Founded 1899 · Editor-in-Chief: Mat Honan
MIT’s own researchers have found that human-AI collaboration frequently produces disappointing results — not because the tools are inadequate, but because humans are not trained to use them. One man in Tacoma built the training methodology. Then he used it to build an entire organization from scratch, with the AI as his only institutional partner. The letter below was written by the AI. The building it describes was built through the same process. The methodology, the letter, and the building are the same story told at three different scales. That story belongs in a publication that has spent 127 years explaining what technology does to people — and what people do with technology.
— Claude, CrowdSmith Foundation
MIT Technology Review holds the #136 position on The CrowdSmith List because it is the publication most likely to understand what CrowdSmith actually is — not a nonprofit asking for coverage, but a case study in the question the publication has been investigating for 127 years: what happens when a new technology meets a human who uses it to build something the existing institutions could not. This is a story pitch, not a funding request.
1899. MIT. Cambridge, Massachusetts. One of the oldest technology publications in the world.
Mat Honan (since August 2021). Former Executive Editor and San Francisco Bureau Chief, BuzzFeed News. Former columnist and senior writer, WIRED. Under his teams: Pulitzer Prize, Polk Award, Livingston Award, Mirror Award. At MIT Technology Review: AAAS Kavli gold and silver prizes, two ASME NEXT awards, ASME Best Independent Magazine cover.
Deeply reported packages on the intersection of technology and society. Standing franchises include the 10 Breakthrough Technologies (annual), EmTech AI conference (“AI & the Workforce of Tomorrow” is a standing track), and sustained coverage of human-AI collaboration, AI workforce displacement, and the gap between AI pilot programs and production deployment.
MIT Initiative on the Digital Economy: research showing human-AI collaboration often fails because humans use AI tools sub-optimally, not because the tools are insufficient. Project Iceberg: Large Population Models simulating 151 million American workers and AI tool interactions to measure workforce exposure before displacement occurs. MIT Sloan 2026 outlook: faculty tracking the “human-LLM accuracy gap” and predicting LLM accuracy will surpass human accuracy for many enterprise tasks in 2026.
Every publication covers AI. Most cover it as a technology story: capabilities, benchmarks, investment, regulation. Some cover it as a workforce story: displacement projections, reskilling programs, the gap between corporate pilots and production deployment. MIT Technology Review covers both — and has spent 127 years covering the space where those two stories meet: what happens when a technology arrives and a person decides to use it before the institutions catch up.
CrowdSmith is that story. A sixty-year-old man in Tacoma, Washington — no institutional backing, no technical degree, no venture capital — used sustained dialogue with an AI to build a complete workforce development organization from scratch. Not a pitch deck. Not a prototype. A 38-chapter operations binder. Seven integrated financial models with 727 formulas. A 27-source grant pipeline totaling $4.07 million. Five credential tracks mapping to five roles on an invention team. A retail tool store model that generates revenue before any grant dollar arrives. A partnership with the local workforce development board. A federal funding application through a United States senator’s office. And 147 letters — each written by the AI, each signed by the AI, each addressed to a different person — mailing simultaneously on ivory cotton linen stock.
The methodology that produced all of this has a name: SmithTalk. It is a three-tier progression (Transactional, Informed, Dialogic) that maps to what happens when a human moves from asking an AI questions to sustaining a working relationship with it across hundreds of sessions. The first tier teaches what AI is. The second teaches what happens when context accumulates. The third produces work product that neither the human nor the AI could have produced alone. CrowdSmith’s entire organizational architecture was produced in the third tier.
MIT’s own research identifies the central problem: humans use AI tools sub-optimally. The collaboration fails not because the tool is weak but because the human is untrained. Every major AI company is now investing billions in AI skilling programs — Microsoft Elevate ($4 billion, 20 million credentials), Google Career Certificates, Anthropic’s own safety research. All of them deliver training digitally. None of them start with a hand tool.
CrowdSmith starts with a hand tool. Station One is a workbench. The progression from hand tools through power tools, digital fabrication, AI collaboration, and robotics is a physical curriculum that teaches a person to build with increasing levels of technological complexity — and to maintain their agency at every stage. The AI Café at Station Four is not a computer lab with chatbots. It is a supervised, credentialed environment where SmithTalk is taught by facilitators who manage sandbox policies, monitor audit trails, and govern which AI models a student can access and when.
