Opportunities
Algorithmic CNC Scrap Reduction
Connected through 6 “incumbent in” links and 1 “applies thesis” link.
Structure
Opportunities
Opportunities
Connected through 6 “incumbent in” links and 1 “applies thesis” link.
Structure
Build difficulty
Hardest Part
Extracting and normalizing high-frequency spindle telemetry from heavily fragmented, proprietary CNC controllers like Fanuc and Siemens without causing latency in the machine's primary control loop.
Min Viable Scope
V1 passively ingests machine telemetry and offline G-code to recommend manual toolpath adjustments for aluminum milling on a single machine brand. Deliberately leave out real-time, closed-loop machine control to eliminate catastrophic crash liabilities.
Cold Start Problem
The models require thousands of hours of mapped toolpaths and failure states to predict scrap triggers accurately. Break this by deploying passive edge sensors to a single mid-market design partner running high-volume, low-mix parts in a single material to force early data density.
Time To First Value
2 to 4 weeks of baseline data collection followed by the first offline toolpath optimization report
Data Moat Available
true
Technical Difficulty
High
Build profile
The gap
Wedge
Start with 5-axis aerospace machining shops cutting high-hardness metals like titanium and Inconel. These materials wear tools rapidly and cost thousands of dollars per billet, meaning a single saved part provides an immediate, indisputable ROI. Expand from high-value exotics into general aluminum and steel high-mix milling, and eventually into automated quoting based on historical shop-floor yield data.
Timing
Edge computing hardware is now cheap enough to run local inference directly at the machine tool. Simultaneously, time-series AI models can process acoustic and spindle-load data with sub-second latency, enabling real-time control loop adjustments that were impossible with previous cloud-reliant architectures.
Why This ICP
Mid-sized job shops run high-mix, low-volume production where dialing in a new part causes disproportionate scrap and eats directly into thin operating margins. They lack the dedicated continuous-improvement engineering teams of large Tier 1 suppliers, making an out-of-the-box algorithmic solution highly attractive.
Size Of Prize
There are roughly 25,000 mid-sized precision machining and job shops in the US. If each shop spends an average of $60,000 annually on scrap mitigation, lost material, and wasted spindle time, the total addressable prize is $1.5B.
Gap Narrative
CNC machine shops lose significant margin to scrap parts caused by tool wear, thermal expansion, and raw material inconsistencies. Current CAM software sets static tool paths that cannot adapt mid-cycle to changing physical conditions at the cutting tool. Machine operators lack the systems to dynamically adjust feed rates and spindle speeds on the fly to save out-of-tolerance parts before they are ruined.
Defensibility
Defensibility compounds through a proprietary dataset of tool-wear and acoustic signatures mapped across thousands of machine, tool, and material combinations. As the system observes more cuts across the customer base, its predictive models for feed-and-speed adjustments become mathematically superior to any new entrant. This creates strict workflow lock-in, as ripping out the software directly equates to degrading shop-floor yield.
Why This Thesis
An Agentic software layer sitting on top of existing CNC controllers fits this gap because the problem requires real-time read-and-write capabilities. The agent acts as an automated operator, reading sensor data and writing G-code overrides dynamically, turning a legacy hardware problem into a scalable software service.
Overview
Sized prize
IllustrativeIllustrative targets and order-of-magnitude estimates — not an achieved track record. This Thing is concept-stage; real figures come from live data once operating.
SAM
~$300-400M for North American mid-market precision machine shops
SOM
~$15-30M
TAM
~40k global precision machining shops x ~$25k/yr algorithmic scrap reduction software spend ≈ ~$1B
Growth Rate
~9-14%/yr, driven by tightening aerospace component tolerances and escalating specialty alloy costs
Paid Comparable Spend
~$60k-120k/yr per shop in scrapped aerospace-grade materials, manual inspection labor, and legacy tool-path simulation licenses
Market sizing
How you know
Kill Thresholds
Leading Metrics
What Proves Right
Mid-market precision machining shops deploy the software and execute the algorithmic toolpaths directly on the shop floor without manual edits. Customers achieve a measurable reduction in aerospace-grade material scrap within their first three production runs. The product secures a $25k annual contract value by proving an immediate 2x return on investment against historical material waste and legacy simulation software costs.
What Proves Wrong
Machinists reject the generated toolpaths due to fears of machine crashes, resulting in manual override rates that negate the efficiency gains. The software requires extensive custom configuration for each legacy CNC controller, which balloons onboarding times and destroys unit economics. The measured material savings on the shop floor fail to outpace the software subscription cost, driving pilot churn.
Win conditions