Breakthrough Tracker explained

What the tracker shows

Frontier fields do not move in headlines, they move in steps: Phase I data, then Phase II, then scale-up. Tokamak plasma 100 seconds, then 1,000, then 10,000. Qubit coherence from microseconds to milliseconds to seconds. A single headline about "the breakthrough" tells you little — the comparison to the previous milestones tells you everything.

That is exactly what the tracker captures. Each track is an ordered list of milestones with a clearly defined target and status: queued, in_progress, done or failed. Alongside sit the target year (when the milestone should be reached), the original pre-registration date (when the claim was first staked), and the sources backing the status.

Tracks exist today for biotech therapeutics (Phase I to market approval), fusion programmes (plasma stability, tritium net-energy), open-weight models (benchmark frontier), and longevity trials. New fields come in via proposals — see below.

Anatomy of a milestone

A milestone has five fields, nothing more: name (what must happen), target value (in which unit), target year (deadline), current status, plus at least one source URL. When a milestone moves to done, the achievement date is added; if it moves to failed, the failure date with reasoning.

Concrete example from fusion: track "SPARC commercial wall-plug energy". Milestone 1 — "First plasma": target 2025, status done (2025-09-12, source: CFS press release + MIT audit). Milestone 2 — "Q>1 demonstrated": target 2027, status in_progress (ongoing plasma runs). Milestone 3 — "Net wall-plug positive": target 2030, status queued. Milestone 4 — "Commercial prototype": target 2035, status queued.

The read is simple: anyone who looks at the track knows in 30 seconds how far the programme sits from the commercial goal. Press releases that shout "breakthrough" without naming the next milestone are immediately flagged as suspect.

Why milestones, not stories

Classical tech coverage runs in story mode: one headline per event, context optional. That works as long as the reader already knows the background — for frontier tech, that is rarely the case. Anyone hearing about spin-photon interfaces for the first time in 2026 cannot evaluate "99.4 % fidelity" without a comparison point.

Milestones supply the comparison point alongside. "Stable plasma 1,066 seconds" lands in a tokamak track immediately: the previous milestone stood at 403 seconds, the next target is 5,000. From which follows: measurable progress, but still far from commercially relevant. A headline without a track does not say that.

The second effect: tracks age. If the target year passes without a status update to done, the milestone automatically switches to failed/delayed. The programme is not "still ambitious" — it has concretely fallen behind, and that becomes visible. Such failed tracks are a source for the Hype Cemetery: high hype at launch, low reality at the target date.

How a track gets created

The tracker runs every Sunday at 18 UTC. A worker agent scans the week's new articles, identifies multi-sourced claims with cleanly stated targets, and writes them as proposals into a pending table. Cost telemetry and token counts are logged per run.

Proposals are not public. They land in the admin cockpit at ${ADMIN_PATH}/breakthrough/proposals and wait for human approval. A reviewer checks: is the target cleanly stated? Does at least one source sit outside the vendor's marketing surface? Is the target year realistic enough to be falsifiable?

The approval window runs Sunday 18 UTC to Monday 08 UTC — deliberately stretched over 14 hours so approval does not happen under newsletter pressure. Anyone who misses Sunday can still admit the track later. What gets approved is referenced in the same week's newsletter and goes live on the tracker page.

Three running tracks

Otoferlin gene therapy (biotech). Phase III done, trial closes 2026, FDA submission target 2027, approval target 2028, scale-up target 2030. Current status: two milestones done, one in_progress, two queued. Reality score of the lead article: 84.

SPARC plasma energy (fusion). First plasma done 2025, Q>1 in_progress, net wall-plug target 2030. Reality score 74; hype score 31. The programme is taken seriously, but the last milestone is far.

Open-weight frontier match (AI). 7B distillation matches math-eval done 2026 (with caveats), coding benchmark target 2027, agentic benchmark target 2028. Reality 46 — the benchmark methodology is contested, more in the linked article.

All three tracks live on the tracker page with the full milestone list and sources.

Suggest a track

If you know a field that is missing from the platform: submit a track proposal via repository issue or contact form. Requirements are a cleanly stated end goal, at least three meaningful intermediate milestones, and sources that do not exclusively come from the lead vendor's marketing material.

We do not accept everything. Tracks without a falsifiable target ("disruptive AI in 5 years") land in a rejected pool with reasoning — transparency about rejections matters as much as transparency about acceptances. Anyone who counters the rejection with justification can resubmit in a revised form.

The methodology page describes the formula side. The community direction page shows which fields currently pull enough attention to justify a new track.