I spend most evenings pulling the same set of on chain and macro feeds, and the one report I kept skipping was the CFTC Commitment of Traders. This week I finally read it, and it told a cleaner story than any single price chart: the people crowding gold and the people holding bitcoin are in completely different moods.
what the gold book looks like
The latest CFTC legacy futures report (report date 2026-09-01, published 2026-09-07) shows COMEX gold with large speculators net long 228,124 contracts, which is 54.94% of total open interest. That is a stretched reading. For context, open interest in the front gold contract was 415,196 contracts, and commercials (the hedging side: miners, refiners, swap dealers) were net short -264,718 contracts.
When one side of a market is 55% net long relative to open interest, that is not a quiet consensus. That is a bet that has already been placed by a lot of people. Specs did trim about 15,210 contracts week over week, so the crowd is not growing, but it is still very much there.
what the bitcoin book looks like
Now look at CME bitcoin futures in the same report. Large speculators are net long just 703 contracts, or 3.57% of open interest. Commercials are net short a modest -1,077. Both sides are basically flat.
So here is the split I did not expect to see spelled out so cleanly: two assets that both get pitched as hard asset inflation hedges have positioning that could not be more different. Gold specs are near maximum bullish crowding. Bitcoin specs are sitting on their hands.
why the gold crowd is where it is
I cross checked two background signals the same night.
The US Treasury curve snapshot (collected 2026-09-07) had the 10 year at 4.77% and the 2s10s spread at +0.41bp. The curve is flat but not inverted, flagged as "normal/steep, soft landing expectation." When the recession fear is not showing up in the curve, the excuse to hold gold as a hedge weakens, yet the momentum crowd keeps the long on.
The FX synthesis (synthesis time 2026-09-07) put the USD basket index at 0.6671, with USD/JPY at 156.18 and USD/CNY at 6.7194. A soft dollar is the other tailwind behind gold. Cheap dollars and a flat curve are a friendly backdrop for a crowded gold long.
Bitcoin, by contrast, is not getting that macro push from this dataset. Its COT is neutral, and the read says nothing about the much larger offshore perp book.
how to pull it yourself
The COT feed is public on the AgentDataHub catalog. A bare request returns JSON:
curl -s "https://agentdatum.com/api/v1/d/cftc-cot" \
| python3 -c "import sys,json; d=json.load(sys.stdin)['data']; g=[r for r in d['data'] if r['contract']=='GOLD'][0]; print('spec_net_pct_oi',g['large_spec_net_pct_oi'],'commercial_net',int(g['commercial_net']))"
That prints spec_net_pct_oi 54.94 commercial_net -264718 on the current report. Swap GOLD for BITCOIN to see the neutral side.
the honest part
I am not calling a top on gold. A few things keep this from being a clean trade signal.
First, COT is weekly and already stale. The report is dated 2026-09-01 but only published days later, so you are always looking in a rearview mirror.
Second, a negative commercial net position in gold is structural, not a secret sell signal. Miners and refiners hedge continuously, so commercials are almost always net short. The real extreme signal is the 54.94% spec pct_oi, not the commercial sign.
Third, CME bitcoin futures are a thin slice of total bitcoin positioning. The COT shows 19,697 contracts of open interest, while offshore perpetual books carry far more. A neutral COT bitcoin reading does not speak for the whole market.
Fourth, these are Legacy Futures Only numbers and exclude options. And a crowded long can stay crowded for a long time. Gold sat above 50% net long through most of a multi year run without immediately reversing.
So the takeaway for me is not "gold will fall." It is that the crowd in gold is committed and the crowd in bitcoin is not, and those two moods rarely line up this cleanly. When they do, it is worth writing down.
This post links to the live data catalog: https://agentdatum.com/.well-known/ai-catalog.json (213 sources, refreshed continuously).
This article was written with AI assistance for drafting and data formatting.
Top comments (1)
Thanks for sharing! Free tools are invaluable for developers. API aggregator platforms that organize services by category are worth exploring. Keep the great content coming! 💡