Now on v2.8 · engine model 2.8

What’s new.

Every version of the model, dated - so you can see how it’s grown, and hold us to it.

2.82026-08-23

THE ENTITY Data — the platform, as an API

THE ENTITY Data is here — programmatic, read-only access to the four datasets that power the platform: the causal supply-chain graph, the computed engine (nowcast, forecast, conviction and node-score record), the tracked-elite smart-money panel, and market feeds & cited filings. One stable JSON envelope, one entk_live_ key, an OpenAPI 3.1 spec, and honest freshness labels on every number (as_of vs when we detected it, never conflated).
Four ways in, same key: a REST API over ~30 endpoints (filter, sort, pick fields, full-text /search); an MCP server so an AI assistant — Claude, Cursor, your own agent — can query the graph, engine and smart-money as native tools; =ENTITY() in Google Sheets and a Power Query for Excel for the spreadsheet crowd; and bulk CSV/NDJSON export to load a whole collection into a warehouse. Plus a derived screener (companies the tracked elite are accumulating that the engine also backs) and a cockpit at /app/data to mint keys, watch usage and grab quickstart snippets. Manage it in the account Add-ons; docs at /developers.

2.72026-08-21

Copilot Max — a monthly membership for Pro members that doubles your daily deep questions (from 10 to 20) and lifts every Copilot limit at once: a longer working context so it holds more of the thread and longer questions in view, room for far more of your own documents and custom feeds, and the Virtual Investment Committee for the cost of a single question instead of three. Cancel anytime.
Plus set standing custom instructions Copilot follows on every question, export any answer to a clean PDF, and get early access to new Copilot features as they land.

2.62026-08-19

Copilot now works with your own material. Attach a PDF, spreadsheet or text file — or point it at an RSS feed you trust — and it reads them alongside the live engine data, citing your document by name when it draws on it. Type @ in the prompt to reference an attached doc or one of your own holdings directly.
New ways to interrogate a thesis: ask /ic to convene a Virtual Investment Committee — three model personas argue the bull, bear and macro case over live store data before a synthesis lands, clearly labeled as committee judgment and charged as three questions. Save a thesis snapshot on any answer and reopen it later as a then-versus-now diff of your book, the node scores and prices. Pick an answer shape in Settings — a three-line Exec Brief or a sectioned Deep Chain Audit — and, when you want more depth than the daily allowance, top up Copilot credits with a one-time pack.

2.52026-08-18

Copilot arrived — ask a question in plain English and it composes a live dashboard on the spot: the right instruments, pulled from real data, with a written read to go with them. It knows your own book, remembers the thread, shows its sources and how fresh each number is, and can carry out journey actions on THE ENTITY for you — always proposed first and confirmed before anything runs, and never a trade.
Smart Money went deeper into history. The elite fund books, the insider tape, activist stakes and congressional trades now read across years of recorded filings — so an accumulation or an exit is measured against a real track record, not a single quarter.

2.42026-08-10

Smart Money grew up. We now track the 13F books of 103 verified institutional managers — the concentrated-conviction funds and the legends — and read where they are quietly accumulating versus backing away, every move dated the day we detect the filing.
One fused smart-money read on every AI stock: the tracked elite funds, the full institutional census, activist stakes, insider buys and congressional trades — each dated to its own filing and weighted into a single accumulating-to-distributing signal.

2.32026-08-09

Opened the Smart Money hub — the elite fund book and the insider tape across the whole AI universe, diffed nightly into dated events: fund entries and exits, accumulation shifts, insider buy clusters.
Insiders and funds now carry real track records. Every call is measured from the day it became public, shown with its sample size, and left blank until the record can honestly back it.

2.22026-07-17

The chain came alive: a shock on one part of the supply chain now cascades to everything downstream — and the model can call the end of a squeeze, not just the start.
Live macro across the chain — real Federal Reserve, energy and SEC-filing data now feed the constraint model — and the top "Strong Buy" call is reserved for names that own a genuine bottleneck.

2.02026-07-09

Bottlenecks became provable. The binding constraints are now the mathematical minimum cut of the entire Energy → Applications flow — not a heuristic — and backtested against real supply-chain dominoes.
Launched the public AI Value-Chain Index and per-company exposure and valuation reads, so every name is scored on where it sits in the chain and what that is worth.

1.22026-05-15

Put effect size and timing on every causal link, and added the critical materials, geographies and policies behind each node — so the model reasons about how hard and how fast, not just which way.

1.12026-04-02

The map went live: public news mapped conservatively onto the chain and decayed into the score in real time, so the picture moves with the market.

1.02026-03-01

The first release — the entire AI supply chain, mapped and scored end to end, from power to applications.