Jack Bookey

vibe cutting @ NOX METALS | industrialist | ex-vc

[Jack Bookey @JackIsLearning]

@zanehengsperger generational aura loss

[Chris Powers @fortworthchris]

I'm going to sit down with @zanehengsperger of NOX Metals soon for a conversation.

Nox is building AI-powered factories that are supplying American manufacturers with metals.

"NOX METALS exists so America can build 100x more factories and technologically abundant industrial capacity in the West."

They recently raised an $11.5M seed round with @PalmerLuckey, @ycombinator, @DTXVentures, and several other great investors on the cap table.

What would you like to know?

[Jack Bookey @JackIsLearning]

if you don't have a real answer to this you're soon to be obsolete

[Christian Keil @pronounced_kyle]

New interview question: what important truth do very few LLMs agree with you on?

[Jack Bookey @JackIsLearning]

@zanehengsperger Steve is a goat. Meanwhile im there in npc mode

[Jack Bookey @JackIsLearning]

Not a lot of suppliers are stoked to support high mix + low volume operations with aggressive lead times. NOX is tho!!

[NOX METALS @noxmetals]

We sat down with Taylor Loehr of Precision Mold & Machining Services to see how Nox Metals has helped his business speed up production and help reindustrialize America at light speed.🇺🇸

[Jack Bookey @JackIsLearning]

The best factories will optimize for the ingress of telemetry and the egress of intelligence to the operational edge

It's cool to have an intelligent ERP, but if it just shuttles data back and forth to your WMS, which then is used by your shift lead who then has to disseminate that knowledge to your technicians- at 80% fidelity per hop then 26% signal survives the company brain<>factory floor round trip

A single schema spans our online instant quotes, our erp, and our edge deployed FACTOS stations. fidelitymaxxing @noxmetals

[Jack Bookey @JackIsLearning]

@ivanvnucec @zanehengsperger Since 2d nesting is an NP-hard problem, we can't verify we have the optimal solution, but we can get highly confident that we're close

[Jack Bookey @JackIsLearning]

good trends !

[Zane Hengsperger @zanehengsperger]

our new quoting/nesting software is ~100x faster than it was 3 months ago ...

10^76 possible combinations to cut this job

there are roughly 10^78 atoms in the universe

and we found one that approximates to the optimal nest

every single person in the history of the world who has done this before us has done with spreadsheets or by hand, taking hours or days... or just a grossly suboptimal solution, that american manufacturers end up paying more for

less than 30 seconds for this to run

a few months ago this would have been impossible for us

nox nesting benchmarks soon

[Jack Bookey @JackIsLearning]

@hyperdima @zanehengsperger fable def shrinks the timelines to apply our proprietary research to the factory floor !

[Jack Bookey @JackIsLearning]

@TuelTaylor @zodiaceng 💯💯💯

[Jack Bookey @JackIsLearning]

The bin packing problem is why we will win

The goal is traditionally to nest a set of items into the smallest amount of space. In our case it's cutting orders from our inventory of aluminum plate. It's an np-hard (computationally annoying) problem and at scale you can't brute force an optimal solution

A greedy solution is to just take the biggest remaining item and place it in the container over and over. Other methods branch out to test various possibilities, judging and pruning branches as they go

At @noxmetals took this hard problem and made it even harder. Our nesting engine considers raw material util (most stop here), forecasted util of the leftover material (compounds with more order data), saw cut count, load balancing across saws, and rush order prioritization

Agents tune the weights of these criteria based on context from our in-house dev'd ERP and then dispatch six unique 2d packing algos. We run multiple waves with various seeds until either the solutions converge, a target objective fn value is met, or we hit a wall clock timeout. Then cut sheets are made and our operators get to work

High material util and fewer saw cuts -> lower cogs and lead time -> faster and cheaper metal to American factories

[Jack Bookey @JackIsLearning]

@zodiaceng That'd work too if we have the inv!

[Jack Bookey @JackIsLearning]

@zodiaceng Hey Ken. We'd cut it from 2" thick plate

[Jack Bookey @JackIsLearning]

average working experience @noxmetals

[Jack Bookey @JackIsLearning]

@noxmetals Absolute cinema

[NOX METALS @noxmetals]

The path a metal block takes on its way to reindustrialize America 🇺🇸

[Zane Hengsperger @zanehengsperger]

building Saas company. am I doing it right?

[Jack Bookey @JackIsLearning]

@67Designs @GaryTan Hi. We don't sell software. FACTOS powers our metal factory here in Detroit. Hope that clears things up. You should swing by sometime!

[Jack Bookey @JackIsLearning]

I want to give you a real answer so sry for the delay- We've been developing a company brain since day one called Nest. This brain is hyper-specialized to optimize our metals business. You can look at my bin packing post for more info here

Factos forward deploys Nest intelligence to the factory floor. Instead of the latency of back-and-forth between our procurement/sales teams and our technicians, software driven decisions are proactively communicated to relevant parties. An example would be a rush order which reallocates scheduled jobs to saws and updates a forktruck operator's pick list in real time