Pipeline

No runs yet

A one-time scan of every parliamentary business from the past 10 years. Each bill goes through the same filters as our daily pipeline so we can build a complete history of crypto-related activity in the Swiss Parliament.

No pipeline runs recorded yet
Step 1
Fetch

Download every parliamentary business filed over the past 10 years from the Swiss Parliament’s open data feed.

not yet run
Not run
Step 2
Dedupe

Remove duplicates: each bill exists in German, French, and Italian, plus anything we already had on file.

not yet run
Not run
Step 3
Keyword Match

Keep only bills that mention a crypto-related term (Bitcoin, blockchain, virtual currency, and similar).

not yet run
Not run
Step 4
LLM Relevance

An AI model strips false positives, e.g. "encryption" in IT-security bills, or "stable currency" referring to the Swiss franc.

not yet run
Not run
Step 5
Centrality Gate

An AI model judges how central crypto is to each bill: main subject, specific provision, mentioned in passing, or unrelated coincidence. Coincidental hits are dropped.

not yet run
Not run
Step 6
Topic Tag

An AI model sorts each bill into one policy area: tax, regulation, anti-money-laundering, monetary, innovation, consumer protection, energy, or other.

not yet run
Not run
Step 7
Stance Scoring

An AI model judges whether each bill is supportive, neutral, or hostile toward crypto. Bills where crypto is only mentioned in passing are kept neutral.

not yet run
Not run
Step 8
Human Review

A curator reviews every candidate by hand before it appears on the public site. Nothing is published automatically.

not yet run
Not run
Step 9
Author Resolve

Link each approved bill back to the politicians who authored or co-signed it, using Parliament’s official records.

not yet run
Not run
Step 10
Politician Recompute

Recalculate each affected politician’s crypto-stance score from their authorship and final-vote record.

not yet run
Not run
Step 11
Party Aggregate

Roll the updated politician scores into a stance breakdown for each party.

not yet run
Not run

Transparency

The exact rules and prompts that drive Step 3 (Keyword Match), Step 4 (LLM Relevance), Step 5 (Centrality Gate), and Step 6 (Stance Scoring).

Step 3 · Keyword Match

A business passes if any high-tier keyword hits, OR ≥2 medium-tier keywords, OR 1 medium + 1 low, OR ≥2 low-tier keywords. Match is substring (lowercase) across title + reason text + tags.

Strong (16)
bitcoinblockchainkryptowährungcryptocurrencykryptowährungendigital währungdigitalwährungcbdcdigitalfrankene-frankencentral bank digital currencydigitaler frankenkryptoripplecardanotether
Borderline (14)
digital assettokenwalletminingcryptoethereumstablecoindefinftaltcoinsmart contractdistributed ledgerpeer-to-peerhashrate

Step 4 · LLM Relevance prompt

Model gpt-4o-mini·Temperature 0·30 lines

Strips false positives where keywords appear incidentally, e.g. encryption in IT-security, stable as in stable currency.

Step 5 · Centrality Gate prompt

Model gpt-4o-mini·Temperature 0·17 lines

Tags each bill primary / targeted / listed / incidental. Drops incidental rows. Listed bills (crypto in a buzzword list, no specific provision) are forced to neutral by the next step. Targeted bills (crypto-specific provision in a broader law, e.g. Stempelabgaben extension to Krypto-Handel) keep their stance.

Step 6 · Stance Scoring prompt

Model gpt-4o-mini·Temperature 0.1·13 lines

Classifies each candidate as supportive, neutral, hostile, or undefined.