AI social media agent: a tutorial with human approval
Our AI social media agent in production: a queue filled 3 weeks ahead, copy rewritten for each network, script checks, human approval, and sending through the Zernio API.
Blog
We regularly share practical resources for startups
Landing page audit with Claude Code: review your SaaS page
Audit a SaaS landing page with Claude Code and Playwright: desktop and mobile screenshots, DOM measurements, Core Web Vitals, a 256-criterion LEVER grid in five passes and quantified A/B hypotheses.
SEO Audit with Claude Code: A Six-Step Method for SaaS Sites
A six-step SEO audit with Claude Code: curl on the served HTML, sitemap HTTP codes, internal link graph, a 6-point E-E-A-T score and a prioritized fix plan.
How to Create Ads with AI for Your SaaS: A 5-Step Method
Create ads with AI without garbled text: a JSON brief, local rendering in 1:1, 4:5 and 9:16, word-by-word proofreading, and a paused Meta import. Commands and thresholds included.
Create shorts with AI: our step-by-step pipeline (2026)
How to create shorts with AI from a long video: the commands, settings and pitfalls behind 40 clips we rendered locally with Whisper, an LLM and ffmpeg.
Free AI Subtitles: The Local Method with Whisper (2026)
Free AI subtitles, run locally: faster-whisper on CPU, a full transcript review, ASS karaoke captions, the 9:16 safe zone and audio at -14 LUFS. Our commands and our thresholds.
Google Search Console and Claude Code: Connect GSC and Triage Your Queries
Connect Google Search Console to Claude Code (CLI or MCP), export 90 days of page × query data and triage your queries with six numeric rules, tested on swanbase.co.
Market research with Claude: a method with no made-up quotes
How to run market research with Claude or ChatGPT without inventing customer quotes: API collection, LLM labeling, script-computed numbers and verified citations.
Meta Ads MCP in Claude: connect your account and audit it across 10 areas
Connect the Meta Ads MCP to Claude and pull your account data level by level. Then audit it across 10 areas using the thresholds from our own audits, with a clear view of the connector's blind spots.
