RESEARCH
Long-form work on testing industrial technology bets before they are funded. Each paper is published in full, with no gate and no form.
How to build a rollout case the finance team will sign: which site conditions to test, which pilot benefits travel, who pays each year, and how to phase the approval.
Validating Equipment-as-a-Service before the OEM carries the risk: measurement disputes, serviceability across the installed base, renewal evidence and the transition inside the OEM.
Testing a partner route into industrial buyers: whether integrators and distributors hold the relationship, can implement, earn enough, avoid conflict and stay accountable.
Commercial due diligence on industrial technology targets: which thesis claims to test with buyers, how to grade them, and what a short diligence window can and cannot evidence.
How a board decides the next tranche for a stalled portfolio growth bet: the evidence each decision needs, a Reset that tests something new, and a Stop that is not about people.
Why industrial pilots succeed and never convert, why the people inside cannot make the call, and what a structured decision, including a defensible Stop, requires.
Why industrial deals stall after the evaluation is won: the champion is not the buyer, polite interest is not a signal, and the seats that stop a deal can be mapped before the sales motion is built.
Physical AI fails the way software does, on an untested commercial hypothesis, and adds three questions software never had to score: the liability model, the sim-to-real gap at the buyer's site, and operational continuity.
Notes from agentic systems we run on our own work: where approval gates belong, why autonomy is set per task from outside the agent, and why controls sit on the dispatch path.
Agents that move workloads, recover from faults and patch nodes at the industrial edge are technically possible. This paper sets out what a buyer has to verify before letting them act, starting with whether the plant's security and OT teams will allow it without approval.
What blocks industrial edge deployments, from fragmented systems and skills gaps to scale, security and brownfield equipment, and what a workable architecture looks like.
What a turnkey edge platform for software-defined industry would have to provide before a plant could run it without a specialist team on site.
The terms that apply when you use thing.company, the papers and the free toolkit.
The free toolkit turns these papers into self-check tools that run inside Claude, ChatGPT, Codex and other AI assistants, organized by the moment your decision is in.
A Sprint Assessment is thirty minutes on the bet you are least sure about.
Or reach us directly at hello@thing.company