ToolTickerBeta

Snapshot #00518 AUG 2026 · 00:00 UTC19 tools·13 indexed·6 watchlist

History buildingNext snapshot · 06:00 UTC

Methodology

ToolTicker measures publicly observable activity — not revenue, private usage or market share.

● Public signals● Daily snapshots● No paid ranking

01Power Score

Claude CodeAnthropic

Current #1 · coverage 30+30+20

0.0Power
Official weights30 / 30 / 20 / 20 + booster · v5
SIZE
88.5× 37.5%0.0
0.0 pts
GROWTH
62.5× 37.5%0.0
0.0 pts
MENTIONS
97.4× 25.0%0.0
0.0 pts
SEARCH
— no signal today (hole)

Coverage 80/100 — SIZE + GROWTH + MENTIONS. Missing components are holes, never zeros: the displayed weights renormalize over the components actually present, and the total (base + momentum booster, when above scale) reconciles exactly to the Power Score above.

The formula is public, the components are named, and the weights were ratified on 2026-08-17 (method prototype-v5): 30% size + 30% growth + 20% mentions + 20% search plus a momentum booster so a genuinely-hot new entrant can rise before it has installed-scale, while an all-round dominant leader keeps the lead. Every score on the site expands to this exact breakdown — click any Power Score on a ranking page.

02Components

Size

30%

Current observable adoption

npm downloadsGitHub starsVS Code installsDiscord membersHF downloads
How is this calculated?

The tool's most representative public usage counter — whichever of npm, GitHub, HuggingFace, the VS Code marketplace or Discord best reflects its adoption. The counter actually used and its raw value are recorded per tool, per day (visible on every tool page).

Growth

30%

↗ 30-day velocity

bootstrapped from daily snapshots
How is this calculated?

The 30-day delta of the size series. Vendors provide history only for npm; for every other tool, growth is bootstrapped from the daily snapshots this site collects — it becomes real for all tools after ~30 days of history.

Mentions

20%

◉ public discussion

Hacker NewsReddit (14-day subreddit volume)X excluded
How is this calculated?

Combined mention signal: Hacker News story mentions (quoted, disambiguated queries) + Reddit post volume over the trailing 14 days, counted per tool in its own subreddit via the Arctic-Shift mirror (native Reddit JSON is OAuth-walled; each subreddit verified to exist). X is excluded by owner decision (metered per-result cost). Two discourse surfaces, percentile-ranked together.

Search

20%

⌕ search demand

Google Trends90-day mean
How is this calculated?

90-day mean of the Google Trends interest index (unofficial API; disambiguated per-tool queries). Rate-limit holes are recorded as holes, never filled with zeros.

Momentum booster · method v5

A tool’s Power is its weighted component sum — 30 size + 30 growth + 20 mentions + 20 search, renormalized over the components present — plus a bounded kick proportional to how far its current heat runs ahead of its installed scale:

booster = clamp((momentum − size) × 0.12, 0, 12)

momentum = mean(growth, mentions, search) percentiles · capped at +12

The point: a genuinely-hot new entrant can overtake a big-but-cold incumbent and reach the top before it has installed-scale — a brand-new model blowing up on HN and Google Trends is no longer permanently anchored by its (still-tiny) download count. The kick is capped and fades as the spike normalizes; an all-round dominant leader (hot and big) keeps #1, and a cold incumbent with no momentum gets nothing. This replaced the 2026-08-14 45/25/15/15 freeze.

03Coverage

Full index

  • Has observable usage data
  • Competes in the global ranking
  • Full Power Score (30/30/20/20 + booster)

Twelve of eighteen tracked tools qualify today.

Watchlist

  • No public usage counter (verified)
  • Scored on attention only — 60 search + 40 mentions
  • Ranked among themselves, never merged into the index

A partial score must not compete with full-signal tools.

Normalization — different counters, comparable percentiles

Raw signal

GitHub Copilot74.4Mall-time installs
Claude Code19.7Mweekly downloads
Midjourney18.7Mdiscord members
Codex13.7Mweekly downloads

Percentile (log-scale, one global cohort)

GitHub Copilot
96.2
Claude Code
88.5
Midjourney
80.8
Codex
73.1

Log-scale percentiles make wildly different counter units (74.4M installs vs 15.1M weekly downloads) comparable before weighting.

Power Score

computed in one global cohort of all 18 tools

Category rank

position among Coding (or Image) tools

Power is global; rank is category-relative. Cross-category Power scores are therefore comparable — today Claude Code (81.0) is more publicly active than FLUX (69.9), even though they never share a leaderboard.

04Confidence

High●●●Two or more independent counter source families for the size signal
Medium●●○One strong size source plus supporting signals
Low●○○Watchlist / attention-only coverage

Confidence affects how much we trust the score. It does not directly increase the score. It is computed from signal coverage, never assigned; missing signals are holes, never zeros.

05Sources

SignalSourceUsed forStatus
Downloadsnpm registry + downloads APISize · GrowthLive
StarsGitHub APISize · GrowthLive
ModelsHuggingFace APISizeLive
InstallsVS Code MarketplaceSizeLive
MembersDiscord invite APISizeLive
DiscussionHN Algolia + Reddit (Arctic-Shift mirror)MentionsLive
SearchGoogle Trends (unofficial)SearchLive
X APIExcluded
Reddit native JSONWalled

All sources are free and public. Every snapshot is versioned JSON in git — the index updates from data, not from a dashboard.

06History

Collected 4× daily (00/06/12/18 UTC) · the index is a time series, not a directory

14 AUG15 AUG16 AUG17 AUG18 AUG19 AUG20 AUG21 AUG22 AUG23 AUG

Snapshot #005 · 18 AUG 2026

History building

24H
✓ unlocked
7D
5 / 8 · 3d to go
30D
5 / 31 · 26d to go
90D
5 / 91 · 86d to go

The shorter windows unlock as history accumulates — the index literally grows more useful every day.

07Limitations

What this index does not measure

Private usageRevenueMarket shareProduct quality“Best AI product”

It ranks what is publicly observable — downloads, stars, installs, community size, mentions and search interest. It structurally favors tools with open, observable footprints. Nothing here is investment advice, and no vendor was consulted or compensated.

Full disclosure & method notes

Scores are computed from frozen, ratified weights (method prototype-v5); every snapshot is versioned JSON in git; holes are holes. Percentiles run over the log-scale value in one global cohort of all eighteen tools.

During the shadow-collection phase the index is a preview of the real product: the first weeks of history are what bootstrap growth and longer timeframes. The methodology and weights are public so that the ranking can be audited, argued with, and rebuilt.

Want to argue with the weights? That is a feature, not a bug — adjustable weights are planned on this very page.