待翻譯：Show HN: An "Evidence Loop" for steering the agents what to build

AI 服務暫時不可用，以下為來源正文，待恢復後補全翻譯。

Agents can build anything. Dipio tells them what to build. Great products aren’t guessed. They’re diagnosed. Dipio interviews your users, applies behavioural science to pinpoint what to build, and drafts the specs your agents and engineers ship. Research that ships → Companies that ship with us Dipio Evidence Loop One loop, from guesses to shipped, evidence-backed features users adopt. The Dipio Evidence Loop is an agentic loop: user needs become evidence, evidence becomes specs, and agents ship them. It collapses time to value, so an engineer, a product manager, or a researcher can ship what users actually need. 01SIMULATE Twins Studio builds synthetic twins, working models of your real users, that run the interview step, so evidence arrives in minutes, not weeks. 01INTERVIEW An AI interviewer runs deep, structured voice conversations with your real users, no moderator needed. 02DIAGNOSE The diagnosis pinpoints what blocks your users and what to build next, grounded in psychological and behavioural science. 03SPEC The diagnosis becomes an implementation-ready spec, grounded in traceable evidence and ready for your coding agents. 04SHIP Your coding agents and engineers build straight from the spec, with no translation step in between. 05LEARN Every release ships as a measurable experiment, and the results become the next question to study. the next question the next question loops back to Interview Who runs each step INTERVIEW Dipio’s AI interviewer runs every conversation, no moderator needed. DIAGNOSE Dipio’s behavioural science pinpoints what blocks your users and what to build. SHIP Your coding agents pull approved specs through the Dipio MCP and ship them. LEARN Your product, in production results feed the next study. Methodologies Know which study to run. Three AI-moderated methodologies feed the loop. Each one starts from a different kind of question. You have an open question Discovery Interviews Jobs-to-be-Done conversations with people who bought, used, or left your product. The interviewer reconstructs the full decision story: what was bothering them, what they tried, what tipped them over. Start here to learn why users really choose you, why they leave, and whether you solve a real problem. A behaviour isn't happening Behavioural Change Diagnostics Users sign up but don't activate; a shipped feature nobody adopts; a habit that doesn't stick. Interviews with people who do and don't do the behaviour diagnose what drives the difference, and the report recommends what to change, grounded in behavioural science. You know what to ask Guided Surveys Write your own questions. The interviewer asks them in order, follows up for specifics, and never goes off-script. Fits pre-product research, feedback on a specific feature, and any structured conversation that follows your script at interview depth. Dipio Evidence Gate Your coding agent builds from the evidence, and traces every change back to it. This is where your agents enter the loop: they connect to the Dipio Evidence Gate natively through the Dipio MCP. A human in the loop - reviews and publishes every draft. Read the evidence Full findings with verbatim quotes, each tied to the specific interview it came from. Your agent builds on evidence it can trace straight back to its source. Pull the spec A spec crafted in evidence: requirements, success criteria, and cautions. Your coding agent calls the Dipio MCP as it reasons about the feature it's going to build and ship. Draft the next study When the research runs out, the agent drafts the study that would answer the open question. A human reviews and publishes. Get started freeSee the tools → Connect from Settings → Agents after signup. Works with Claude Code, Cursor, Conductor, and any MCP client. In practice Know your users. Then ship the right thing. What product, growth, and engineering teams run through the Dipio Evidence Loop today. Product Decide what to build next Turn interviews into a ranked, evidence-backed view of what your users actually need, and go from open question to shipped feature in one cycle, not one quarter. Engineering Ship straight from the evidence Your coding agent pulls specs crafted in evidence through the Evidence Gate and builds them, with the research each change stands on attached. Research Peer-reviewed science underneath. Our first paper is in press at Behavioural Public Policy (Cambridge University Press), independent proof that our approach works. The work is grounded in behavioural science from the UK’s #1 ranked university in Psychology, the London School of Economics and Political Science. Paper I In press Can AI represent the way real people think and decide? We compared Dipio’s synthetic twins against real human responses, across three products and four leading AI models. The twins matched real humans closely enough to act as a reliable stand-in for them. r = .81–.91very strong 00.51.0 · perfect Correlation between synthetic twins and real human responses, across 3 products and 4 frontier models. Early access The next Evidence Loop runs at simulation speed. We are building synthetic twins of your userbase that take the interview step, so evidence arrives in minutes, not weeks. The correlation results above are the foundation. Twins Studio, the product built on this research, is in early access. From matching outputs to matching minds. Our research goes a layer deeper. Today’s synthetic agents can match what people say. The next generation needs to match how people think, the psychological building blocks of human decision-making: traits, values, heuristics, biases. We are developing a new method to make synthetic twins reason from the same psychological foundations as the real humans they represent, not just mimic their outputs. Get started Ready to build from evidence? Get started free