Key Takeaways

  • Most software pipelines average conflicting data feeds into a single consensus number, which hides real market risk.
  • Moshi founder Anekha Sokhal watches direct social feeds from LNG facilities because operators post operational status updates there faster than wire services.
  • When Reuters, Bloomberg, and The Wall Street Journal report conflicting numbers on pipeline flows or cargo ships, Moshi treats the split itself as a tradeable data point.
  • Instead of stripping out outliers, Moshi pairs single claims with a confidence score and maps the surrounding conversation to show where certainty breaks down.

Why Consensus Models Destroy Trading Signals

Most data pipelines treat conflicting information as noise. If one source says twelve cargo ships passed through a canal and another says eight, an engineering team usually averages the numbers or discards the outlier. That approach works for web analytics, but it ruins physical commodity trading.

In physical commodities, real money moves when the market cannot agree on physical reality. Anekha Sokhal traded commodities before founding Moshi. As she told swyx at AI Engineer NYC, forcing a consensus strips out the exact information traders need to price risk. If reporters at Reuters, Bloomberg, and The Wall Street Journal look at the same transit corridor and report contradictory figures, that disagreement proves that nobody actually knows the physical supply status.

“If Reuters is saying something or Bloomberg is saying something and Wall Street Journal is saying something they may not all agree,” Sokhal noted. “But the fact that they don't agree is a point of data because is there an opportunity there because no one knows.” The spread between those reports represents genuine uncertainty, and uncertainty creates price swings.

Operator Feeds Beat Mainstream Financial Wires

The fastest operational intelligence rarely starts on terminal feeds. It starts with engineers and facility managers on the ground. Sokhal observed that operators of liquefied natural gas facilities frequently post updates on Twitter before public relations teams issue statements.

“They'll be the suppliers of commodities, they'll be the LNG, all of these facilities who will be active on there because it's the quickest way for them to also give a status update,” Sokhal explained. Traders identify specific individuals who run physical terminals and track their accounts directly.

Moshi builds around this behavior rather than replacing it with an automated black box. When an operator posts an update about a plant outage or a shipping delay, Moshi displays the claim alongside context from other sources. “When we are showing it to someone, we give a confidence score of okay, someone is saying this one thing, but this is everything else happening around the conversation,” Sokhal said. Users keep the raw, biased, early post, but they see the surrounding skepticism or corroboration at a glance.

What to Do With This

Audit your data pipeline to see where your models take an average of conflicting external sources. Replace that logic by logging the variance between those sources as a distinct metric, and alert your team whenever that spread hits a 30-day high.