A junk-drawer upgrade became my “still moving” project.
I have been dealing with some pretty severe nerve pain, and at points it has felt genuinely debilitating. Long, deep work or creative sessions have not been realistic. But I am not willing to stop making things, learning, and exploring.
That is why this junk-drawer project was perfect. It was small enough to carry, easy to pause, and paced by print cycles instead of brute force. Overall, I spent about a week printing nonstop: first several days printing baseplates, then several days printing bins. Individual prints were typically in the 5–12+ hour range in both phases.
What Is Gridfinity?
If Gridfinity is new to you, it is an open modular storage system built around a repeatable grid. You print baseplates and bins, then rearrange them as your needs change. Zack Freedman created the format, and the maker community has expanded it in every direction.
- Gridfinity specification (dimensions, units, and compatibility rules)
- Gridfinity documentation (design details and implementation notes)
- Gridfinity Generator (the browser tool I used for this project)
- Gridfinity for Dummies (quick video intro if you want a fast visual overview)
How I Started
I started with a tape measure. I measured the drawer, then used ChatGPT to sanity-check the numbers and confirm the size was reasonable for planning.
The key math was straightforward but useful: the drawer is about 571.5 mm x 482.6 mm, which mapped cleanly to a 13x11 Gridfinity layout (546 mm x 462 mm), leaving enough margin to be practical. From there, ChatGPT helped me pick a baseplate tiling strategy and call out real-world constraints like tolerances and edge margins. I laid the baseplates out on the floor first, but not to iterate design. It was simply a fit check before committing to the drawer.
Tooling Choices
For the design side, I used Gridfinity Generator for all bins, faceplates, and baseplates. Mostly defaults. No magnets.
That turned out better than expected. For this drawer, friction fit is solid. I might try magnets later, but they were not required to make this useful.
ChatGPT stayed in the loop the whole time as a planning partner. I used it to build a full print plan (layout strategy, bin recommendations, organization approach), generate diagram-style planning output, and turn the plan into PDF so I could revise it as constraints changed.
The biggest redesign came when I accounted for my 256x256 printer limit. With ChatGPT, I reworked the baseplate tiling strategy and validated what could and could not physically print before wasting filament.
A few technical details from that plan were especially useful in practice:
- Treat 5x5 as the largest reliable baseplate tile on a 256x256 printer.
- Assemble the drawer with a mixed tile set (5x5, 3x5, 5x1, 3x1) instead of forcing giant prints.
- Print order mattered: one 5x5 test tile first, then full baseplate set, then bins.
- Use bin height intentionally: 3u for shallow items, 4u as the default, 5u for deeper hardware.
The planning principle was simple: front for frequently used tools, middle for hardware, back for overflow. Real usage still won in the end, but this gave me a strong starting blueprint.
The Setback: Corner Warping
The first real blocker was corner warping. On larger prints, corners started curling just enough to break fit. Worse, from some angles it looked fine until it wasn't.
The fix was raising bed temperature from the 35-40C range to 50C. That one change stabilized corners and got the project moving again.
From Fix to Flow
After the bed-temp adjustment, I did not jump straight to "final." I ran fit checks, validated bin engagement, and watched for repeatability. I needed confidence, not one lucky print.
Once the geometry behaved consistently, I went back to a steady rhythm: print, wait, test, adjust. Baseplate runs were long enough that most progress happened between sessions, not during them.
The Loop That Actually Worked
My process was less "design everything first" and more "keep moving while life is hard." This is what I actually did:
- Keep long prints running continuously (several days of baseplates first, then several days of bins, typically 5–12+ hours per print).
- Print a handful of bins and place them sort of randomly.
- Move all drawer contents into a cardboard box.
- Slowly move items from the box into bins as bins became available.
- Rearrange as I learned, then choose what to print next.
- Share before/after/refinement photos with ChatGPT and get feedback on layout, bin usage, and next prints.
- Adjust toward real behavior, not idealized structure (for example, keeping one main screw pile).
Over time the workflow became very clear: measure → design → plan → build → adapt → reflect. ChatGPT was integrated at every stage, but final decisions were grounded in how we actually use the drawer.
Stable State Progression
This was not a one-shot transformation. It evolved in visible steps:
This was exactly the kind of project I needed: no long heroic sessions, just steady loops that tolerated interruption. The printer ran almost nonstop for about a week — several days of baseplates, then several days of bins, often 5–12+ hours per run — and that pace let me keep moving forward, build something genuinely useful, and made my wife happy to boot.