Before the run used on this site, an earlier extraction in July 2026 processed 667 source pages with the model claude-opus-4-8 and found 1,854 unique positions (1,074 skill-tagged). The October run processed 975 pages, including all 667 earlier ones, with claude-opus-5-5, and found 4,022 unique positions.
Because the runs share most of their sources and differ in model and coverage, the comparison is a consistency check, not a trend. If the ranking of skills were fragile, it would change noticeably between runs.
Result
Among the 50 of the top 50 skills that the July run also named, the rank correlation between the two runs is 0.84 (Spearman). The same five skills lead both runs, in nearly the same order; differences grow further down the list, where counts are small.
| Skill | Oct rank | Oct share of tagged | Jul rank | Jul share of tagged |
|---|---|---|---|---|
| Python | #1 | 29.2% | #1 | 25.8% |
| AWS | #2 | 21.9% | #2 | 15.5% |
| React | #3 | 16.8% | #4 | 9.6% |
| SQL | #4 | 15.3% | #3 | 14.2% |
| TypeScript | #5 | 14.5% | #5 | 8.1% |
| CI/CD | #6 | 12.6% | #13 | 6.5% |
| Java | #7 | 11.5% | #6 | 7.7% |
| Kubernetes | #8 | 11.1% | #12 | 6.7% |
| JavaScript | #9 | 10.7% | #9 | 6.9% |
| PostgreSQL | #10 | 9.0% | #15 | 6.2% |
| GCP | #11 | 8.8% | #7 | 7.4% |
| Go | #12 | 8.7% | #11 | 6.8% |
| Node.js | #13 | 8.4% | #20 | 5.0% |
| Docker | #14 | 8.3% | #17 | 5.7% |
| Azure | #15 | 8.2% | #10 | 6.8% |
| Git | #16 | 7.7% | #16 | 5.9% |
| Terraform | #17 | 7.4% | #25 | 4.4% |
| LLMs | #18 | 6.6% | #18 | 5.5% |
| Snowflake | #19 | 6.0% | #14 | 6.5% |
| REST APIs | #20 | 5.8% | #53 | 2.0% |
| Agile | #21 | 5.4% | #72 | 1.3% |
| Databricks | #22 | 5.1% | #19 | 5.2% |
| Linux | #23 | 4.9% | #37 | 2.7% |
| CSS | #24 | 4.7% | #28 | 3.7% |
| dbt | #25 | 4.3% | #21 | 4.6% |
| HTML | #26 | 4.3% | #29 | 3.5% |
| MySQL | #27 | 4.2% | #32 | 3.3% |
| ETL | #28 | 4.0% | #38 | 2.6% |
| Kafka | #29 | 4.0% | #31 | 3.3% |
| Infrastructure as Code | #30 | 3.9% | #133 | 0.7% |
The July run also classified a larger share of positions as data roles and fewer as "Other", consistent with its narrower source set.