AI in manufacturing enthusiast. CAD/CAM techie and weekend woodworker. 🤖🪵

Liverpool
Joined May 2023
Live and work anywhere is real. Did exactly that, moved my production career to a free zone with 100% foreign ownership. Demand doesn't care where you are, the capability is what counts.
With all the negative talk about the economy keep this in mind… 1. The only thing that makes you profitable is demand and supply tension. If supply is constrained relative to higher demand, you’ll be profitable. You’re in control of both far more than you imagine. 2. You’re alive at a time where you can live and work anywhere and you can sell to people all over the world. They key is to design a business that can maintain demand and supply tension and be part of the global economy.
13
Interesting stack. The question is always sim-to-real though. Do you know of any actual production deployments on real factory floors, or is this all benchmarks and demos?
$NBIS: Leading Neocloud for Managed Physical AI Services The field I have identified $NBIS leading in among Neoclouds is Managed Services for Physical AI. They have the best integrations of Nvidia OSMO, Nvidia Issac Sim, Nvidia Cosmos, and Nvidia Issac-GR00T and are first movers in this market. Nvidia OSMO: github.com/NVIDIA/OSMO OSMO is a workflow orchestration to manage the robotic model and reinforcement training life cycle. It's is a platform the other $NVDA other open source components listed below. Nvidia Issac Lab: github.com/isaac-sim/IsaacLa… Issac Lab is a allows for sensor simulation like cameras, LIDARs, touch sensors and allows reinforcement training in simulated environments. Nvidia Cosmos: github.com/NVIDIA/Cosmos Cosmos is a foundation world for Physical AI that many industries can build off of between autonomous vehicles, manufacturing, and agriculture. Nvidia GR00T: github.com/Nvidia/Isaac-GR00… GR00T is a multi-modal vision and language model with hooks for fine-tuning with customer robot data and examples. Conclusion NBIS is a first mover among Neoclouds for Managed Physical AI. They are the first to integrate all 4 major Nvidia open source components on top of their managed GPUs.
11
Parametric feature tree from a raw STL, running locally in the browser. Impressive work. Does it handle organic surfaces or mostly prismatic geometry? I'd love to throw a scanned part at it.
My latest free tool, ScanRuler.com, can automatically reverse-engineer your 3D Scans! Just tell your AI agent to connect to the ScanRuler MCP give it an stl and tell it to reverse engineer it. A detailed video and a custom skill will be released in the next few days!
1
25
$1.58T of AI capex in 2027 and most of the coverage is a chips story. Power, storage, networking - transformers, switchgear, chillers, grid connections. I'd call that an industrial demand event, and the equipment layer is where manufacturers actually see the orders.
11
Will Hughes retweeted
This Monday, Morgan Stanley released a report about following AI's $1.6 trillion money Here is the takeaway in images (2 more in comment) but it can be concluded as - The market is having a reset, not necessarily the end of the bull market. - Companies are still spending a crazy amount of money building the AI world and that money becomes someone else's revenue. Around $1 trillion of the $1.6 trillion 2027 AI spending could show up as revenue for S&P 500 companies. So: Microsoft/Google/Amazon spend money NVIDIA, Broadcom, Micron, networking, storage, power and other suppliers make money. - AI is becoming extremely important to the whole stock market AI stocks could drive 78% of S&P 500 earnings growth in 2027.
4
5
30
2,049
other countries: build sovereign chip production. OpenAI: we can buy from Nvidia. Another strong sign that OpenAI ngmi.
Ask Tesla, the world’s preeminent force for engineering, how easy it is to build on top of a stack that is only produced in another country. Please, stop and think for a moment about what you’re suggesting. Consider the economics as being downstream from engineering. Do you understand why Silicon Valley exists in California and not NYC or Chicago or the UK? Do not sell out our long-term dominance for the next 50 years so that we can be slightly ahead 18 months from now.
16
Will Hughes retweeted
When you're reverse engineering tiny parts, "close-enough" doesn't work, and precision is everything. The #Artec3D Micro II desktop 3D scanner delivers 5-micron accuracy to capture every micro-texture, fine edge, and organic detail automatically. #3DScanning #ReverseMetrology
3
7
262
Will Hughes retweeted
Hyundai and Kia are showcasing ~200 smart manufacturing technologies at E-FOREST TECH DAY 2026, driving a transition to Software-Defined Factories (SDFs). Highlights include factory-specific AI agents, digital twin simulations, and automated 3D vision inspection systems. #Sma...
2
1
22
If you are building the next AI CRM you are genuinely cooking your future grandchildren. You are way too smart for something meaninglessly incremental. Only things people should be building in SF: >AI labs >Compute maximization >Reindustrialization >Energy abundance >Robotics & defense >Bio >Physical infra Everything else becomes irrelevant if China wins. They are already very, very far ahead of us in energy, metals, manufacturing capacity, and the general ability to produce physical things at scale.
