AILANTA
← All research
GitHub · Market formation

Where new GitHub projects outpace the active market

Two equal cohorts of 695 repositories reveal which themes are disproportionately strong among new projects compared with the established, actively maintained ecosystem.

16.07.2026695 + 695 repositoriesemerging edge vs active market
Main finding

Agent development remains the strongest structural gap

ai-agent represents 11.2% of new projects versus 2.3% of the active market. Claude Code is 5.3 times stronger, Codex 14 times, and agent-skill 8.2 times. Expanding the sample strengthened rather than overturned the original finding.

01

Emerging edge versus active market

Share of all repositories in each cohort using the topic.

Active marketNew projects
ai-agent
2.3%11.2%
4.9×
claude-code
1.6%8.5%
5.3×
codex
0.4%5.6%
14×
agent-skill
0.6%4.9%
8.2×
prompt-engineering
0.3%2.4%
coding-agent
0%1.7%
new
local-first
0%1.6%
new
mcp
2%3.5%
1.8×
02

Three leading layers

01

Agent infrastructure

ai-agent and agent-skill both outpace the installed base, indicating a distinct infrastructure category.

02

Agentic software development

Claude Code, Codex, and coding-agent are disproportionately strong among new repositories. Software development remains the principal agent market proving ground.

03

Local-first as an adjacent shift

local-first represents 1.6% of the new cohort and is absent from the active baseline. It is a small but independent early layer, supported by the separate runtime study.

03

Key market gaps

GitHub topicActive marketNew projectsGap
ai-agent2.3% (16)11.2% (78)+8.9 pp
claude-code1.6% (11)8.5% (59)+6.9 pp
codex0.4% (3)5.6% (39)+5.2 pp
agent-skill0.6% (4)4.9% (34)+4.3 pp
prompt-engineering0.3% (2)2.4% (17)+2.1 pp
coding-agent0% (0)1.7% (12)+1.7 pp
local-first0% (0)1.6% (11)+1.6 pp
mcp2% (14)3.5% (24)+1.5 pp