Summer Projects - Taylor Tam

The premise

Some fun projects I'm working on outside my summer on Apple's iCloud Privacy team.

This is the running log: what I tried, what broke, what shipped, and what I learned along the way.

Project One

Ender 3 Printer

After the keyboard phase, naturally, I have to have a 3D printer phase. Right now, I have an Ender 3, and I'm going to set it up and customize it over the summer.

Ender 3 printers are common open-source machines, but they are still fairly manual and a little rough around the edges. There is no automatic bed leveling by default, setup can be finicky, and getting consistently good prints usually takes iteration.

The goal of this section is to track upgrades, fixes, calibration notes, and what changes actually make the printer better over time.

What I'm Tracking

Printer upgrades and quality-of-life improvements
Better calibration and more reliable print settings
Experiments with materials, parts, and workflows

Week of 06/22

Weekly update photo for the Ender 3 project — First troubleshooting session after powering the printer on.

Broken printer part on the Ender 3 — The broken extruder part that stopped the first setup attempt.

The first signs of trouble showed up fast. I found the correct cable to plug my printer in, and after the moment of truth, nothing happened.

The first step was to get the printer running. Unfortunately, with a seven-year-old second-hand printer, something was bound to go wrong, and a bearing for loading the PLA into the extruder broke off. So I ordered a replacement piece so I can start printing.

Next steps are getting a first successful print, adding automatic bed leveling, and setting up a Raspberry Pi server so I don't have to load files manually with an SD card.

Week of 07/06

Made a tiny mistake on the extruder piece. Turns out that, for an Ender3, there are two variants on the extruder stepper motor. One variant has a small screw that allows you to remove the gear, and another has a press-on gold gear you can't remove. I had the latter (unfortunately). So I reordered another extruder replacement matching the stepper motor and gear I had, and replaced it!

Replacement extruder kit installed on the Ender 3 — Installed replacement extruder assembly.

Correct replacement kit parts for the Ender 3 extruder — The correct kit after ordering the wrong variant first.

Ender 3 stepper motor with the press-on gear — The press-fit gear that caused the mismatch.

This also came with a new stepper and the first gear component I mentioned-- but the wasn't sticking, so I put some adhesive foam under the gear to make them mesh and I had a *working* extruder. I spent the red of the night tuning some print settings to produce two prints the next day: a calibration cube and, of course, benchy.

Early failed print attempts from the Ender 3 — Early failures while dialing in the settings.

Calibration cube printed on the Ender 3 — First calibration cube after the extruder fix.

Benchy test print on the Ender 3 — Benchy, because it had to happen.

Project Two

Apartment Finder

So, I've been a bit of a bum. 1) I start work in one day and didn't find an apartment in Cupertino. 2) I haven't set up OpenClaw, which is also pretty bummy, considering I probably hear about OpenClaw more than the weather.

To kill two birds with one stone, I'm going to build out a widget that helps me find short-lease apartments in Cupertino under a certain price and helps me actually get an apartment.

I'm setting up the frontend as a Swift app that posts to a local HTTP server acting as a bridge between the frontend and, soon, my OpenClaw agent. This gives the frontend a stable API so I can select an area, max price, min price, and lease length, then send those values to the local server.

From there, OpenClaw sits behind the server as a reasoning layer. It interacts with the internet, filters postings, and returns apartments I can hopefully afford.

Search Payload

POST /apartments/search payload:
{
  "state": "California",
  "maxPrice": 3300,
  "minLeaseDays": 90,
  "maxLeaseDays": 110,
  "area": "Cupertino"
}

Once this JSON is received by the backend, it gets logged and routed to either live or demo mode. A JavaScript runner builds a natural-language prompt and launches the agent search.

openclaw agent --local --json \
  --session-key agent:apartment-scraper:default \
  --timeout 180 \
  --message ""

OpenClaw browses the internet and returns JSON-like output, which I then parse and normalize into the app's listing schema. Since the OpenClaw response often comes wrapped in metadata, the runner checks fields like:

payloads[0].text
finalAssistantVisibleText
finalAssistantRawText

After that, a `normalizeListings` script cleans the results, verifies that required fields exist, checks that URLs work, and deduplicates repeated listings. The server then returns clean JSON to the Swift frontend, where I can view apartment results.

Swift results frontend — Apartment search results rendered in the Swift frontend.

That part is already pretty cool. But I still don't want to click every link, email each place manually, and track everything myself.

The next step is selecting apartments I contacted, moving them into a separate tab, and saving them into `UserDefaults` so I don't have to rerun queries every time.

Swift apartment tabs frontend — The tabbed view for tracking contacted apartments separately.

After that, I want to add details in one place and use the Apple Mail API to poll my inbox every five minutes for any emails from apartments. I can also add a status tracker to manage applications.