Rafa Schwart (@rafa_flows) header
Rafa Schwart (@rafa_flows) profile picture

Global Operations @motosolutions | Building @arqentia

The Metaverse
Joined May 2010
Photo by Rafa Schwart (@rafa_flows)Photo by Rafa Schwart (@rafa_flows)Photo by Rafa Schwart (@rafa_flows)Photo by Rafa Schwart (@rafa_flows)Photo by Rafa Schwart (@rafa_flows)Photo by Rafa Schwart (@rafa_flows)Photo by Rafa Schwart (@rafa_flows)Photo by Rafa Schwart (@rafa_flows)Photo by Rafa Schwart (@rafa_flows)Photo by Rafa Schwart (@rafa_flows)
A near miss is a free warning. I want it written down the same day. On one line, a pallet stack nearly tipped. Everyone talked about it, nobody wrote it down, and it was stacked the same way again. Where does a near miss get written down in your shop?
1
1
1
25
A promise to the customer only gets kept if the crew knows it exists. On one ramp, a packaging detail we agreed with the customer never reached the work instructions. Final inspection caught it. Where do sale promises get written down in your shop before the work starts?
1
19
Scanning every rack shows you what is there. What keeps that data useful is the mismatch. When the scan and the system disagree, one named person owns it and closes it the same shift. Who owns the mismatch on your floor? teapaw.org/lukas_m_ziegler/status…
Andrei is building robots that can scan an entire warehouse, rack by rack, and turn what they see into live inventory data. His company, @dexoryHQ, builds the hardware, the software, and increasingly the machines that make the robots. I visited their factory outside Oxford to see what that looks like in practice: an eight-metre scanning tower, CNC machines making robot parts, and a robot arm being trained to help automate the machine shop. Andrej told me why his team brings more work in-house as Dexory scales: “If you rely on someone else’s roadmap, if you rely on someone else’s timelines, you have no real power and control.” This is a tour of a full-stack robotics company, and a conversation about speed, reliability, and what it takes to build robots for the real world. 00:00 Intro 01:15 Factory tour: getting started 02:40 Continuous tower testing 04:10 Electronics and bringing components in-house 06:00 Scaling production 07:21 CNC machining and robots making robot parts 09:20 AI-driven manufacturing plans 11:00 Fixtureless machining and custom cutting tools 13:06 Robot-arm and motor reliability testing 15:54 Capabilities of the robot with the CTO 18:05 Live demo: eight-metre tower and warehouse scanning 19:00 Remote inventory images and autonomous charging ~~ ♻️ Join the weekly robotics newsletter, and never miss any news → ziegler.substack.com
26
Job shop owners: a hot job lands on Wednesday and the week is already full. Something gets bumped. What decides it in your shop: the due dates, the customer, or whoever pushes hardest?
21
Stock at the back of the shelf is the stock that expires, rusts or gets lost. The fix I'd start with is boring: write the arrival date on every box and pull the oldest first. An AI can flag aging stock for you, but only if the date made it onto the box.
1
26
A customer calls about a bad part. First thing I'd want: the material lot it came from. On the traveler, you find it fast. In someone's memory, you start guessing. AI can search every traveler, but only what got written down. Where does the lot number live in your shop?
13
A terminal across the shop gets filled in at the end of the shift, from memory. Put it at the machine and it gets filled in while the part is still there. Same with an AI tool. If using it means leaving the job, the input comes late and thin.
19
Small-batch shop owners, I want to know where a long changeover really loses its time. Hunting for tools and fixtures? Dialing in offsets? Waiting on material and the program? Which one eats most of yours?
18
Floor Words EP 02: spaghetti diagram. Follow one part through a shift and draw every move as one line. The tangles show where the time goes. AI can draw it from your walk notes. Why it doubles back, ask the operator. What's the longest walk on your floor?
27
Floor Words EP 01: golden sample. The approved unit kept near the line. You hold today's piece next to it. AI can compare photos of the two and list every visible difference. Color under your shop light, finish and feel still take a person holding both.
1
1
33
A golden sample is one finished unit checked against every requirement and kept protected near the line. In a cabinet shop it can be the sample door with the approved stain. Where does yours live, and who on the floor knows where?
7
CNC shop owners: how do you decide when to change a cutting tool? By part count, by the sound of the cut, or when a part measures out of tolerance?
1
35
Tu proveedor puede estar fabricando con la revisión B cuando tú ya aprobaste la C. La IA te ayuda a encontrar el plano. Lo que no te dice es con cuál está trabajando él.
1
51
Ahí la IA sí ayuda: encuentra el documento correcto. Y el flujo de trabajo registra si el proveedor ya confirmó qué revisión recibió. Lo que no puede hacer es aprobar la revisión. ¿Cómo sabes hoy qué revisión tiene tu proveedor?
1
5
68% of the 700 small and mid-size US manufacturers MxD surveyed reported a disruptive cyber incident in the past 24 months. More than 70% have office IT and factory systems at least partly connected. Before an AI tool gets a login on that network, map everything it can reach.
1
46