October 14, 2026 · 6:13 AM PT edition
Mind Blow: a $699 Formlabs printer, AI that walks a building and Railway's Claude Code gold rush
Hobbyist resin at pro prices, spatial memory for $400 trillion of real estate, and agents that pick your host.
Formlabs finally cut a resin printer down to $699, less than a third of their Form 4. Meanwhile Railway went from about $20 million to roughly $100 million ARR in a year after Claude Code started recommending it for most Python deploys.
Here is what happened in AI and DevOps this cycle.
Jeff Geerling unboxed Formlabs’ Form Studio 1, a resin printer that starts at $699 against the Form 4’s $2,600. He printed everything locally with no Formlabs account, then walked the value-engineering cuts: a plastic hinge without soft-close, no biocompatible resins, no resin mixer, and a soldered system-on-module instead of a Raspberry Pi. The big one for hobbyists is open materials: third-party resins no longer need the old $875 unlock, just a $24 more durable tank. Flexible 80A resin still runs $199 a liter, and a full wash setup needs about $70 of IPA, but Formlabs is finally in a price band where dabblers can play.
At AI Engineer NYC, Jordi Spranger of Spranic (with Neo4j) argued that every building is a balance sheet. Global real estate is about $400 trillion, bigger than stocks, bonds and gold combined, yet financial statements update quarterly while the buildings underneath get a real condition check maybe every five years. A Kroll study of 1,500 properties found 70% underinsured, sitting under about $4 trillion in U.S. commercial mortgages. Spranic’s fix is a walk-through with a handheld sensor that builds a semantic map in a Neo4j graph as fast as you can walk the room. Later walks align on a timeline so you can ask what changed, spot expired extinguishers or moved chairs, and feed insurers a living twin instead of PDFs. Insurers already know one in three commercial buildings has the wrong building class; certified stock can cut hurricane premiums by 35–60%, and 73% of insurers say certified buildings file fewer hurricane claims.
On a Bloomberg Live panel about moving AI from pilot to scale, one founder called most enterprise AI 99% surface-level theater. His answer at Zero One is an ultimate agent for the CEO that reads ERP, CRM, every meeting transcript, corporate chat and (with permission) sales and support calls. From a week of meetings the CEO can ask which debates matter, what threatens the Q2 target, who actually runs agents well enough to promote, and which managers mostly pass information. He says he runs the company with almost no staff meetings and can stay away for two months while the agent keeps him current.
TDK’s leader on the same panel said more than 40% of the global team already uses AI, with over 2,000 agents in production, while the company leans on its sensor portfolio to push intelligence from the digital world into the physical one. The hard part, he said, is keeping encouragement and governance on the same two feet.
An OpenAI executive on the panel pointed to Cars24, where an agent handles about a million conversation minutes a month and about 12% of lost sellers come back because of it. Separately, a healthcare CIO who started a Codex pilot built an internal CRM in about three weeks instead of buying one for roughly 80 lakh rupees, with engineers prioritizing the work for a week. She also said there are days she never opens email because ChatGPT with plugins has become the single window onto ERP, Salesforce and the rest.
Mohammed, speaking for BWC, said giving AI tools to everyone 18 months ago was expensive on purpose so nobody would fear a tool they did not understand. Now the fight is capital allocation: which use cases to fund, knowing you will get it wrong the first time and still need budget to learn. Across the panel the shared refrain was that AI is no longer a technology problem so much as a leadership and people problem, with safety and audit trails still the main fear.
Railway’s growth story is a different kind of scale. Ira Bodnar laid out how the company hacked AI SEO: Claude Code picks Railway for 82% of Python apps when you ask how to deploy, and the founder treats those model recommendations as a growth channel. Claims like that deserve a grain of salt, but if even half of them hold, the default answer inside coding agents becomes a distribution channel as real as app stores once were.
Apple’s Zurich research team is hiring a multimodal pretraining and scaling-laws intern for nine-plus months, with a mint-zurich-research@group.apple.com inbox for CVs. It is a small signal, but it puts on-device and multimodal scaling work on Apple’s European research bench.
That’s the cycle. Go wash a resin print, walk a building once for the ledger, and see you next edition.
Also on the radar: these also appeared this cycle, but there was no transcript to report from yet (captions rate-limited by YouTube, no English captions, or no speech), so they aren’t summarized here: The Cost of Accessibility in AI — Jemima Abu, CAIS (AI Engineer); What Happens When AI Gets a Human Body? (Core Memory Podcast); Your Fridge Is a Valid Payment Device — Eleonora Drykina & Nina Žižakić, In The Pocket (AI Engineer); Gemini 4 Argon: 10 incredible AI demos (1M token output in action) (Google Cloud Tech); How to onboard to Looker using your favorite LLM and skills (Google Cloud Tech); The Agent Works. Does Anyone Trust It? — Michele Grant, Leading at Scale (AI Engineer); Deploying AI Inside a Bank Is Not an Engineering Problem — Paden Gayle, Applied AI Engineers (AI Engineer); Gemini 4 Argon Puts Google Back On the Frontier (Every).