AI Showcase: Working Tools and Project Blueprints

Try a working local CSV inspector, explore six ambitious AI project blueprints and build a concrete brief for your next coding, research or creative project.

What if you could just make the thing?

A research collection. A tool for a recurring annoyance. A practice space for a skill you care about. Start with the working example, then take a project brief into your own workspace.

A working CSV inspector

Paste a small CSV or explicitly select a UTF-8 file up to 200 KB. The tool checks blank or missing expected fields, uneven records and exact repeated records after the first. It processes text locally in the browser without calling an AI model or uploading your input. It does not modify input or persist it across reloads.

The first non-empty record is a header. Empty lines are skipped; record numbers include the header and count decoded records, so quoted line breaks remain within a record. Duplicate comparison preserves spaces and case. The demo accepts at most 10,000 data records.

The synthetic example contains six data records, three columns, three blank or missing expected fields, one exact repeat and one uneven record. Download a plain-text inspection report in the interactive version. These are structural signals for review, not a claim that a repeated record is wrong. No semantic or business-rule validation is performed.

Build your own version.

Six project blueprints

These are projects to build, not completed customer case studies.

Your own research observatory

Bring: Five approved source records with dates, links and uncertainty notes.

Make: A searchable collection with a timeline and source detail pages.

Request: Build a local source library from these approved records. Start with search and topic filters, then add a timeline that distinguishes exact and approximate dates. Every claim must link to its supporting source. Keep uncertain locations out of maps until reviewed.

Check: Open every source in the first five records. Search for a known term, combine filters and check a record with an approximate date.

Push further: Add a comparison view for sources that disagree, then a reviewed public export.

Read the walkthrough

A tiny tool for a recurring annoyance

Bring: A synthetic CSV that includes blanks, a quoted comma and an exact duplicate.

Make: A browser tool that reports structural issues and exports a review note.

Request: Build a local CSV inspector with a paste box, explicit file selection and a synthetic sample. Show blank fields, uneven rows and exact duplicate rows. Preserve input and label all findings for review. Test quoted commas, line breaks, empty files and invalid quotes.

Check: Compare known sample counts, change the input and open the downloaded report.

Push further: Add required-column rules for one real export after its owner approves the definitions.

Read the walkthrough

A practice space for something you care about

Bring: Ten questions you have checked, with answers and explanations.

Make: A quiz that gives feedback and lets you retry missed concepts.

Request: Create a local practice app from these ten reviewed questions. Ask one at a time, explain mistakes after an attempt, and let me retry missed questions. Keep progress on this device, with a clear reset. Make it work with a keyboard and on a phone.

Check: Answer one wrong on purpose, retry it, reload and check progress, then reset only the demo entries.

Push further: Add a spaced review queue using an explicit rule you can explain, and keep the question source visible.

Read the walkthrough

A workshop for your creative process

Bring: Three approved references and a description of the audience and mood.

Make: A direction board with alternatives, editable notes and an exportable brief.

Request: Build a local creative direction board. Let me compare three distinct approaches, annotate what works, and export the selected direction as a brief. Label reference images and their sources. Keep the selected composition and typography decisions explicit.

Check: Change directions, preserve a note, reload and inspect the exported brief for the correct choices.

Push further: Add side-by-side revisions so you can evaluate a change without relying on memory.

Read the walkthrough

An atlas of your scattered files

Bring: A working copy of a small approved folder and an inventory.

Make: A searchable index with tags, paths and an unresolved-items view.

Request: Inventory this working folder and build a searchable index. Preserve original paths and files. Add tags only when supported by content, and keep ambiguous classifications in an unresolved view. Show which files were unreadable.

Check: Locate three known files, inspect an ambiguous item and reconcile the inventory with the working folder.

Push further: Propose a reversible rename plan with collision checks and a before/after manifest.

Read the walkthrough

The weekly report that remembers its method

Bring: Two approved sample exports and a definition of the reporting period.

Make: A repeatable report, supporting table and visible exceptions list.

Request: Create a weekly report workflow using these approved sample exports. Define admitted rows and denominators, retain source identifiers and show unresolved exceptions. Run it on both samples. Package the method and instructions so I can repeat it with a new export.

Check: Recalculate one group, remove a required input and confirm the workflow reports the gap.

Push further: Schedule only after manual runs are useful; notify on completion or missing data with source timestamps.

Read the walkthrough

Build your project brief

Choose a blueprint, your specific goal, audience, available context and action boundary. The interactive builder combines them into an editable request to copy or download. It runs locally and does not call an AI model.

The complete AI guide | Codex handbook.

Related Power of AI pages

Keep reading with Start here, everyday uses, the tool guide, the writing workshop, and sources and standards.