
Your AI App Has a Backup. Can It Actually Recover?
Run a disposable SQLite restore drill that catches wrong business values even when database integrity passes. Includes tested Python code and failure checks.
BY JASON EAD
Original guides and analysis about coding tools, models, GitHub projects, agents, and creative AI workflows.

Run a disposable SQLite restore drill that catches wrong business values even when database integrity passes. Includes tested Python code and failure checks.

Turn AI artwork into a reviewable business asset with an approved-copy brief, crop checks and a clear human release decision.

A valid JSON file can still contain the wrong price. Build a read-only Python review that exposes every service-catalog change before an AI-assisted edit reaches your business.

Use tiktoken to preserve required instructions, include whole business records and report omitted context before an AI request.

Build a local Datasette console to inspect AI workflow records, filter review queues and understand the limits of read-only snapshots and export controls.

Render structured project briefs through a reusable Typst template, preserve review artifacts and test the actual PDF output. Includes complete Python and template code.

Build and locally test a check-only Python formatting hook, separate style feedback from business correctness, and verify Claude integration before relying on it.

Run a local embedding model over business procedures, return traceable sources and test the questions your library cannot answer. Includes complete Python code.

Build a local DOCX-to-Markdown review bundle with source snapshots and hashes. A tested example shows why successful conversion can still lose business context.

Give business agents approved service details through two read-only MCP tools. Build the Python server and test real stdio calls with a fictional catalog.

Build a silent portrait promo from AI artwork and editable campaign copy. Includes a tested FFmpeg renderer, file checks and a fictional business example.

Use a transactional outbox and stable action IDs to recover CRM handoffs, with a tested local example that exposes duplicate-task and lost-reply failures.

Separate an AI coding task from unfinished business-app work with Git worktrees, a reusable task prompt and a tested review-to-cleanup example.

Build a tested Python scorer for AI ticket routing. Use fictional cases, severity weights and strict run checks to catch regressions hidden by overall accuracy.

Build a local review queue for possible duplicate CRM companies. Includes tested RapidFuzz code, a fictional dataset, threshold results and limits for AI workflows.

Create a campaign hero with AI-generated artwork, editable HTML copy and a responsive layout. Includes the exact prompt, complete source and a tested fictional example.

Build a tested local DuckDB workflow that calculates weekly order metrics, produces a reproducible JSON brief and gives an AI assistant a focused business-review task.

Package a recurring business-brief workflow as an agent skill, with a reusable template, a worked handoff and checks that keep missing decisions visible.