Been thinking about second brains for AI.
Same concept as the human versionβoffload context so you can pick up where you left off. Except it's md files the model reads, not you.
Works surprisingly well.
bridger.to/llms.txt
I've found that this approach helps guide the AI to focus more on customer impact rather than simply listing features.
Using AI for marketing / GTM copywriting?
Consider asking the LLM to evaluate text using the value equation:
Value = (Dream Outcome Γ Perceived Likelihood of Achievement) / (Time Delay Γ Effort & Sacrifice)
imho: mastering AI is the highest ROI skill right now. not just 'using it'. mastering it. the gap between those two is massive.
π Feature
The prompt I use for adding premium / subtle animations to a website with Claude Code and Cursor
git.new/animation-p...
I'm bullish on .app domains
The Dolomiti
dub.sh/dont-quit
Install it with `npm install payload-translate`
Then add it to your Payload config π
For translation is uses Gemini 2.5 Flash for now since it did very well at WMT25 (more to come).
Introducing Payload Translate πππ
An @payloadcms plugin for AI translation across your locales.
Works off of Payload's amazing localization setup and allows translation from a source document to any other locale you have pre-configured in your Payload backend.
Something new
Interesting how AI is simplifying UI down to:
Chat
Table
Canvas
Buttons
Terminal
Weird realization: I've used more skills from management books than programming books when working with AI lately.
Getting it to work within your system feels more like coaching than coding.
JSON-LD used to be "nice to have" for rich snippets.
Now? AI models are using structured data to understand and reference your content. Skip it and you're invisible to both Google and the answer engines.
5 minutes of implementation. Two optimization layers.
The simplification of UI happening with AI is underrated.
Instead of changing the interface to change the workflow, now you just update the tools and prompts behind a chat.
Same UI, completely different capabilities. Iterate without disrupting users.
We also stripped out the minuscule set up pieces by added robots.txt to block crawlers from indexing the WP backend. WordPress is just your CMS now, not your website.
If you've ever wanted to use WordPress for content but actually build your frontend in React, this is the fastest way.
The revalidation piece is the part I'm most happy with. You edit a post in WordPress, it pings your Next.js app, cache clears instantly. No waiting, no manual rebuilds
So we built a one-click Railway template that handles all of it:
β MySQL database with persistent storage
β Custom WordPress image with the revalidation plugin pre-installed
β Next.js frontend already connected to the WP REST API
β Headless theme that redirects any WP visitors to your Next.js site