AI Search Guide, Docanban MCP, Performance Wins, and New Hardware
AI search formatting, a massive performance win, Docanban getting an MCP server, and new hardware arriving next week.
This was one of those weeks where I shipped more than I planned and slept less than I should have. Five major guides published, 90+ articles restructured, a Cloudflare R2 setup that’s saving me real money, and a site speed improvement that genuinely surprised me. Here’s everything.
The AI Search Formatting Guide
This one consumed most of Monday and Tuesday. I spent two weeks before that searching my own articles across Perplexity, ChatGPT, Claude, Gemini, and Microsoft Copilot. Testing which ones got cited. Tracking the patterns.
The uncomfortable finding: my highest-traffic post on Google, a 4,800-word guide with strong backlinks, didn’t get cited once by any AI search engine. Meanwhile a 1,900-word post I almost didn’t publish? Cited by Perplexity eleven times in one month.
The difference wasn’t quality. It was structure.
I wrote the whole system up: How to Format Your Blog Posts for AI Search. Covers answer-first paragraphs, entity-dense writing, which schema types actually matter (only three: FAQ, HowTo, Article), and a comparison table showing how each AI engine handles citations differently.
The article has six custom SVG graphics, a formatting checklist block, and a comparison table across all five AI platforms. It’s the kind of piece that takes days to build properly. But I’ve already applied the formatting changes to my top 20 articles, and early results show AI citations jumping from about 4 per month to 31. Small numbers. Clear trend.
442ms to 38ms: The Performance Win
This one felt good. I’ve been working with WP Multitool by @MythThrazz on optimizing the backend performance of gauravtiwari.org, and the numbers this week were something else.
The homepage went from 442.8ms slowest callback to 38.4ms. The about page dropped from a 2,849ms total to under 1,200ms. That’s not a CDN trick or a caching layer masking the problem. That’s actual server-side execution time dropping by over 90%.
The fix involved cleaning up expensive database queries, optimizing hook execution order, and removing redundant callbacks that were firing on every page load. WordPress sites accumulate this kind of cruft over years, especially when you’ve got 2,000+ posts and dozens of active plugins.
I posted the before/after on X and the response was immediate. People want to know how. I’ll write it up properly once we’ve finished the remaining pages.
Docanban: MCP Server and REST API
I love this little product. Docanban is a Kanban board I built for editorial workflows, designed specifically for Markdown documents. This week, I got the MCP server and REST API done.
What this means: you can now manage your Docanban boards through AI assistants. Move cards between columns, create documents, update statuses, all through natural language. The REST API opens it up to any automation tool. It is under limited beta. Send me a message to get access.
I also shipped UX improvements this week. The Writing System board now tracks 364 documents across 4 columns with tag filtering. It sounds simple, but getting the board to feel like a real workspace (not a fragile demo) took serious iteration. Stable board URLs, refresh-safe state, drag-to-reorder, keyboard shortcuts for search and new docs.
Again, Docanban is still in free invite-only beta. If you work with Markdown documents and want a visual way to track them, reply and I’ll send you access.
FlyingPress vs WP Rocket: The Real Benchmark
This comparison had been sitting in my drafts for months. I finally ran it properly: FlyingPress vs WP Rocket.
Not a feature-list comparison. Real benchmark data across 12 test sites. Time to First Byte, Largest Contentful Paint, Total Blocking Time, the full Core Web Vitals panel. Same hosting (Cloudways), same theme, same content. Just the caching plugin swapped.
FlyingPress won by a meaningful margin. Not on every single metric, but on the ones that actually affect your PageSpeed score and user experience. The critical CSS generation is noticeably better, and the automatic unused CSS removal works more aggressively.
I’ve used WP Rocket since 2014. Writing that FlyingPress is better now wasn’t easy. But the numbers are the numbers.
GEO vs SEO: The Playbook Nobody’s Writing
Published GEO vs SEO: The Complete Playbook on Sunday. GEO is Generative Engine Optimization. The practice of making your content visible to AI search engines, not just Google.
The key insight from writing this: GEO and traditional SEO overlap about 40%. That’s less than most people assume. The 60% that doesn’t overlap is where the opportunity lives. Things like answer-first formatting, entity density, and FAQ schema have a bigger impact on AI citations than they do on Google rankings. Conversely, backlink building and keyword clustering matter way more for Google than for any AI engine.
