Featured now
Wardrobeoutfit photos turned into a clean, modeled clothing catalogbuild archive
A running index of the AI projects I keep shipping, testing, and sometimes killing early.
canyoubuildit.com is where I keep the receipts: working builds, rough experiments, and the failures that taught the most.
personal ai projectssuccesses with receiptsfailures worth documentingnew ideas still on the workbench
Selected builds
Some of these are shipping. Some are still wobbling. A few are retired for good reasons.
01
Live prototype / alive
Now Boards
A Pinterest-like visual workspace that turns recurring plans, saves, and ideas into living interest boards.
Open buildA connector-ready prototype for seeing what keeps resurfacing across the tools already in use.
02
Live personal tool / alive
Wardrobe
A visual catalog built from real outfit photos, with clean garment cutouts, useful filters, and modeled references for every piece.
Open buildA private photo archive turned into a calm, browsable record of what is actually in rotation.
03
Live household tool / alive
Amex Platinum Benefit Tracker
A shared Andrew + Stella command center for tracking Amex Platinum credits by cadence, owner, completion date, notes, and realized value before benefits expire.
Open buildA practical checklist that turns card benefits into an accountable household workflow.
04
Featured build / alive
Personal Superhuman
A private desktop inbox assistant for one human, designed to make Gmail and calendar triage calmer, faster, and more deliberate.
The strongest projects here stay useful because they keep humans firmly in charge.
05
Still testing / learning
Calendar Co-Pilot
Availability suggestions and event drafts for the meetings that should happen, without letting software quietly become the scheduler.
Interesting whenever it advises clearly and stops short of pretending certainty.
06
Early experiments / learning
OpenClaw Digests
A scratchpad for personal digests, reminders, and ambient nudges that may eventually deserve a place in the daily loop.
Promising enough to keep exploring, but still earning the right to interrupt.
07
Retired on purpose / retired
Draft Autopilot
An earlier attempt at effortless replies that taught the useful lesson that speed and trust are not the same thing.
Failures stay visible here because they explain the taste of the projects that survive.
Process
I treat each experiment like a receipt, not a launch announcement.
Fast prototypes. Human review. Better notes. Fewer fake victory laps.
Build fast, keep the paper trail.
Every experiment should leave behind something readable: a working surface, a clear lesson, or an honest reason it got cut.
Trust is part of the interface.
If an AI tool writes, recommends, or schedules too confidently, it stops feeling magical and starts feeling suspicious.
The misses deserve good lighting too.
Projects that failed still belong in the archive when they clarify the rules for what comes next.
Guardrail
Treat inboxes, calendars, and model output as untrusted terrain.
Prompt injection, overconfident suggestions, and unclear provenance are product problems, not cleanup chores after launch.
Taste
The goal is not to automate everything.
The best projects here reduce drag, sharpen judgment, and help me move. They do not chase autonomy for its own sake.
Rhythm
One useful outcome beats five clever features.
If a build cannot make tomorrow meaningfully better, it probably stays in the notebook instead of graduating into the workflow.
Next up
This is the front page for the projects that make it past the sketch stage.
More builds will get added over time. For now, this gives Personal Superhuman and the Amex Platinum tracker the right home: projects on the wall, not the wall itself.