docs: fix and annotate research links after link check

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
Lucas Winther
2026-10-30 00:23:11 +01:00
co-authored by Claude Opus 5.5
parent 15740a054f
commit 071910f5f5
5 changed files with 15 additions and 6 deletions
+2 -2
View File
@@ -218,7 +218,7 @@ There is no importer. Windmill itself states: "No, there is no automated n8n-to-
— [windmill.dev/compare/n8n](https://www.windmill.dev/compare/n8n)
- The CFA Institute case study (vendor-published, January 2026): the SRE team evaluated Windmill for about a year and fully migrated off n8n in early 2025. "We were very quickly able to get our scripts out of n8n and transitioned over"; "only brief training sessions" were needed. Their reasons for leaving were wanting real programming capabilities and Git workflows. They run 25 synthetic tests every 5 minutes on Windmill Cloud EE. No workflow count or effort figure is given. — [CFA Institute case study](https://www.windmill.dev/case-studies/cfa-institute)
- n8n workflows can be exported as JSON via the CLI (`--backup` = `--all --pretty --separate`), but credential secrets stay encrypted in the n8n database. — [n8n CLI import/export](https://www.mintlify.com/n8n-io/n8n/cli/import-export)
- n8n workflows can be exported as JSON via the CLI (`--backup` = `--all --pretty --separate`), but credential secrets stay encrypted in the n8n database. — [n8n CLI import/export](https://docs.n8n.io/hosting/cli-commands/)
- One third-party anecdote: Windmill's rollout took 9 days including SSO, against 3 days for n8n, for two UK clients. The author flags the small sample. — [automationatlas.io](https://automationatlas.io/guides/n8n-vs-windmill-2026-comparison/)
### Inferences
@@ -238,7 +238,7 @@ The recurring criticisms are these: it is a developer-only tool with a steeper l
- On HN commentary about Windmill's benchmark claims, commenters said predictable per-job performance under concurrency matters more than raw speed. — [GeekNews summary (2023-11)](https://news.hada.io/topic?id=12005) (secondhand, older)
- An unverified community anecdote describes a problematic workflow affecting a broader instance. The source suggests checking worker isolation and resource limits. — [airframe.ai analysis](https://www.airframe.ai/product/windmill-dev/analysis) (low reliability)
- "Windmill's curve is steeper"; the integration catalog is "smaller and closer to developer-leaning targets"; teams need per-language worker pools. — [automationatlas.io](https://automationatlas.io/guides/n8n-vs-windmill-2026-comparison/)
- Self-hosting is a "slightly more complex due to multi-service architecture" setup. The question is framed as "do you want to click or do you want to type?", and for non-developers setting up webhooks, "n8n wins". — [Railway compare](https://railway.com/deploy/n8n-vs-windmill) (promotional)
- Self-hosting is a "slightly more complex due to multi-service architecture" setup. The question is framed as "do you want to click or do you want to type?", and for non-developers setting up webhooks, "n8n wins". — [Railway compare](https://railway.com/deploy/n8n-vs-windmill) (dead link as of 2026-10-30) (promotional)
- Postgres 18 is the default, and older data volumes (16 or earlier) block startup until you migrate. This is an upgrade friction point. — [Self-host docs](https://www.windmill.dev/docs/advanced/self_host)
- The legacy low-code app editor is now labeled "legacy" in favor of full-code React or Svelte apps. — [Docs intro](https://www.windmill.dev/docs/intro)