The playbook covers both channels without doubling your workload. One content strategy that feeds both discovery channels.
The Internal Linking Project
Two related guides dropped this week: Internal Linking for WordPress and How to Fix Orphan Pages.
These came out of a project I ran on my own site. I audited all 2,100+ posts and found 679 that had zero internal links pointing to them. Orphan pages. Google can find them through the sitemap, but the site architecture treats them like they don’t exist.
I built a Python script that crawled my site’s internal link graph, identified every orphan, and automatically injected contextual links from related articles. 679 posts went from 0 internal links to 3-5 each. The script analyzed content similarity to pick the right anchor text and placement.
Early signals: crawl depth improved across the board, and several orphaned pages that had been stuck on page 3-4 of Google started climbing within two weeks.
Functionalities Plugin: The Meta Module
Quick one, but it got good engagement on X. The Functionalities plugin now handles full copyright and Dublin Core metadata injection automatically. Author attribution, content licensing, creation dates, modification dates, syndication sources, all through clean meta tags.
You need Functionalities plugin solely for this. One lightweight plugin replacing the five or six separate ones most people use for metadata, redirects, schema, and SEO cleanup. It’s the kind of invisible infrastructure that makes a site properly machine-readable without any manual work.
Cloudflare R2 for WordPress Assets
Published this one a couple weeks ago but it’s been generating steady traffic: How I Serve WordPress Assets from Cloudflare R2.
The setup: all static assets (images, SVGs, PDFs) served from Cloudflare R2 instead of the origin server. The cost difference is absurd. I was paying roughly $25/month in bandwidth through my hosting. R2’s free tier covers the first 10 GB of storage and the first 10 million requests. My total R2 bill last month was $0.47.
TTFB for images dropped from 180-220ms to 40-60ms. Every image on this blog now loads from an edge server within 50ms of wherever you are.
Speaking of Cloudflare, they quietly bumped Workflows step limits to 25,000 this week. If you’re building anything on Workers, that’s a significant upgrade for complex logic without recursion workarounds.
New Hardware: MacBook Pro M5 Pro
The 16” MacBook Pro in Space Black with M5 Pro is arriving March 13th. 18-core CPU, 20-core GPU, 16-core Neural Engine, 24GB unified memory, 1TB SSD. With AppleCare+ because I don’t take chances with work machines.
The timing is deliberate. My current setup handles everything fine, but the AI-heavy workflows I’m running now (Claude Code sessions, local model testing, heavy Playwright automation, multiple Docker containers) are pushing it. The M5 Pro’s Neural Engine should make a real difference for on-device inference.
I’ll share the migration process and first impressions once it’s set up. The developer workflow comparison (build times, Docker performance, AI tool responsiveness) is what I’m most curious about.
The Rupee Situation
Couldn’t ignore this one. The Indian Rupee fell to an all-time low against the US Dollar this week. For anyone running an international business from India, this is a double-edged sword. My revenue is mostly in USD, so the conversion helps on paper. But hosting costs, tool subscriptions, and hardware (hello, MacBook at Rs 2,99,900) all get more expensive.
The practical takeaway for creators and freelancers billing in INR: reduce spending, increase sales. Not profound advice, but it’s what I’m doing. Audit every subscription, cut what isn’t earning its keep, and focus energy on the revenue-generating work.
What’s Coming Next Week
Content consolidation. About 25 articles in the merge-or-redirect queue. Thin pages that should either be combined into stronger guides or redirected to better content.
SEO vs AI search data update. The existing piece needs real citation tracking data now that I’ve been monitoring for six weeks.
MacBook Pro M5 setup. Arriving March 13th. Full migration and developer workflow comparison incoming.
Substack Notes. Starting this week: 3-5 notes (not newsletters) per week minimum. Short observations, behind-the-scenes moments, and teasers for upcoming guides.
That’s the week. Heavy on content infrastructure, product development, and backend performance.
If you’re working on anything similar, especially the AI search formatting, the performance optimization, or editorial Kanban workflows, I’d love to compare notes. Reply anytime.