{"type":"rich","version":"1.0","provider_name":"Transistor","provider_url":"https://transistor.fm","author_name":"Manufacturing Hub","title":"Ep. 268 - David Nichols of Loupe on Claude Code, AI Retrofits, and the End of the SI Moat","html":"","width":"100%","height":180,"duration":4081,"description":"David Nichols of Loupe explains what actually changed in automation engineering this year, and why control retrofits just got far cheaper.\nNearly twenty years into building high performance controls, David Nichols is blunt about what AI did to his business: what used to take six weeks now takes two, and his team ships roughly five times more software per sprint. An engineer dropped the raw documentation from a failing robotic cell into a folder, asked Claude for a state diagram, and got a reverse engineered flow chart in thirty minutes that beat anything the customer had. The error recovery gap causing the downtime was visible right there in the chart. A SASE member handed a model a hundred page ethernet specification and got working code for every function, plus tests, on the first try. That was a six week project. It took a day.\nThe implication is the sharpest part of the conversation. Software rewrites used to be the dumbest thing an engineer could propose, because software was expensive and full of unknowns. That logic has inverted. If a rewrite takes a week and produces better documentation, better tests, and better coverage, then brownfield modernization stops being a risk calculation. His words: open season on old controllers. For an industry sitting on Windows 7 machines and undocumented cells one PC failure away from stopping a line, that is a different economic picture entirely.\nVlad walks David through the full project lifecycle, and the answers stay practical. Discovery becomes a context gathering exercise where interview transcripts and a drawer full of legacy PDFs turn into control diagrams and proposals in days. Development becomes worth treating as a real software engineering project, because source control, digital twins, and automated testing are exactly what these models are fluent in. David is equally direct about platforms: the bottleneck in automation was never technical, it is cultural. His analogy is a shop that insists on carburetors...","thumbnail_url":"https://img.transistorcdn.com/yoKAvzBXZ3YjQTekFk7KFGXeuwJ29WgXvop3dVEfhLs/rs:fill:0:0:1/w:400/h:400/q:60/mb:500000/aHR0cHM6Ly9pbWct/dXBsb2FkLXByb2R1/Y3Rpb24udHJhbnNp/c3Rvci5mbS9zaG93/LzE3MjEzLzE2MDk0/MzA1OTgtYXJ0d29y/ay5qcGc.webp","thumbnail_width":300,"thumbnail_height":300}