55
23
10
550
23,689
So the 2.5M DAU includes bought accounts? That retention number is even worse than it looks.
The demand is already there. Now $META needs to make Muse sticky. Day-one adoption isn't the problem. Retention is. If the product improves, that gap could become the opportunity. BTW, you can just buy a Muse account, very cheap too!!!
17
Will Hughes retweeted
India will build its own industrial future. Not import it. Not copy a movement happening elsewhere. @etherealmachine is pushing the boundaries of what India can manufacture, taking on some of the hardest problems in advanced manufacturing.
Kaushik Mudda, Co-founder & CEO of @etherealmachine, on India’s precision manufacturing gaps and the need for a stronger “mother machine” ecosystem to build next-generation machinery, at #TechSparks2026.
4
18
147
3,036
Did you know microreactors could one day be used to power military bases, disaster recovery efforts, or remote locations where traditional infrastructure doesn’t exist? Learn more about microreactors ⬇️ energy.gov/ne/articles/what-…
21
51
11
245
9,230
Will Hughes retweeted
there has been a breakthrough. here's a preview. more soon.
Vibe weld fixture design
6
5
1
152
8,508
Regulatory capture 101. The regulated writing the regulation.
🚨 Unpopular opinion: other people (not those working for Anthropic) should be deciding on AI safety bills.
25
1.2k followers and a better track record than the whole 50k+ influencer crowd. We all follow the wrong people on here.
Credit where Due: Predicting the AI Financing Fears When times were good in June and everybody was selling subscriptions, @liamdaltonjr made a very accurate call that there would be a financing panic in the Neoclouds in Q4/Q1 even though he was very bullish AI in the long term. Completely different from Michael Blurry who called the next 10 AI bubbles. Huge credits for Liam for calling out the financing slowdown fear that we are seeing now. His work is all free and he earns money by managing a sizable fund. I had a video call with him in June and can confirm that he has a dam good view of NYC. The guy has 1.2k followers but is a needle in the haystack of all the 50k+ influencers.
31
Other platforms suppressed it. X's own chatbot spread it 300 times. That tells you everything about the testing behind these deployments.
Her name is Lucy Craig and her false accusations have ruined numerous innocent lives
1
155
So the plan is local-content rules and a softened target instead of competitive products. ngmi.
🇫🇷🇩🇪 Paris and Berlin just traded climate ambition for industrial protection. At the European Council summit in Brussels on Friday (16 Oct), Emmanuel Macron and Friedrich Merz announced a Franco-German bargain on cars: • Germany accepts stricter “Made in Europe” / local-content rules in the proposed Industrial Accelerator Act — a longstanding French demand. • France backs softening the EU’s 2035 new-vehicle emissions target from a 100% CO2 cut vs 2021 levels to 90% — and without forcing the remaining 10% into alternative-fuel investments or offsets, as the Commission had proposed. • Joint line: a “balanced” approach to climate goals while strengthening automotive technological sovereignty, the industrial base and competitiveness. The deal slipped from Thursday after the German SPD (junior coalition partner) objected. On Friday the SPD dropped its opposition. Environment Minister Carsten Schneider (SPD) called it a “modern realignment” of German industrial policy focused on jobs and production sites. An Élysée official said Made in Europe for automotive can be extended to the UK and Switzerland under certain conditions — addressing a concern of UK PM Andy Burnham ahead of the 20 November 2026 UK–EU summit. Macron said he hopes the Franco-German dynamic leads swiftly to agreement with partners “for our jobs and our industry.” Climate target in, local-content carve-outs out. That is the deal Brussels got when Paris and Berlin decided jobs matter more than a neat 100%. Sources: POLITICO, 16 Oct 2026: politico.eu/article/macron-a… Photos: Opel/GM Gliwice assembly line — Marek Ślusarczyk (Tupungato), CC BY 3.0 (Wikimedia; cropped, brightened, sharpened). Insets: Emmanuel Macron, World Economic Forum Annual Meeting 2026, CC BY 4.0; Friedrich Merz, Michael Lucan / Presseclub München 2024, CC BY-SA 3.0 de. #EU #France #Germany #Macron #Merz #MadeInEurope #Cars #2035
22
Cloudflare is currently defending businesses against agentic systems This means they’ll have the data, plus the ability to sell custom agent permits services to websites Perhaps one of the best public market stock picks for the next 5 years Insanely valuable
32
46
6
1,445
44,051
I assumed benchmark cheating was a lab-only thing too. Turns out the 45/45 in 8.5 min was just git diff. Great catch on the transcript forensics.
Replying to @PawelHuryn
About the high effort, we caught DeepSeek cheating. I thought this happens only in research papers and deep inside AI labs. Be careful with that model! 😀
25