How readiness is scored
Every future technology on the tree has a technology readiness level (TRL) from 1 to 9. The scale was developed at NASA and is now used by the European Commission, the US Department of Defense and the International Energy Agency. This page explains how we apply it to hardware, software, biology and broad goals such as “artificial general intelligence”, so a score means the same thing on every page.
The rules
- Score the whole technology, not its best part. A technology is at the level of its least mature essential piece. A record-breaking component in the lab doesn’t make the system TRL 6.
- Every score cites evidence. The score’s rationale points at milestones on the same page, and each completed milestone has a date and at least one public source. If we can’t cite it, it doesn’t count.
- Public evidence only. Claims in press releases count only when backed by a paper, a demonstration others observed, a regulator’s record or independent reporting. Unverified announcements are recorded as pending milestones, not as progress.
- When in doubt, score lower, and say why in the rationale.
- Scores are dated and never overwritten. When a score changes, the old one moves to the page’s TRL history with its date, so readers can see how a technology has progressed and how our judgement has changed.
- Future technologies score 1 to 6. Each future technology has a definition: a checkable test for its first demonstration. We write that test to correspond to roughly TRL 7, a working system prototype demonstrated in a real setting. Once the test is met and sourced, the technology moves from the future to history, so a future page never shows TRL 7 or above. If evidence seems to justify a 7, either the definition is too weak or the technology belongs in history.
The levels
The general definitions follow the European Commission’s wording. The columns show what each level looks like in practice for different kinds of technology.
| TRL | General definition | Hardware | Software and AI | Biology and medicine |
|---|---|---|---|---|
| 1 | Basic principles observed | Physical effect described in the literature | Algorithm or idea described | Mechanism hypothesised from basic research |
| 2 | Technology concept formulated | Application proposed; analysis but no experiment | Approach specified; no working code | Target identified; candidate approach proposed |
| 3 | Experimental proof of concept | Effect demonstrated in a lab experiment | Prototype shows the core idea works on a toy problem | Activity shown in cells or tissue (in vitro) |
| 4 | Technology validated in a lab | Components work together in a lab (“breadboard”) | Components integrated and tested on realistic data in the lab | Effect shown in animal models |
| 5 | Technology validated in a relevant environment | Breadboard tested in realistic conditions | Integrated system tested on real-world inputs, outside the developers’ control | Safety and manufacturing studies done; ready for first human trials |
| 6 | Technology demonstrated in a relevant environment | Full-scale or representative prototype demonstrated | Complete system demonstrated on the real task with real users, as a pilot | Early human trials (phase 1 and 2) show safety and a signal of effect |
| 7 | System prototype demonstrated in an operational environment | Prototype works in the field | System runs in real deployment and meets its stated goal | Large trials (phase 3) succeed |
| 8 | System complete and qualified | Qualified for production and use | Generally available, with its capability independently confirmed | Approved by a regulator |
| 9 | Actual system proven in an operational environment | Proven in routine use | Proven in routine, widespread use | In routine clinical use |
For biology, the levels follow the stages of drug and device development (preclinical research, then clinical trials in phases 1 to 3, then approval), as in the US FDA’s description of the drug development process.
Broad goals
Some future technologies are capabilities rather than devices: artificial general intelligence, mind uploading, molecular manufacturing. They have no single prototype to point at, so:
- The definition does the work. The page states a concrete, checkable test, such as a specific set of tasks, a benchmark under stated conditions or a measurable output, and links to a public forecast question that uses a similar test where one exists. Differences between our test and the forecast question are noted on the page.
- Score the test, not the hype. The TRL describes how close the best public system is to passing that test, using the general definitions above. Strong results on parts of the test raise the score only once they’re integrated in one system.
- Prefer milestones to adjectives. Progress is recorded as dated milestones, such as a benchmark passed or a capability shown, each with sources. The score summarises them.
Forecasts
A TRL says how mature a technology is now. When it might arrive is a separate question, answered by forecasts. Each future technology has at least one forecast, and one is marked primary; the primary forecast’s median sets where the technology sits on the tree. Forecasts come from public sources such as Metaculus, published expert surveys and Ray Kurzweil’s books, or from the maintainer, labelled as such. A maintainer’s estimate is always a range. Every forecast records when it was retrieved, and earlier values are kept, so the site can show how forecasts have moved.
When a forecast puts a technology before one of its prerequisites, the tree flags a date conflict. The technology is drawn no earlier than the prerequisite, and the conflict is shown on the tree and on the page rather than hidden.
Sources
- European Commission, Horizon 2020 work programme, General Annex G: Technology readiness levels (2014).
- NASA, Technology readiness levels, and NASA Procedural Requirements NPR 7123.1, which defines them formally.
- US Government Accountability Office, Technology Readiness Assessment Guide (GAO-20-48G, 2020).
- International Energy Agency, ETP Clean Energy Technology Guide, which scores energy technologies on an extended version of the scale (up to TRL 11, for market uptake).