Test flightNew, and still being tuned
Three intelligence briefs covering biotech, AI and technology, and global affairs. Key claims linked to sources, counterarguments made explicit, forecasts tracked publicly, and corrections recorded.
Choose your briefs See the sources
Signal > NoiseImplications > HeadlinesPrimary sources > Commentary
The record so farEvery published edition · to Tue 10.20.2026
We show a Brier score once a brief has 20 graded forecasts. That's a floor for showing a number, not proof of accuracy or calibration: early scores can swing a lot, and a few dozen forecasts say little about the next one. A Brier score measures how close forecasts came to what happened: for each one, take the probability we gave, subtract the outcome (1 if it happened, 0 if not) and square it, then average. Lower is better. Zero is perfect, and calling every forecast 50% scores 0.25. Every number links to the edition or ledger behind it. Track records: Biotech, AI, Worldview.
01 Why read a brief written by AI?
Because you can check it.
It shows where it looked
Registries, filings, dockets, journals and data releases, linked and dated, so you can read what the brief read.
It shows its work
Key claims carry a label and a source: the document, the passage and a verdict. Partial and open verdicts get printed too.
It keeps score
Forecasts carry odds and a date, and get graded in public. Misses come first.
How it's made An AI system and workflow using expert agents writes every brief, working from editorial rules that set what to cover, which sources to trust and how to check a fact. An editor approves every edition before it's sent. Approval isn't a line-by-line check, so mistakes will get through. That's what the sources are for.
Why we built it The Architecture of Signal, the essay on holding machine-assisted analysis to its sources.
Why the name The name dates to 2020, when a few friends got talking about the causes they cared about. It's a club because you can ask, correct and check the record. Read why
02 Proof of work
Straight from the latest briefs.
One matched claim from each brief, as published. The record above counts every verdict, partial and open ones included, and each row links to its full table.
Claim Check sets a claim beside the evidence, such as the endpoint in the trial registry, and records what the evidence supports and what's still unresolved. A verdict covers that one claim. It isn't a quality score for the study or proof that a treatment works.
Biotech & Pharma IntelligenceEdition 025 · 10.20
- Registry says
- YMRS change, baseline to week 3. N=253. Primary completion 08.05.2026.
- Headline says
- 3.1-point gap vs placebo at week 3 (p=0.045), N=253
Match
AI & Technology IntelligenceEdition 025 · 10.20
- Primary document says
- Revenue $18.5B to $19.5B vs $14.5B to $15.5B guide; gross margin 12.0% to 12.5%; unaudited
- Headline says
- Big beat on AI demand
Match
Global Citizen & Worldview IntelligenceEdition 004 · 10.17
- Claim
- 9,277 confirmed Bundibugyo cases and 4,478 deaths in DR Congo
- Source
- ECDC, 10.16
- The passage
- "reported a total of 9 277 confirmed cases, including 4 478 related deaths"
Match
The proposition
Governments should freeze evictions of vulnerable tenants who have nowhere else to go until the housing shortage eases, as Spain's rejected decree would have done until 2030.
Blue team defends · Kautilya, Arthashastra
A state that lets its people fall into ruin weakens itself. An 87-year-old evicted with nowhere to go is the kind of visible ruin a competent state prevents.
Best evidence: in 2024, 28.1% of Spain's market-rate renters spent over 40% of disposable income on housing, against 19.2% across the EU. Fact
Red team attacks · Hayek, The Use of Knowledge in Society
Prices carry knowledge no planner has; when the state overrides them, people respond in ways it can't foresee. A four-year freeze tells owners to rent less.
Best evidence: in San Francisco, landlords whose buildings came under rent control cut the rental housing they supplied by 15% (American Economic Review, 2019). Fact
Where they really disagree Which margin counts: the tenant in front of us now, or the future tenant who never finds a flat. We don't pick a winner.
03 The shelf
Or read the whole thing.
The latest edition of each brief, as published. Subscribers get every past edition on the shelf, searchable by title.
Open the shelfEvery edition, for subscribers. Your link is in any Mind Club email.
Biotech Signal · New for Biotech & Pharma readers
Go beyond the brief.
Biotech Signal is a companion dashboard for biotech subscribers, covering trial catalysts, capital activity, therapeutic-area trends, and forecast records. Access is included with either biotech subscription option during the test flight.
- Trial catalysts, graded A to F
- Capital activity by area and deal type
- Therapeutic-area and target trends
- Forecast records, and how each one resolved
Choose Daily or Weekly Already subscribed? Open Biotech Signal
Biotech SignalSample
Signal index
58.0
Therapeutic-area momentum
Trial catalysts
APhase 3 vaccine meets its co-primary endpoints
BFrontline lymphoma bispecific, interim PFS
DPhase 3 lung cancer drug misses central-review PFS
Catalysts · capital · forecasts · every row sourced
Sample · information only, not investment advice
Ask Mind Club
Ask a question. We answer one a week, with sources.
Something in a brief you want unpacked, or a question you think we should be asking? Send it in. We pick one each week and answer it in the brief.