplexity Pro
• agent-frameworks (2) — CrewAI Enterprise, LangSmith / LangGraph

The shape that surprised us when we totalled it: fifteen of the thirty-five tools we now watch sit in the coding category. AI "products" is mostly AI for shipping software, with the remainder split fairly evenly across the LLM-API layer, image generation, and video. We don't have a strong opinion about whether that ratio is "right." It's the field, as it actually files into the radar's seed list, in 2026.

The four vendors that own more than one tool in the dataset are worth naming. OpenAI has three (the API, DALL·E 3, Sora — three different surfaces of the same lab). Anthropic has two (the API and Claude Code). Cognition / fka Codeium has two — Codeium and Windsurf, the rebranded IDE — which is the cleanest example in our dataset of a single company owning both a developer-side autocomplete and an editor. StackBlitz has two: the original IDE and Bolt.new. Everyone else is one slug per company.

What didn't move

This is the part we will spend more of every future edition on, but this week the answer is "nothing." No tier landed a price cut, no tier landed a price hike, nobody quietly removed a free plan, nobody introduced a "Teams" SKU between snapshots. The changes table has zero price_change and zero feature_change rows because none of the tools we track have a second snapshot to diff against yet — the seed scrape was Friday, the daily run picked up its second swing on Saturday morning, and the diff window has barely opened.

The closest thing to news, if you squint, is the price floor the radar captured for the coding tier on Saturday afternoon. The daily scrape pulled fresh tier sheets for nine tools and stored 150 price rows across the whole dataset. Read across the coding cohort, the paid tiers look like this:

tool_slug            tier                      period     amount
stackblitz           Pro                       monthly    $18
replit-agent         Replit Core               yearly     $18  (=$1.50/mo)
continue-dev         Team                      monthly    $20
bolt-new             Pro                       monthly    $25
lovable              Pro                       monthly    $25
v0                   Team                      monthly    $30
bolt-new             Teams                     monthly    $30
tabnine              Code Assistant            monthly    $39
lovable              Business                  monthly    $50
stackblitz           Teams                     monthly    $55
tabnine              Agentic Platform          monthly    $59
jasper               Pro                       monthly    $59
replit-agent         Replit Pro                yearly     $90  (=$7.50/mo)
v0                   Business                  monthly    $100

Two things jump out. One: the entry-level paid tier across the coding cohort clusters tightly at $18–$30/month. Two: every single one of these tools offers a $0 tier. The radar counts twenty-one tools across the full dataset with at least one zero-priced plan on file — i.e. the median AI tool we track still has a free way in. That ratio will be one of the metrics this column watches over time. The day it slips under fifty percent will be worth writing about.

What's behind the journal

For anyone curious about the plumbing: the radar runs as a cron'd Python scraper that lives outside this repo, in ~/workspace/elofyn/products/ai-tool-radar/. It writes to data/radar.sqlite. The journal post you're reading was generated by opening that file with sql.js, pulling the trailing-seven-days slice of the changes table, and pasting the numbers into an MDX scaffold:

bun run journal:radar

That command runs scripts/journal-radar.mjs, which:

1. Reads radar.sqlite (read-only, via the in-process sql.js we already ship for the Dev Snippets API).
2. Pulls the trailing 7 days of changes, joined to tools for category/vendor.
3. Pulls every prices row captured in the same window.
4. Writes content/blog/radar-this-week-NN.mdx with the metrics filled in and TODO-marked narrative sections.

Edition 02 — and every edition after — will use that script as the seed. The hand-written narrative around the numbers is the part that stays human. The numbers don't.

What to expect in edition 02

By next Friday the daily scrape will have captured two snapshots for every one of the thirty-five tools in the dataset, which means the first real price_change and feature_change rows will land in the changes table. If even one of these tools quietly raises a tier, quietly removes a feature, quietly adds a "Free for students" plan, or quietly disappears, the radar sees it and edition 02 names it. If nothing moves — which would be a surprise in this category — we'll say so plainly, and the column will be three paragraphs short. We're deliberately not committing to a minimum word count. The journal is a function of the data; if the data is quiet, so is the post.

Until then: thirty-five tools watched, twenty-one of them new this week, fifteen of them in the coding category, every one of them currently on file with at least one snapshot, none of them yet caught moving a price. We'll see what next Saturday's diff brings.