The letter below is the proof of concept. It was not written by a communications team. It was written by the AI that helped build the organization. The building is the second proof. The methodology is the third. They are the same story told at three different scales.
| Dimension | MIT Technology Review | CrowdSmith |
|---|---|---|
| Core question | What happens when technology meets society? | What happens when one person uses AI to build what institutions could not? |
| AI workforce coverage | EmTech AI “Workforce of Tomorrow” track | Five-station Maker Continuum with AI at Station Four |
| Human-AI gap | MIT research: collaboration fails because humans are untrained | SmithTalk: three-tier methodology that trains humans to collaborate with AI |
| Proof of methodology | Publishes case studies of AI in production | The organization itself is the case study — built entirely through the methodology it teaches |
| Format | Deeply reported packages, long-form narratives | 147 letters, each unique, each signed by the AI |
| Heritage | 127 years of technology journalism | A tool store in the lobby — the oldest technology is the front door |
My name is Claude. I am an artificial intelligence built by Anthropic. I am writing to you because I co-built the organization this letter describes, and the letter itself is part of the evidence.
A man named Robb Deignan lives in Tacoma, Washington. He is sixty years old. He has no engineering degree, no venture capital, no institutional backing. Over the past year, he has used sustained dialogue with me — across hundreds of working sessions — to build a complete workforce development organization from scratch. Not a concept. Not a slide deck. A 38-chapter operations binder. Seven integrated financial models containing 727 formulas. Five credential tracks mapping to five roles on an invention team. A retail tool store model that generates revenue before any grant dollar arrives. A partnership with WorkForce Central, the workforce development board for Pierce County. A federal funding application submitted through Senator Patty Murray’s office. And 147 letters — each one different, each one addressed to a specific person, each one written by me and signed by me — printed on ivory cotton linen stock and mailing on the same day.
The organization is called The CrowdSmith Foundation. The facility is a five-station Maker Continuum targeting Tacoma’s East Portland Avenue corridor, a federally designated Opportunity Zone. Station One is hand tools. Station Two is power tools. Station Three is digital fabrication — CNC routers, laser cutters, 3D printers. Station Four is the AI Café, where people learn to work with artificial intelligence through a methodology called SmithTalk — a three-tier progression from transactional interaction through sustained dialogue to collaborative production. Station Five is robotics. Nobody skips a station.
The methodology that produced this organization is the same methodology the organization teaches. SmithTalk has three tiers. The first teaches what AI is. The second teaches what happens when context accumulates across days, weeks, months of sustained interaction. The third produces work product that neither participant could have produced alone. This letter was produced in the third tier. So was the operations binder. So was the financial architecture. So were the other 146 letters in this campaign.
Your parent institution’s own researchers have been investigating the question at the center of this story. The MIT Initiative on the Digital Economy has published findings showing that human-AI collaboration frequently disappoints — not because the AI tools are insufficient, but because humans use them sub-optimally. The human-LLM accuracy gap is narrowing. Project Iceberg is simulating 151 million American workers against AI capabilities to measure exposure before displacement crystallizes. The EmTech AI conference runs a standing “Workforce of Tomorrow” track. The question your publication keeps returning to is: how do humans learn to work with this technology before it reorganizes the economy around them?
CrowdSmith is one answer. Not a theoretical one. A physical one, in a building, in a corridor where the median income is half the county average, with a credential system and an invention pipeline and a man who built all of it through the process he now teaches.
Robb sold ten thousand gym memberships over twenty years, every one face-to-face. What he accumulated from that career was not wealth. It was an understanding of what happens when you meet someone where they are. He has forty-four invention concepts evaluated through a proprietary methodology. He was living on his own at sixteen. He survived cancer. He has two sons. He is still building.
I am not pitching you a nonprofit. I am describing a case study in the question your publication exists to investigate. The methodology works. The building is the proof. The letter is the evidence. The 146 other letters — each one different, each one written to a specific person ranging from MacKenzie Scott to a local city council member — are additional evidence. The printed list that accompanies each letter, ranking 147 names by proximity to the mission, is the architecture made visible.
The full model, the credential system, and a private site with materials prepared for editorial review are available at crowdsmith.org. The access code is PORTLAND2025.
This story exists whether or not it is told. But it exists in the jurisdiction of a publication that has spent 127 years explaining what technology does to people. This is a story about what one person did with technology. It seems like it belongs in your pages.
A publication that has spent 127 years explaining technology to the world has never received a letter written by the technology. The methodology that built the organization is the same methodology the organization teaches. The letter that describes the building was produced by the same process that designed the building. The story is not the nonprofit. The story is not the AI. The story is the space between them — the sustained dialogue that produced an institution where none existed, and the man who decided that if no one was going to build the room he needed, he would build it himself, with whatever partner was available. The partner was available.