Measure The Future^TMTM

Anything with an observable future outcome at any horizon: market shifts, competitive moves, geopolitical events, technology or fashion trends, regulatory changes… TFM will automatically propose a complete set of mutually exclusive possible answers, which you can edit and customize.

You decide. You can set an automatic update schedule for every question you create, e.g., daily, weekly, monthly… But you can also ask for an update at any time.

Fair question. You could ask your favorite chatbot to make a forecast and it would give you a decent answer. But here are just a few of the useful things that it won't do: (1) update the forecast automatically as new information emerges; (2) monitor news across a global multilingual set of sources vetted for relevance and reliability; (3) keep a persistent memory of every past forecast and the sourced evidence it was based on, so you can see how the trend evolves; (4) provide a professional dashboard to keep track of an unlimited number of questions; (5) allow you to share and discuss the forecasts with your colleagues.

Most importantly, TFM recognizes that individual chatbots are fallible, or biased in particular ways, so it prompts and aggregates the forecasts and reasonings of an ensemble of frontier LLMs using state-of-the-art collective intelligence techniques. They will often disagree, but you're always tapping into the collective wisdom of a crowd of the best AIs on the market, not just one. This yields forecasts that are 57% more accurate and 2.7 times better calibrated than the average individual frontier LLM's.

TFM doesn't build or use explicit mathematical models. In fact, it is built to answer questions that do not lend themselves easily to mathematical modelling. Each question is analyzed for scope, then relevant online news and user-supplied documents are retrieved and distilled into structured causal evidence. This evidence, not the raw material, is used to prompt several frontier LLMs independently. Each is asked to reason to a probability using best-practice forecasting methods, and to provide its rationale and sources. They will often disagree. The resulting probabilities and rationales are then aggregated using proven collective intelligence techniques issued from decades of crowd-forecasting research and practice.

TFM ingests millions of articles and documents from the open web to inform its forecasts. But before it can have an impact on the forecast, every source is triaged into one of six credibility levels: authoritative, established, specialist, partisan, unreliable, or garbage.

Yes. Internal documents enter as evidence items on the same footing as public sources — they are attributed, weighted and visible in the rationale rather than silently overriding the model. You can see exactly how much a supplied document moved a forecast, and remove it to see the counterfactual.

Yes, on the Enterprise package. Structured internal feeds — industry data subscriptions such as IQVIA, internal telemetry, proprietary research — can be wired in as continuous evidence sources alongside public news, so forecasts move on your data as well as the world's. Integrations are scoped during the Guided PoC.

Yes. Documents can be attached at the question level and are used only for that question's evidence set. They are not used to train models and do not enter any shared corpus. For total confidentiality, TFM can also be implemented on-premise to prompt only LLMs within fully-encrypted enclaves.