Pinned Tweet
// 23, Bay Area
// Building multi-agent systems, manufacturing software, and hardware/physical AI experiments
Long-term goal is a role in physical AI, industrial automation, or space systems. Here to learn from people in the field, share what I’m working on, and find others building real hardware + software. DMs open.
Demo of the aluminum plate nesting + quote tool I’m building to understand the real flow from parts list to cut plan.
What it currently does:
- Packs parts with kerf onto 48×96 plate
- Shows material cost, usable remnant value, and waste cost (everything that isn’t a part or usable remnant)
- Labels usable vs borderline remnants
- Mixed-order batching across plates
- Work order popup with part positions
Still learning:
- Free-rectangle tracking and scoring matter more than I expected
- Usable remnant rules and waste cost change what “good” packing looks like
- A quote needs more than density — it needs remnant value and a clear handoff to the operator
Next: stronger multi-order batching, mill-cert parsing, cut-time/lead-time estimator, and inventory-aware plate selection.
Learning how instant plate quotes actually work by building a simple nester.
The version I’m using scores placements with Best Short Side Fit (minimize the shorter leftover dimension). It works, but gaps still appear with mixed part sizes.
This is one piece of the larger system: customer parts list → nest → quote (cost + remnant value) → cut plan.
Planning to tighten remnant detection and add a basic export next.
in just over a month, we've had
- flying cars
- AI solving a Millenium Math problem
- A new antibiotic discovered by AI lab
- 722 AI written Math papers released at once
- A solution to AI alignment
- Google reducing errors 40× in quantum computing, making it practical
- Opus 5 level AI running fully offline
- Opus 5.5 solving motion graphics and automations.
And that's just in the last 5 weeks. IYNYK.
Incredible time to be alive






