@dshukertjr — Supabase database via MCP server to Cursor
You can connect your Supabase database via MCP server to Cursor!
The Cursor agent can get the schema and any data from your database through the official Postgres MCP server!
This helps Cursor understand your database better when it needs to modify it to add a new feature!
Smart setup — the Postgres MCP server gives the agent real schema awareness instead of hallucinated column names. If you're using Cursor on a Supabase project this is the obvious next step.
@illyism — Shadcn redesign killed conversions and rankings
🤯 Want proof of how Google actually works?
I changed this site's colors to look more "modern" & "clean" (shadcn ui: 100% black & white basically, no colors, etc)
Look what happened:
- Conversion rate: 70% → 58%
- Time on page: 📉
- Rankings: 🪦
- Traffic: -50% in 7 days
Honest experiment with real numbers attached — that's what makes it worth saving. The conventional wisdom is that clean, minimal design wins. In practice, stripping color signals to users (and apparently to Google) that something changed, and not in a good way.
@tom_doerr — Browser API for AI automation
Browser API for AI automation
Short post, but browser-use (the tool being linked) is worth knowing about. It exposes browser actions as a clean API that LLM agents can call — navigate, click, extract. Less duct tape than Playwright wrappers, more purpose-built for agentic workflows.
@tom_doerr — Open-source RAG infrastructure for AI applications
Open-source RAG infrastructure for AI applications
The space is crowded but RAG infrastructure that you actually own and run yourself is still underserved. Saved this to look at the indexing and retrieval primitives — most of the hosted options make it hard to control chunking strategy or swap embeddings.
@GithubProjects — Run your own AI cluster at home with everyday devices
Run your own AI cluster at home with everyday devices 📱💻 🖥️⌚
That's exo, which lets you pool heterogeneous devices — phones, laptops, desktops — into a single inference cluster. NetworkChuck also posted a video about chaining five Mac Studios together for the same thing. The more interesting angle is the phone support: most distributed inference projects ignore ARM mobile hardware entirely.
@tom_doerr — Crawl site, generate GPT knowledge file
Crawl site, generate GPT knowledge file
Honestly just saved this so I don't forget it exists. Generating a structured knowledge file from a crawl is a useful primitive — feed it into a custom GPT, use it as RAG context, or just have a local copy of a third-party API's docs without hunting through pages.
