What I learn building on Base44 and with AI tools, written up while it's still fresh. New posts land here first.
blog to-do
14 posts so far
9 min read
On 4 October my GitHub sync copied two Base44 backend functions into my app, I published, and both still answered 404 "not found or not deployed". How to test a function with curl, redeploy it with the CLI or one Apps API call, read secrets and logs properly, and catch bugs first with Deno.
14 posts
9 min read
My Base44 Partner leads land in my own CRM through a backend function with a public URL, so it checks every delivery first. Signing is required (no secret, 503), the HMAC is checked on raw bytes in constant time, retries are skipped by event id, and a failed save returns 500 so Base44 tries again.
8 min read
Google says AI Overviews need no special schema or AI files: a page just has to be indexed and eligible for a snippet. So for my own name I'm removing everything that makes "Aatman Jain" ambiguous, from a wrong canonical to AI-written fallback text and mismatched profiles. No citations yet. Here's the plan.
7 min read
I made aatmanjain.com readable for AI agents: llms.txt, agents.md, a Markdown copy of every page, a read-only JSON API and discovery files, for about 124 extra bytes of gzipped JavaScript. What goes in each file, how I generate them on Base44, the deploy traps, and what static hosting still can't do.
7 min read
My Base44 site was built with Claude Code: it works on a branch, I read the pull request and merge into main, and the merge syncs to Base44 for publishing. The setup, the workflow, and three traps I hit: Base44's bot commits to main, a redeploy isn't a rebuild, and the build thinks it's 2022.
8 min read
Hattle, my World Cup sweepstake app, went from a paper sketch to shipped in 8 hours. The draw is a seeded Fisher-Yates shuffle anyone can re-run and check. Here's how it works, an illustrative code sketch, and my own Node test showing why the popular sort-with-random shuffle is quietly unfair.
9 min read
A UK GDPR cookie banner on Base44 needs equal Accept and Reject buttons, and code that stops anything optional running first. On my site, Base44's own analytics logged page views and stored a visitor ID before anyone clicked. Here's how I gated it, with the code and the checks.
7 min read
Shipping 20+ apps for real clients taught me that code is rarely the hard part. The hard part is finding the real problem behind the first ask, then checking the boring details before users do. Here's what freelancing at 15 has taught me, with the reviews to back it up.
9 min read
The JavaScript my Base44 site loads before first paint went from 336 KB to 245 KB gzip, and phone scrolling went from 54 janky frames to none. No speed change could merge until screenshots of every route matched the old build. What worked, what I threw away, and why Lighthouse disagrees.
7 min read
Base44's SendEmail only reaches people who signed up to your app, unless you're on a paid plan with a verified custom domain. Here's the rule as a table, the two domain settings people mix up, a save-first contact form pattern, Resend's DNS records, and my checklist for missing emails.
8 min read
Every entity on my site has an explicit rule for create, read, update and delete. Here are the four RLS patterns behind all ten, two more for apps with users (clearly marked untested), when to reach for the service role, and the checklist I run with Base44's security scan before launch.
8 min read
AI-built apps look the same because every decision you skip gets a safe default. Here's how I make those decisions myself, with real examples from this site and Hattle, and exactly where to put your rules in Base44: AI Controls, the Theme panel, Plan mode and Discuss mode.
9 min read
My Base44 portfolio kept going blank until a reload. Nothing had crashed: the pages were stuck behind framer-motion animations that never finished. How to tell which blank you've got, the two bugs behind mine, and the four fixes that took permanent blanks to zero in about 1,000 Chromium and 71 WebKit checks.
8 min read
Base44 hands crawlers a prerendered snapshot of each page, but only for routes written out as literal paths in App.jsx. Anything on a /:slug route gets a bare shell with your default title. How I caught it on my own site, the canonical tag that made it worse, and a two-curl test for yours.
Newsletter
New posts, what I built and the Base44 and AI news worth knowing. One email, every Friday at 8am.
Ask me. If it's a good question, it might turn into a post.