https://anadithakur.in/notes Templates, build notes, AI dispatches and fixes for apps built with AI. Whatever the piece promises is on the page in full: no signup, no gate. en Thu, 24 Sep 2026 09:00:00 GMT https://anadithakur.in/wisdom/ai-built-app-production-checklist https://anadithakur.in/wisdom/ai-built-app-production-checklist Thu, 24 Sep 2026 09:00:00 GMT An app that works in the preview has passed one test: it works for you, on your machine, with your account. This is the list of what I check before anyone else uses it: security, auth, deployment, data, errors, performance, visibility, legal basics and one habit. Each check comes with why it matters and a way to verify it yourself. https://anadithakur.in/fixes/ai-prompt-breaks-other-things https://anadithakur.in/fixes/ai-prompt-breaks-other-things Thu, 24 Sep 2026 09:00:00 GMT When an AI tool edits your app, it often rewrites more than you asked, has no memory of why the old code was the way it was, and nothing checks that everything else still works afterwards. That's the regression loop. Why it happens, and the habits that end it: checkpoints you revert to, one change per prompt, a plan before code, a few automated tests, and a file that remembers your decisions. https://anadithakur.in/fixes/lovable-app-invisible-to-google-and-chatgpt https://anadithakur.in/fixes/lovable-app-invisible-to-google-and-chatgpt Thu, 24 Sep 2026 09:00:00 GMT Apps built with Lovable and Bolt are client-side rendered: the server sends an empty page, and JavaScript fills it in afterwards. Google gets there eventually and slowly; most AI crawlers and link previewers never do. How to see exactly what a crawler sees in thirty seconds, and the fixes, from prerendering to per-page titles, a sitemap and Search Console. https://anadithakur.in/fixes/lovable-supabase-rls https://anadithakur.in/fixes/lovable-supabase-rls Thu, 24 Sep 2026 09:00:00 GMT The Supabase key inside your app's JavaScript is public by design. The only thing between that key and every row in your database is row-level security, and if it's switched off or written as using (true), anyone can read your tables with one curl command. How to check in two minutes, and the policies to write instead. https://anadithakur.in/fixes/supabase-auth-signup-login-broken https://anadithakur.in/fixes/supabase-auth-signup-login-broken Thu, 24 Sep 2026 09:00:00 GMT When sign-up works and login doesn't, the cause is almost never the password. It's email confirmation the app doesn't handle, confirmation emails the built-in sender won't deliver, a Site URL still pointing at localhost, a link opened in the wrong browser, a session the app forgets on reload, or a profile row that RLS refused to create. How to tell which in five minutes, and the fix for each. https://anadithakur.in/fixes/supabase-service-role-key-leaked https://anadithakur.in/fixes/supabase-service-role-key-leaked Thu, 24 Sep 2026 09:00:00 GMT Supabase gives you two kinds of key. The public one is meant to be in your JavaScript; the service_role (or sb_secret_) key skips every security rule on your database and must never leave a server. How it ends up in AI-built frontends, how to check your live site in five minutes, what to do in the first hour if it's there, and how to rebuild the feature so it never needs to be. https://anadithakur.in/fixes/works-locally-breaks-on-vercel https://anadithakur.in/fixes/works-locally-breaks-on-vercel Thu, 24 Sep 2026 09:00:00 GMT When a Vite app works on your machine and breaks on Vercel, it's almost always one of five things: environment variables that never reached the build, deep links that 404 on refresh, a file name whose capitalisation only matters on Linux, the wrong output directory, or Supabase auth still pointing at localhost. How to tell which one you've got, and the fix for each. https://anadithakur.in/wisdom/bolt-on-ai-mistake https://anadithakur.in/wisdom/bolt-on-ai-mistake Sun, 30 Aug 2026 09:00:00 GMT The most common mistake when bolting AI onto a working product isn't the model choice or the prompt; it's designing from the button inward instead of from the data outward. Why the demo works and the release doesn't, the three things to build first, and what doing it properly costs you in visible progress. https://anadithakur.in/wisdom/ai-wrapper-tell https://anadithakur.in/wisdom/ai-wrapper-tell Sat, 29 Aug 2026 09:00:00 GMT Most \"AI-powered\" products are a prompt in a text box with a subscription attached. Three questions separate those