Dirac, Inc.

Your customer doesn't experience your factory. They experience your lead time. They don't see the rework loop, the change order stuck in someone's inbox, or the line waiting on a revision nobody pushed yet. They see one thing: the gap between when they ordered and when it showed up. Every internal coordination problem eventually gets paid for at the loading dock. A delay that started as a small misalignment between engineering and the floor becomes a date you have to apologize for. This is why "operational" problems are never just operational. The speed you can coordinate internally becomes the speed you can promise externally. Your org chart shows up in your delivery dates whether you want it to or not. The companies that tighten the loop inside are the ones that get to make promises their competitors can't.

Every engineering change order starts a clock that nobody can see. The change gets approved in the PLM. Now it has to reach the floor. It travels through emails, meetings, updated drawings, a planner who has to re-sequence the work, and a supervisor who has to brief the line. Somewhere in that chain, days disappear. Meanwhile, parts are still being built based on the old revision because the new intent hasn't physically arrived yet. Most teams measure the change itself. Almost nobody measures the propagation time: the lag between "decided" and "actually happening on the floor." That lag is one of the truest measures of how fast a company can really move. Shrink it, and everything downstream gets faster.

A work instruction is a snapshot of how someone thought a part should be built, frozen at the moment it was written. Then the design changes. The fixture gets swapped. An operator finds a faster way and never tells anyone. The instruction stays exactly as it was while the work moves on, until nobody on the floor bothers to follow it. Walk any floor and you'll find the real instructions living somewhere else. A sticky note on the machine. A senior tech who "just knows." A photo on someone's phone from the last time they ran this job. The document says one thing. The build says another. And the gap between them is invisible until a new hire follows the document exactly and gets it wrong. Instructions that update when the design updates are the difference between a company that gets smarter over time and one that resets when your best people retire.

In a typical hardware company, three systems claim to know what's happening: the PLM, the ERP, and the MES. They don't agree. PLM knows what was designed. ERP knows what was bought. MES knows what was built. Each was implemented at a different time by a different team for a different reason. The technician at station 14, who actually needs to know which revision of the bracket to install at 9:13am on a Tuesday, is left to triangulate. That triangulation is where the cost lives. A single source of truth is one of those phrases that gets thrown around without meaning much. Done seriously, it requires a fundamental rethink of which system is upstream of which. The companies willing to do that work will out-coordinate everyone else in their category.

Today, we’re thrilled to announce Smart Model-Based Revisions, now live in BuildOS. In every other manufacturing tool, a CAD revision means starting over: New model in, new work instructions out, and hours of re-authoring the work you've already done. We think that's a broken way to sync design and manufacturing. So we fixed it. Smart Model-Based Revision changes how work instructions evolve with your model: → Approve your first work instruction once. It becomes a locked, traceable baseline. → Import the next CAD revision in a single click. → BuildOS automatically maps what stayed the same, flags what changed, and carries the rest of your production plan forward. Manufacturing engineers go from re-authoring to reviewing. The grunt work disappears. This is a fundamental shift in enabling manufacturing teams to stay focused on building great things.

Your digital systems think a part is done. Your floor knows it isn't. There's a quiet gap in almost every factory between what the system believes is happening and what's actually happening. The ERP says the order shipped, but reality says it's sitting in rework. The plan assumes the standard cycle time, and the operator knows the real one and quietly absorbs the difference. Everyone adapts around this gap. Managers walk the floor to find out what's "really" going on, and trust shifts from the system to a few people who can translate between the two worlds. That gap is expensive. Closing it doesn’t take more sensors or more dashboards. All it takes is the system and the floor aligning when the part is “done”. When they agree, you stop sending people to the floor to find out what’s “really” happening.

Proud to see Dirac, Inc. Co-founder & CEO Filip Aronshtein on this panel. Physical AI is the next layer of manufacturing software. The factory floor of the next decade will run on context-aware production planning: systems that read what's actually happening on the line, replan as it changes, and earn the trust of the people running them. Glad we were part of the conversation.

Most manufacturers are sitting on a data goldmine, and they don’t even know it. Every machine cycle, every quality check, and every assembly step generates a signal. But because systems are isolated, most of that data vanishes into thin air. The shop floor produces more operational data than almost any other environment on earth. The problem is connection. When those signals finally reach the decision-makers in engineering, planning, and procurement, lead times shrink, defects are caught before they escape, engineers stop guessing and start knowing. The factories that figure this out will learn faster than any competitor can keep up with.

Your most important factory asset is invisible. Very few companies can tell you how much manufacturing knowledge they own, yet process knowledge is vital: ! The fixture setup that saves 18 minutes. ! The assembly order that prevents downstream defects. ! The inspection pattern that catches failures before rework. ! The torque sequence the veteran technician “just knows.” These decisions compound across every unit shipped. Most organizations store them in people’s heads, email chains, static documents, and tribal memory. That means the asset disappears every time someone leaves. The next generation of industrial companies will treat manufacturing knowledge like software: versioned, structured, searchable, and continuously improving. That shift will redefine industrial competitiveness.

The factory is becoming a software problem. The biggest bottleneck in manufacturing is coordination. Modern factories already have the robots, CNCs, scanners, sensors, and ERP systems. What they don’t have is a unified operational layer connecting the intent to physical execution. That gap appears everywhere: engineering changes that take weeks to propagate, work instructions written from scratch, tribal knowledge trapped in pdfs and inboxes, or production delays caused by confusion. Software solved this problem for code decades ago. Manufacturing is next. The companies that win this decade will have the fastest path from design to production to iteration. That’s a software problem now.