Keep Up to Date
The technology gaps that matter — and what to watch.
By Justin — ProtoForge Technologies
Not everything is an AI problem. Some gaps need different technology. Some need better process. Here's what's genuinely worth watching — and where the market hasn't caught up yet.
Physical templates — 3D scanning, not AI
That steel template on the bench with the complex profile and six holes? That's a scanning problem, not an AI problem. Drafting tables are fragile, need constant recalibration, and when they drift they don't tell you.
Phone-based scanning (LiDAR) isn't accurate enough for CNC. Handheld 3D scanners — Revopoint MetroX, Creality CR-Scan Raptor — can reach 0.02–0.03mm accuracy, but they output 3D mesh, not DXF. You need QuickSurface to extract clean profiles from that mesh. Logic Trace digitising systems are purpose-built for this: trace with a stylus, get DXF directly.
Each option in the chain solves part of the problem. The workflow still requires manual steps and software that most cutting shops don't have. Worth watching as handheld scanner software matures.
PDF-to-DXF conversion — the gap nobody's solving
The biggest surprise in this space: every major CAD vendor is adding AI for 3D-to-2D — generating drawings from models. They're looking the wrong direction.
The "receive a PDF, convert to clean DXF for nesting" workflow is invisible to them. Nesting vendors expect clean DXF input. Nobody's solving intake.
The closest options right now: Scan2CAD (~$49–89 USD/month, replaces splines with true arcs) and Print2CAD 2026 AI (~$300–400 USD, arc-with-bulge tracing). Both help. Neither is a complete solution.
That's a genuine gap in the market. If someone builds it properly — PDF in, clean DXF out, no manual cleanup — cutting shops will pay for it immediately.
Nesting quality — experience, not algorithms
Auto-nesting is improving. The software can pack parts efficiently and generate reasonable cut paths. But the gap between software output and an experienced programmer's output isn't about computation — it's about physical reality.
How parts unload. Skeleton integrity. Cut direction relative to grain and heat distortion. These things come from years of walking to the floor and watching what actually happens. The software doesn't know what the programmer knows.
Could you imagine a nesting evaluator — something that scores a nest against physical criteria and flags issues before the job runs? Maybe. But anything slow or intrusive won't survive in production. The workflow is already under time pressure. Tools that add friction get switched off.
This one is worth watching as AI reasoning gets better. It's not there yet.
These gaps are closing.
I'll tell you when they do. No spam — just updates when something genuinely changes.