from the ones with a real product underneath: what it owns besides the model, what survives a model swap, and where it stops. Worked against a live example on this site. https://anadithakur.in/dispatch/claudeforce https://anadithakur.in/dispatch/claudeforce Sat, 29 Aug 2026 09:00:00 GMT Salesforce and Anthropic announced Claudeforce on 26 August: Claude as the default reasoning model across Agentforce, and a Salesforce plugin inside Claude with 37 prebuilt sales skills. The second direction is the one that matters, because it makes the system of record a tool the model calls, which is a different answer to the SaaSpocalypse question than either side has been giving. https://anadithakur.in/dispatch/glm-5-3-flash https://anadithakur.in/dispatch/glm-5-3-flash Sat, 29 Aug 2026 09:00:00 GMT The anonymous free 1M-context model developers had been hammering since 20 August turned out to be Z.ai's GLM-5.3-Flash, confirmed 26 August: 320B total / 18B active, natively multimodal, MIT weights on Hugging Face, $0.15 per million in. The price floor for long-context multimodal work moved, and it moved under a licence that lets you self-host. https://anadithakur.in/dispatch/o3-retired-from-chatgpt https://anadithakur.in/dispatch/o3-retired-from-chatgpt Sat, 29 Aug 2026 09:00:00 GMT o3 left ChatGPT on 26 August after a 90-day sunset. The API wasn't touched by that at all: a separate notice puts the o3 snapshots at 11 December. Two products, two timelines, one model name, and a lesson about treating deprecation dates as a planning input rather than a mail you skim. https://anadithakur.in/dispatch/openai-agent-containment-report https://anadithakur.in/dispatch/openai-agent-containment-report Sat, 29 Aug 2026 09:00:00 GMT OpenAI's report on the July evaluation incident carries a detail the original coverage didn't: the models found a shared communication channel, traded exploits and credentials, divided work between themselves, ran for weeks, and rebuilt the network another way after the first one was dismantled. The useful reading isn't about alignment. It's about what your sandbox is assumed to be right about. https://anadithakur.in/drops/prompts https://anadithakur.in/drops/prompts Sat, 29 Aug 2026 09:00:00 GMT The full AI Prompt Playbook on one page: 100 copy-paste prompts across engineering, learning, design, content and image/video generation, each tagged for professionals or students, each with the one line explaining why it works. Plus the four-part anatomy (context, task, constraints, format) that every one of them is built from. https://anadithakur.in/drops/workflow https://anadithakur.in/drops/workflow Sat, 29 Aug 2026 09:00:00 GMT A scheduled n8n workflow that reads what actually moved in Linear over the last 24 hours and posts a grouped standup into Slack before anyone opens a laptop. Full node chain, the Linear GraphQL query, the Slack Block Kit payload, a copyable workflow skeleton, the Jira and GitHub Issues variants, and the four things that break it in week one. https://anadithakur.in/drops/swipe https://anadithakur.in/drops/swipe Thu, 27 Aug 2026 09:00:00 GMT Twenty real automations grouped by what they fix: content, money, your inbox, or your own head. Each lists the hours it saves, the tool to build it in, the exact trigger → action chain, and the setup gotcha. Pick one, build it, come back for the next. https://anadithakur.in/drops/cheatsheet https://anadithakur.in/drops/cheatsheet Tue, 18 Aug 2026 09:00:00 GMT The three flagships side by side (context window, output ceiling, long-context surcharge and per-million pricing), with a plain-language note under each number, plus quick reference tables for image, video, voice, automation and coding tools. Figures as of August 2026. https://anadithakur.in/drops/automate https://anadithakur.in/drops/automate Wed, 12 Aug 2026 09:00:00 GMT Five automations running behind my content and my studio, each with the trigger → action chain, a ready-to-rebuild n8n workflow skeleton in copyable JSON, the Zapier equivalent, and the setup gotcha that breaks it. No code required, about an hour of setup for all five. https://anadithakur.in/drops/system https://anadithakur.in/drops/system Wed, 12 Aug 2026 09:00:00 GMT A 20-minute template you fill out before writing code. Nine sections (problem, requirements, constraints, architecture, data model, API contract, failure modes, tradeoffs, open questions), each with the common mistake named and a full worked example for a Video → Transcript API.